Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Optical measurement of tissue perfusion changes as an alternative to electrocardiography for heart rate monitoring in Atlantic salmon (Salmo salar)

Optical measurement of tissue perfusion changes as an alternative to electrocardiography for... Background: Welfare challenges in salmon farming highlights the need to improve understanding of the fish’s response to its environment and rearing operations. This can be achieved by monitoring physiological responses such as heart rate (HR) for individual fish. Existing solutions for heart rate monitoring are typically based on Electrocardiog- raphy (ECG) which is sensitive to placement and electrode orientation. These factors are difficult to control and affects the reliability of the principle, prompting the desire to find an alternative to ECG for heart rate monitoring in fish. This study was aimed at adapting an optical photoplethysmography (PPG) sensor for this purpose. An embedded sensor unit measuring both PPG and ECG was developed and tested using anesthetized Atlantic salmon in a series of in-vivo experiments. HR was derived from PPG and compared to the ECG baseline to evaluate its efficacy in estimating heart rate. Results: The results show that PPG HR was estimated with an accuracy of 0.7 ± 1.0% for 660 nm and 1.1 ± 1.2% for 880 nm wavelengths, respectively, relative to the ECG HR baseline. The results also indicate that PPG should be meas- ured in the anterior part of the peritoneal cavity in the direction of the heart. Conclusion: A PPG/ECG module was successfully adapted to measure both ECG and PPG in-vivo for anesthetized Atlantic salmon. Using ECG as baseline, PPG analysis results show that that HR can be accurately estimated from PPG. Thus, PPG has the potential to become an alternative to ECG HR measurements in fish. Keywords: Salmo salar, Heart rate, Implant, Biosensors, Photoplethysmography management and operation of such salmon farms entails Background a broad range of interrelated operations exerting convo- Atlantic salmon (Salmo salar) is one of the most impor- luted effects on the fish. To ensure acceptable fish health tant species in fish farming with more than 2.6 Mt pro - and welfare conditions during production, relevant data duced globally in 2019 [1]. A typical salmon production to describe these must be collected and evaluated in site consists of 8–10 flexible sea cages usually 50  m in conjunction with operational data. Such evaluations are diameter that holds a volume of around 40,000 m . The largely subjective and experience based, and are carried out as part of the daily inspection and feeding routines. *Correspondence: eirik.svendsen@ntnu.no 1 About 15% of all farmed salmon are lost during pro- Department of Engineering Cybernetics, NTNU, O.S. Bragstads Plass 2D, 7034 Trondheim, Norway duction, a loss partly attributed to the lack of objective Full list of author information is available at the end of the article input to production control [2]. To address this loss, © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 2 of 12 recent research efforts have focused on determining Changes in treatment and/or living conditions chal- what constitutes good fish welfare and which indicators lenges an animal’s homeostasis and may be expressed as one should quantify (i.e., measure) to obtain a more various stress responses [18]. Recently, heart rate meas- objective foundation for making welfare critical deci- ured using DSTs was demonstrated as being linked to sions [3], and safeguard ethically sound production. welfare [19] and stress [20] in Atlantic salmon and in Sensing such indicators should be done using as unob- other salmonids such as rainbow trout [21]. Heart rate trusive measures as possible to avoid compromising may, therefore, serve as a proxy for stress, thereby provid- safe and efficient operations. However, this is in conflict ing information on the fish’s welfare in aquaculture pro - with the challenge of collecting data from a biomass vided that heart rate can be obtained in real time. Heart contained in a volume of roughly 40,000 m of water, rate estimates for fish are most commonly derived from as this calls for distributed or mobile data collection an electrocardiogram (ECG) measured by electrodes systems which are based on, e.g., environmental sens- integrated in the tag’s encapsulation. However, the reli- ing networks, sonars/echo sounders, passive acoustic ability of this method has been shown to be sensitive to monitoring and different types of optical methods such tag deployment as the derived heart rate depends on the as underwater cameras [4]. Such systems are likely to lateral placement in the fish as well as the tag’s orienta - be intrusive on the production, and may thus have to be tion [21]. Exploring other potentially more robust meas- removed from the aquaculture cage before operations urement principles for their ability to sense heart rate are carried out to reduce the likelihood of, e.g., equip- is, therefore, desirable, as it might contribute to making ment damage. Furthermore, such technologies provide heart rate data more reliable and feasible to obtain for data with different spatial and temporal resolution and free swimming fish in aquaculture. provide data on a population level since the recorded Photoplethysmography is an optical sensing technique data are obtained from a sub-volume in the sea cage quantifying changes in tissue perfusion (i.e., blood vol- rather than from a specific group of fish. Sonars and ume) by optical absorption. The photoplethysmogram cameras can provide data on a group level if smaller (PPG) is a convolution of many components. In humans, parts of the biomass are present within the sensor’s slow-varying components may arise from breathing [22], field of view, a feature that has been used to estimate, while the superimposed pulsatile component changes e.g., fish size and cage biomass [5 ]. Although group with the cardiac cycle [23]. Using the pulsatile compo- and population level data can provide a certain insight nent from several wavelengths of light enables estima- into the dynamics in animal husbandry operations, the tion of arterial blood oxygen saturation (SpO2) provided sensing technologies’ intrusiveness on operations and a mapping function compensating for tissue scattering limited understanding of the measurements’ link to effects is known [24]. Independent from this, the pulsatile welfare imply that alternative or supplemental solutions component can be used to obtain a heart rate estimate for welfare evaluations are needed. as shown for both humans and other mammals [25, 26]. Individual level data provides the highest possible data u Th s, PPG is considered a likely candidate to address the resolution with respect to biomass, a feature recognized challenges associated with ECG obtained using implants in precision farming both on land [6] and at sea [4]. In provided a suitable sensing solution can be designed. aquaculture, the main method for obtaining individual Implantable PPG solutions have been developed for, level data is using miniaturized, encapsulated electronic e.g., sheep [27] and other mammals [28]. Although systems commonly referred to as “tags”. Typical parame- potentially relevant, such solutions are proprietary and ters that can be measured in an operational setting today either require additional, external interfaces for power are activity and swimming depth [7], heart rate (HR) [8], and data collection/processing, or depend on radio- swimming speed [9] and position [10, 11]. Tags are avail- based data transfer, making them infeasible for use in able in two main types: Data storage tags (DSTs) and fish in seawater. Measurement of PPG has been dem - telemetry transmitter tags [12], and are usually surgically onstrated for aquatic animals such as zebrafish (Danio implanted into the fish’s peritoneal cavity. rerio) using imaging PPG [29]. This is a suitable tech - Tagging has been used successfully to study fish in dif - nique when tissues with low opacity can be remotely ferent situations ranging from tracking fish in rivers and imaged in a controlled environment. This is the case sea [13–15] to shedding light on fish behaviour in sea for zebrafish in the larval stage, but not for Atlantic cages during aquaculture operations [7, 16]. Thus, tagged salmon in a context where using implants is consid- individuals may act as representatives (i.e., “sentinel fish”) ered relevant. Correspondingly, obtaining heart rate for the rest of the biomass in an aquaculture cage [17] from PPG has been demonstrated by drilling holes and to facilitate safer and less operationally intrusive welfare inserting PPG sensors through the shell of the Medi- evaluations. terranean mussel (Mytilus galloprovincialis) [30]. This S vendsen et al. Anim Biotelemetry (2021) 9:41 Page 3 of 12 approach requires stationary subjects as the sensors are Methods wired to external hardware for power and data logging. Sensing and logging equipment These approaches are, thus, not directly suited for fish The MAX86150 optical biosensor module [33] from implants, but demonstrate the feasibility of applying Maxim Integrated (San Jose, USA) that offers both PPG PPG for heart rate sensing for different aquatic animal and ECG output in a miniature package (3.3 × 5.6 × mm) groups. with ultra-low power consumption, was selected for the In recent years a market desire for mobile medi- measurements in this study. The physical and electrical cal equipment and consumer products containing properties of the module make it suitable for integration such sensors (e.g., smart phones and sports watches as the biosensor element of a prospective DST or acoustic [31, 32]) has driven development and miniaturiza- tag with such sensing capability. A printed circuit board tion of PPG sensors. Such innovations have increased (PCB) was designed as a platform for miniaturizing the the potential to include such off-the-shelf sensors in reference circuit design for the MAX86150. The PCB was implants for fish, as this is an application with simi- subsequently cast in a cylindrical epoxy casing such that lar requirements with respect to size and power con- the sensing direction was perpendicular to the cylinder’s sumption. However, the anatomical and optical tissue longitudinal axis. Seven thin electrical wires protruded property differences between mammals and fishes, the rear of the cast and was used to extend the ECG raise the question of whether PPG captured in Atlan- electrode connections of the MAX86150, and to enable tic salmon using the common intraperitoneal implan- power connection and data transfer from the sensor. tation approach results in useful data. The purpose A battery powered MAX32630FTHR microcontroller of this study was, therefore, to adapt an off-the-shelf board was used to relay data to a PC using Bluetooth. PPG/ECG biosensing module for in-vivo testing in The cast also included a threaded aluminium end piece Atlantic salmon to investigate its potential to provide for attachment to a fixture rod. The fixture rod allowed an alternative heart rate estimate from Atlantic salmon horizontal and vertical sensor orientations as well as that does not suffer from the same limitations as ECG, rotation of the sensor around its longitudinal axis. The while facilitating a future low-cost implant based on complete assembly is shown in Fig. 1. Maxim DeviceStu- standard electronic components. dio (V 5.3.03289.0) was installed on a Dell Latitude 7490 Fig. 1 A (left): 1. Fixture rod, 2. Wire extensions for MAX86150 ECG electrodes, power connection and data transfer, 3. Threaded aluminium end piece, 4. MAX86150 biosensor module. B (right): 1. Epoxy cast with MAX86150 biosensor module, 2. ECG electrode extension leads, 3. Data and power supply extension leads, 4. MAX32630FTHR microcontroller board for wireless data transfer to PC Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 4 of 12 laptop computer and used for real-time data inspection and logging. To determine the biosensor’s feasibility when applied to Atlantic salmon, all sensor settings were kept the same for all data sets (sampling frequency fs = 200  Hz, RED (660  nm) and IR (880  nm) LED pulse amplitudes A = A = 20 mA, LED pulse width PW = 400  µs, and 660 880 no sample averaging), thus enabling signal quality com- parisons across data sets. Note that when using PPG to estimate heart rate only, no reference measurements apart from the ECG baseline used for HR comparisons are required. Experimental procedures Individual fish were captured from a holding tank using a knotless dip net and immediately transferred to a tank containing water from the same water supply and anes- −1 thetic in a knock-out solution (70  mg  l Benzoak Vet). The tank was covered using a Styrofoam plate after the fish was placed inside. Using a wheeled trolley, the tank was then moved to the indoor experimental location. When deemed to have reached level 3 anesthesia [34], the fish was placed in a specialized surgical table. By Fig. 2 A (top): Rotation angles for the horizontal sensor orientation. inserting a hose connected to a small pump into the fish’ B (bottom): Rotation angles for the vertical sensor orientation. buccal cavity, continuously aerated water with mainte- Illustration by Mats Mulelid, SINTEF Ocean AS −1 nance anesthetic (35  mg  l Benzoak Vet) was used to irrigate the gills and keep the fish sedated. The wire elec - trodes for ECG measurement attached to the MAX86150 were inserted into the muscle on the left and right lateral sides close to the heart. A 1–2  cm incision was made in the abdominal wall along the sagittal plane just anterior of the pelvic fins (position 1, Fig.  2A). For position 1, the sensor was inserted horizontally into the peritoneal cav- ity and one data set collected for orientations of 0° (ven- tral), 90° (right lateral), 180° (dorsal), 270° (left lateral) and 0° (ventral repeated) degrees rotation, respectively (Fig.  2A). Following data collection in position 1, a sec- ond 1 cm incision was made in the abdominal wall along the sagittal plane just posterior from the transverse sep- tum (position 2, Fig.  2B). For position 2, the sensor was inserted vertically into the peritoneal cavity and one data set collected for orientations of 0° (anterior), 90° (right lateral), 180 degree (posterior), 270 degree (left lateral) Fig. 3 Data collection and 0° (anterior repeated) degrees, respectively (see Figs.  2B and 3). Data was collected for 1  min for both positions and all orientations for each individual (10 data sets per fish / 50 data sets in total), after which the fish preventing valid ECG signals. Data collection was suc- −1 was euthanized by an anesthetic overdose (> 70  mg  l cessful from the remaining 6 fish (from now on referred Benzoak Vet) and exsanguination. to as Fish 1 through 6), resulting in 10 data sets for fish A total of 7 fish were used for the experiment. No 1, 2, 4, 5 and 6, and 6 data sets for fish 3 which suffered valid data was obtained for the first fish due to electri - cardiac arrest after 6 data sets. The analyses were thus cal interference likely caused by a nearby water pump based on 56 data sets in total. S vendsen et al. Anim Biotelemetry (2021) 9:41 Page 5 of 12 Data processing QRS peaks were 40% higher than neighbouring peaks. Valid 20  s data subsets were manually selected and HR was then determined by the average time difference labelled based on the data evaluation criteria presented (∆t ) between peaks using: avg by Elgendi [35]: HR = · 60 BPM (1) avg • Salient ECG baseline. • Salient PPG signal for one or both wavelengths. Because raw PPG data containing both a stochastic • Similar PPG waveform morphology throughout the trend and measurement noise was logged, a preprocess- entire subset. ing step to remove these was necessary. The slow-varying (i.e., low-frequency) trend was removed using a third Data analysis using Python 3.8 (Anaconda Inc., Aus- order Butterworth highpass filter [38] with a cutoff fre - tin, Texas, USA) could then be conducted for valid data quency of 0.25  Hz. Measurement noise was reduced sets labelled”GOOD” (accept) or”FAIR” (accept). Data using a second-order Savitzky-Golay filter [39] with a 50 sets labelled “BAD” (reject) were judged to not contain sample (i.e., 0.25 s) window. This window size was chosen the required information for signal quality calculations to retain as much information in the signal as possible and were, therefore, omitted from further analysis. A based on the lowest expected HR of 15 BPM for Atlan- complete overview of data sets and corresponding wave- tic Salmon [37] and comparable species [21]. An example lengths considered valid based on these criteria is given illustrating the result of these processing steps is given in Table  1, while representative samples of PPG data in Fig. 5. HR from PPG was then calculated from the de- sets labelled “GOOD”, “FAIR” and “BAD” can be seen in trended and noise suppressed signal using autocorrela- Fig. 4. tion defined by: The default filtered ECG signal output from the MAX86150 was used to find HR by first scaling the fil - y(m) = X(n)X(n + m) tered output to [0, 1] and then running a peak detection (2) n=1 algorithm. Scaling was done so the same peak detec- tion parameters for all ECG time series could be used. where N is the number of samples in the signal X, and Peaks were identified using the “find peaks()” method in m ∈ [0, N − 1]. By applying the same peak detection Python’s statistics module utilizing both a time window method as that used for ECG, the index of the domi- (i.e., the minimum required distance between peaks) and nating peak (i.e., the first peak) in the autocorrelation a prominence limit to identify the QRS peaks (i.e., ventri- series was identified. Because the autocorrelation series cle depolarization) [36]. The time window was set to 160 featured softer (but not flat/diffuse) peaks compared samples (i.e., 0.8 s) based on a maximum expected HR of to the ECG signal, the prominence setting in the peak 80 BPM for Atlantic salmon [37] and comparable species detection algorithm was relaxed to 0.3 to avoid too [21]. Prominence was set to 0.4, thus demanding that the Table 1 Summary of data set evaluations for both orientations, “Or.”, (Hor = horizontal and Ver = vertical) and all rotations, “Rot.”, (see Fig. 2) for all fish Or. Rot. Fish1 Fish2 Fish3 Fish4 Fish5 Fish6 Hor 1 0 NVD NVD λ2 NVD λ2 NVD Hor 90 NVD NVD λ2 NVD NVD NVD Hor 180 NVD NVD λ1, λ2 NVD λ2 NVD Hor 270 NVD NVD λ1, λ2 NVD λ1, λ2 NVD Hor 2 0 NVD NVD λ1, λ2 NVD NVD NVD Ver 1 0 λ1, λ2 λ1, λ2 λ1, λ2 λ1, λ2 λ1, λ2 NVD Ver 90 λ1, λ2 λ1, λ2 N/A λ1, λ2 NVD λ1, λ2 Ver 180 λ1, λ2 NVD N/A NVD NVD NVD Ver 270 λ1, λ2 λ1, λ2 N/A λ1, λ2 λ1, λ2 λ1, λ2 Ver 2 0 λ1, λ2 λ1, λ2 N/A λ1, λ2 λ1, λ2 λ1, λ2 NVD signifies “No Valid Data”, i.e., no part of the time series fulfils all selection requirements. λ1 and λ2 represents valid data for 660 nm and 880 nm, respectively “N/A” (Not Applicable) denotes the data series which could not be collected for Fish 3 due to cardiac arrest. Note that the 0 degrees measurement has been repeated for both orientations Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 6 of 12 Fig. 4 Signal examples Fig. 5 Top: Raw data. Middle: De-trended and smoothed data. The red and blue dots illustrate the detected peaks and valleys, respectively. Bottom: Scaled ECG. The red dots illustrate the detected ECG peaks aggressive peak rejection. As for ECG HR, the time calculate PPG HR in beats per minute (HRBPM) using window was set to 160 samples. The index of the identi - Eq. 1. fied peak, therefore, represented the average time delay To evaluate which sensor orientation and rotation gave between peaks in the PPG time series, and was used to the highest quality data, the signal quality index (SQI) for S vendsen et al. Anim Biotelemetry (2021) 9:41 Page 7 of 12 Fig. 6 Signal processing flow the noise suppressed (but not de-trended) PPG signals, was Discussion calculated. SQI is defined for individual PPG pulses by [35]: The high accuracy for HR computed from PPG com - pared to the ECG baseline from our experiments indi- PPG pa cates that PPG is a viable alternative sensor principle for P = · 100 SQI (3) PPG avg monitoring heart rate in Atlantic salmon. Moreover, the MAX86150 unit appears to be a suitable sensor module where PPG is the peak-to-peak PPG pulse amplitude and pa for implementation in future DSTs or transmitter tags PPG its mean value. The PPG signal where only noise avg aiming to measure and report HR over time. Although the suppression had been applied was used to retain access sensor is designed for integration in optical HR measure- to the pulses’ mean value which was removed during de- ment applications for humans and, thus, with human tis- trending. A pulse was defined as the valley to valley noise sues and blood in mind, the HR estimate relies solely on suppressed data subsets. Valley indices (i.e., positions) were the time-varying tissue perfusion. While human and s fi h identified by inverting the de-trended (i.e., mean-centered) blood is different with respect to composition and cell signal and reapplying the peak detection algorithm with the morphology, the essential functionality is the same, as same window and prominence settings as those for auto- is the hemoglobin [40]. The wavelengths emitted by the correlation peak detection. Using the valley indices, the MAX86150 sensor are chosen due to their absorption pulses were extracted from the noise suppressed (but not sensitivity to oxy- and deoxyhemoglobin which deter- de-trended) data sets, and the average SQI (SQI ) for all avg mines the blood’s oxygenation and thus its colour. Because pulses calculated to represent the subset SQI. An example tissue colour is affected by perfusion, processing either or illustrating the peak detection result is given in Fig. 5. both wavelength will be a feasible approach to obtain HR To evaluate the accuracy of the PPG HR estimates, the in both humans and fishes. difference between the ECG baseline and both wave - The data used in this study originates from 6 different lengths separately were calculated in percent using: fish, yielding 28 data sets. Although up to 10 data sets were collected from each fish, the data sets are consid - HR − HR PPG ECG = · 100 (4) HR ered independent because each data set is separate in HR ECG either time, relates to different tissues, or both. When The signal processing flow is illustrated in Fig. 6. reviewing Table  1, most data available for processing came from the anterior part of the fish. This is prob - Results ably because the measurements in this region coincide ECG was labeled “GOOD” in all data sets, while one with locations with a high blood supply such as the or both PPG wavelengths were judged to qualify as liver and gut (thus implying high perfusion) compared “GOOD” or “FAIR” in 28 data sets. Thus, 28 (of 56) to the posterior region where tissues with lower perfu- data sets were omitted from analysis. The results from sion (mostly white muscle tissue) are present. This is the data processing are given in Table  2. The results supported by that the average SQI (Table 2) was highest show an average accuracy of 0.7 ± 1.0% for 660 nm and for locations and orientations associated with the ante- 1.1 ± 1.2% for 880  nm wavelengths relative to the ECG rior part of the fish, especially for the 0 degree rotation HR baseline. The average SQI (SQI column in Table 2) which is towards the heart. This indicates that future avg indicates that the vertical orientation and 0 degrees implementations of PPG sensors for Atlantic salmon rotation (i.e., measuring towards the heart) resulted in should facilitate data collection in this area. the best data quality. Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 8 of 12 Table 2 Results sorted by average SQI where the columns named Fish is the fish number, Or. is the orientation, Rot. is the rotation, HR is the heart rate in BPM calculated from ECG, HR is the calculated heart rate in BPM for the 660 nm time series, HR is the ECG 660 880 calculated heart rate in BPM for the 880 nm time series, ∆HR is the difference in HR estimate between the 660 nm time series and ECG in percent, ∆HR is the difference in HR estimate between the 880 nm time series and ECG in percent, SQI is the SQI for 880 660 660 nm, SQ I is the SQI for 880 nm and SQI is the average SQI, respectively 880 avg Fish Or. Rot. HR HR HR ∆HR ∆HR SQI SQI SQI Likely tissues ECG 660 880 660 880 660 880 avg 5 Ver 270 50.4 50.2 50.4 0.40 0.00 0.03 0.02 0.02 M, F 3 Hor 2 0 53.1 53.3 53.3 0.38 0.38 0.04 0.05 0.05 F 5 Hor 270 50.6 50.4 50.6 0.40 0.00 0.02 0.08 0.05 M, F 5 Hor 1 0 35.9 N/A 35.2 N/A 1.95 N/A 0.08 0.08 F 5 Hor 180 46.0 N/A 46.0 N/A 0.00 N/A 0.10 0.10 F, I 2 Ver 270 43.0 42.7 42.4 0.70 1.40 0.16 0.14 0.15 M, F 5 Ver 1 0 49.0 48.6 48.2 0.82 1.63 0.28 0.06 0.17 L, G 3 Hor 90 41.1 N/A 40.7 N/A 0.97 N/A 0.18 0.18 M, F 3 Hor 1 0 41.1 N/A 40.7 N/A 0.97 N/A 0.18 0.18 F 3 Hor 180 41.7 41.2 41.0 1.20 1.68 0.21 0.21 0.21 F, I 6 Ver 90 27.5 27.3 26.8 0.73 2.55 0.36 0.17 0.26 M, F 3 Hor 270 42.9 42.9 43.5 0.00 1.40 0.48 0.24 0.36 M, F 5 Ver 2 0 52.6 52.6 52.6 0.00 0.00 0.48 0.25 0.36 L, G 4 Ver 270 37.7 37.7 37.9 0.00 0.53 0.81 0.51 0.66 M, F 2 Ver 90 35.0 35.1 35.2 0.29 0.57 0.95 0.74 0.84 M, F 6 Ver 270 36.0 35.8 35.7 0.56 0.83 1.51 0.55 1.03 M, F 2 Ver 2 0 48.6 48.8 48.8 0.41 0.41 0.87 1.59 1.23 L, G 1 Ver 270 18.0 18.2 18.2 1.11 1.11 2.35 0.63 1.49 M, F 1 Ver 90 13.9 13.8 13.8 0.72 0.72 1.95 1.24 1.60 M, F 1 Ver 180 19.0 18.9 18.9 0.53 0.53 2.55 0.69 1.62 G, F 6 Ver 2 0 33.6 33.6 33.6 0.00 0.00 3.27 0.66 1.96 L, G 2 Ver 1 0 32.8 32.0 32.0 2.44 2.44 2.66 1.84 2.25 L, G 4 Ver 2 0 70.2 70.2 70.2 0.00 0.00 2.96 1.61 2.28 L, G 3 Ver 1 0 59.4 59.4 59.4 0.00 0.00 3.75 1.32 2.54 L, G 4 Ver 1 0 32.4 30.7 31.3 5.25 3.40 4.69 1.38 3.04 L, G 1 Ver 1 0 19.4 19.3 19.3 0.52 0.52 13.2 2.90 8.05 L, G 1 Ver 2 0 26.7 26.5 26.5 0.75 0.75 14.59 6.20 10.40 L, G 4 Ver 90 37.2 37.6 39.2 1.08 5.38 24.68 7.92 16.30 M, F The table entry “N/A” (Not Applicable) is for the cases where no valid data could be identified for the respective wavelength in a particular data set. Entries in the Likely tissues column are: M: Muscle, F: Fat, I: Intestines, L: Liver and G: Gut, and denote tissues likely present in the sensing volume for the associated sensor orientation and rotation In our results,  HR was reported using one decimal The range between the lowest (13.9 BPM) and high - because additional decimal points are likely inaccurate. est (70.2 BMP) heart rates may be explained by that This conclusion stems from considering the HR-depend different fish individuals would have had different phys - - iological baselines (e.g., stress levels and health) prior ent quantification error for our highest HR (70.2 BPM). to the experiments. These differences would, thus, have When sampling at 200  Hz, the change in BPM resulting yielded different individual responses to handling and from what is considered the maximum timely offset from anesthesia. Differences in heart rates were, therefore, be true peak position can be calculated. This is achieved by expected. Reported heart rates for Atlantic salmon and first finding beats per second (BPS) which in this case is comparable species are between 15 and 80 BPM [21, 70.2/60 = 1.17 BPS. The time between HR peaks will in 37]. With the exception of one individual (13.9 BPM) this case be ∆t = 1/1.17 = 0.855  s. A timely offset in true all measured heart rates fell within this expected range. peak placement exceeding 50% of the sampling interval The individual having a heart rate below 15 could have implies that a peak will be associated with the previous or been more susceptible to sedation thus explaining this next sampling point. Hence, for our quantification error result, or the reported HR range is conservative. we get Eq = 0.855 + (1/200) · 0.5 = 0.8575 s. This peak to S vendsen et al. Anim Biotelemetry (2021) 9:41 Page 9 of 12 peak distance which includes the 50% offset, then gives a component. This is partly remedied using high pass filter - shifted BPM of BPM = 60/0.8575 = 69.995. The BPM dif - ing in the analyses, as this effectually reduces or removes ference can then be calculated as ∆BPM = 70.2 − 69.995 long term trends and any changes therein. Moreover, an ≈ 0.2 BPM. The corresponding number for our lowest estimate of HR relies solely on the frequency content in baseline HR (13.9 BPM) is 0.008 BPM. the PPG signal measurement signal, and not the ampli- The quantification error considerations are closely tude. Based on these observations, movement of tissues related to our measurements’ sensitivity. By first accept - relative to the sensor during data collection were unlikely ing that a peak cannot be placed “between” two sampling to have had impact on the results. points, the sensitivity can then be evaluated in the same Motion artefacts may also have been caused by tissue way as the quantification error, only using the whole sam - contraction (e.g., the heart) if it was within the sens- pling interval. Thus, the sensitivity can be considered to ing volume during data collection. This is particularly be twice that of the quantification error, i.e., 0.4 BPM per relevant for the vertical orientation with 0° rotation sample offset for 70.2 BPM, and 0.016 BPM per sample (i.e., when the sensor was pointing towards the heart). offset for 13.9 BPM. Because this is of particular concern, a subsequent The deviation from the baseline of 0.7 ± 1.0% for post mortem dissection of Atlantic salmon was done to 660  nm and 1.1 ± 1.2% for 880  nm can be explained by assess the sensing volume for this orientation and rota- different factors. One potentially important source tion. The dissection revealed that the tissue observed of error is that PPG is sensitive to motion artefacts. with the present method was likely dominated by low- Such artefacts can be divided into two types: Perfusion perfusion fatty tissues surrounding larger blood ves- changes in tissue caused by motion and relative motion sels such as the hepatic arteries and veins (Fig.  7) [40]. between the sensor and the sensing volume. The former Fat has a high optical scattering [41] coefficient due to is not considered relevant when evaluating the results lipid droplets inside the fat cells. Due to the size of the because all fish were in level 3 anesthesia (surgical) and, scatters, this scattering is highly forward directed and thus, motionless. However, such artefacts are likely to be almost independent of the wavelength of the light. This important when the method is applied to free-swimming implies that the light from the sensor is strongly scat- non-sedated fish. A logical next step on the path towards tered by the tissue while the intensity decays exponen- an operational measurement method would, therefore, tially with distance from the light source in accordance be to apply the sensor to fish exhibiting normal swim - with the (modified) Beer–Lambert law [42], thus lim - ming behaviour, and collect concurrent PPG and motion iting the distance light travels. This is indicated by the data to assess the potential impact of specific motion fact that the SQI for both wavelengths was generally low patterns. due to the big difference between the pulsatile PPG com - To minimize the effect of the latter, the setup was ponent’s amplitude and the signal mean. A large mean designed to be as rigid as possible to ensure a stable value implies that a lot of light is scattered back to the sensing environment during data collection (Figs.  1 and receiver without having penetrated far into the tissue. It 3). However, although the setup was mechanically sta- is, therefore, likely that the data originates from tissues ble and the fish in level 3 anesthesia, motion artifacts close to the light source and that the heart is not part of caused by potential tissue movement such as peristalsis the sensing volume. [40] may have caused transient changes in the trend and Accuracy may also have been affected by the physiolog - potential changes in the amplitude of the PPG’s pulsatile ical state of the fish. The fish used in the experiment were Fig. 7 A (left): 1. Epigastric vein/artery, 2. Pylorus caeca/lipid deposits, 3. Liver. B (right) 1. Septum transversum, 2. Hepatic arteries/veins Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 10 of 12 lab grown and showed no signs of deteriorated health. Autocorrelation was chosen for PPG analysis because The anesthesia had both an analgesic and a paralyzing its low computational demand and robustness against effect. It must, therefore, be expected that the secondary potential transient motion artifacts makes it a likely can- circulation system driven by the caudal heart and move- didate for implementation in a microcontroller suitable ment was impaired during data collection. In addition, for integration in a size and energy constrained fish tag. the fish underwent a surgical procedure and the incisions However, it cannot be ruled out that other, more compu- as well as the sensor insertion may have further disturbed tationally demanding approaches such as singular spec- parts of the circulation system. When reviewing Table  2 trum analysis [44], wavelet analysis [45] or a fast Fourier for fish where data for both iterations of the horizontal transform [46] approach could have given better results. and vertical 0° are available, similar SQIs for both itera- The perfusion index (Eq.  3) was used to evaluate sig- tions appear to be the trend. Although this indicates that nal quality because this is considered the “gold standard” the physiological state of the fish remained stable during for PPG signal quality evaluation, even though alterna- data collection, it is likely that this state differs from that tive methods (e.g., skewness, kurtosis and entropy) [35] a fully awake and moving fish would exhibit. This under - which might lead to better results, exist. Such methods, lines the necessity of conducting further experiments however, are derived using reference data readily avail- with fish exhibiting more normal behaviours and physi - able for humans. To the authors’ knowledge, no reference ological function. PPG data for fish exist for comparison. Such data would The peak detection procedure may have affected the be a very useful resource in developing new methods for accuracy of the method because the detrended PPG sig- quality evaluations of PPG data from fish, particularly if nals consisted of an oscillating curve with wide peaks aspiring to quantify SpO2. compared to the ECG peaks. The accuracy in PPG peak Overall, valid data was identified for all orientations detection was, therefore, lower, thus resulting in slight and rotations. Although certain combinations of orien- differences in the intervals between the detected PPG tation and rotation yielded fewer data sets fulfilling the peaks and their corresponding ECG peaks. For long subset selection criteria than others, this does not nec- time series containing many peaks, such differences are essarily mean that data are harder to obtain for these expected to cancel out, but this may not have been the orientations and rotations. The same low-energy output case for 20 s data sets. This may have been further exac - settings were used for both orientations and all rota- erbated by the fact that the various orientations and rota- tions, thus implying that more valid data could have been tions would have illuminated tissues and capillary beds obtained if sensor settings, such as output power, had supplied by haemal arches connected to different points been increased. PPG, therefore, has the potential of being along the dorsal artery. The resulting differences in pulse a robust alternative to ECG for HR measurement in fish. transit times (PTT, i.e., times between the heart beat and when it is observable in the sensing volume), may have Conclusions shifted the PPG in relation to the ECG [43]. Further- A PPG/ECG module has been successfully adapted to more, if different capillary beds with different PTTs were measure both ECG and PPG in-vivo for anesthetized present in the sensing volume it could have distorted Atlantic salmon. Using ECG as baseline, results from the PPG pulses by dragging them out in time, thereby an analysis of the PPG signals show that that HR can be explaining the differences in morphology seen between accurately estimated from the PPG measurement, thus the “GOOD” and “FAIR” pulse examples in Fig. 4. having the potential to become an alternative to ECG HR The criteria used for the selection of valid data subsets measurements in fish. for analyses are subject to interpretation as highlighted Based on the encouraging results from this experiment, by Elgendi et al. [35], meaning that other evaluators could the MAX86150 has been integrated with an inertial motion have included some rejected data sets and vice versa. unit, a temperature and a magnetic field sensor in a stand- Although this is an inherent weakness in this method for alone cylindrical implant measuring 13 × 40 mm [47]. This assessing data quality, it is of greatest importance when implant will undergo testing in swim tunnel trials logging using PPG for determination of SpO2 where the PPG motion data, ECG and PPG. Data from both wavelengths shape is paramount for the validity of the SpO2 estimate will be logged to evaluate the possibility of deriving SpO2 [24]. When estimating HR only, PPG morphology is of from the data. These studies will enable the evaluation of lesser concern since only the frequency content of the how motion artefacts due to swimming motion impacts pulsatile PPG component is required. The PPG based HR the data, and potentially how such artefacts can be rem- estimate is, therefore, considered robust against varia- edied in post-processing. tions in interpretation of the subset selection criteria. S vendsen et al. Anim Biotelemetry (2021) 9:41 Page 11 of 12 2. Sommerset I, Walde C, Bang Jensen B, Bornø B, Haukaas A, Brun E. Fiske- Because this implant measures both ECG and PPG, co- helserapporten 2019. report 5a/2020 (Norwegian Veterinary Institute, processing of ECG and PPG for an even more robust HR Oslo, Norway, 2019) (2020) estimate is made possible. This is already a topic in human 3. Noble C, Nilsson J, Stien LH, Iversen MH, Kolarevic J, Gismervik K. Velferdsindikatorer for oppdrettslaks: Hvordan vurdere og dokumentere medicine to improve accuracy and robustness for heart fiskevelferd. 2. utgave (2018) rate variabilty estimates beyond what is currently possible 4. Føre M, Frank K, Norton T, Svendsen E, Alfredsen JA, Dempster T, Eguiraun using ECG alone [48]. H, Watson W, Stahl A, Sunde LM, et al. Precision fish farming: a new frame - work to improve production in aquaculture. Biosys Eng. 2017. https:// doi. Acknowledgements org/ 10. 1016/j. biosy stems eng. 2017. 10. 014. The authors thank the Research Council of Norway for funding the work (see 5. Knudsen F, Fosseidengen J, Oppedal F, Karlsen Ø, Ona E. Hydroacoustic “Funding”) as well as SINTEF Ocean AS for supporting the PhD project through monitoring of fish in sea cages: target strength (ts) measurements on their in-kind contribution. We also thank the personnel at NINA’s research sta- Atlantic salmon (Salmo salar). Fish Res. 2004;69(2):205–9. tion at Ims for facilitating the experiment. 6. Berckmans D. General introduction to precision livestock farming. Anim Front. 2017;7(1):6–11. Authors’ contributions 7. Føre M, Svendsen E, Alfredsen JA, Uglem I, Bloecher N, Sveier H, Sunde ES and JAA conceived the idea for the setup and testing the sensing principle LM, Frank K. Using acoustic telemetry to monitor the effects of crowding in fish. ES is responsible for the completion of the PhD in the funding project, and delousing procedures on farmed Atlantic salmon (Salmo salar). designed the required equipment, planned and executed the experiment, Aquaculture. 2018;495:757–65. prepared and processed data, and did the main job in preparing the manu- 8. Brijs J, Sandblom E, Axelsson M, Sundell K, Sundh H, Huyben D, Broström script. FØ participated in the planning and execution of the experiments with R, Kiessling A, Berg C, Gräns A. The final countdown: continuous physi- emphasis on anesthesia and surgery. MF, JAA assisted in equipment design ological welfare evaluation of farmed fish during common aquaculture and planning of the experiment. LLR has participated in data processing, practices before and during harvest. Aquaculture. 2018;495:903–911 while BF and REO have contributed with the anatomical and physiological 9. Hassan, W., Føre, M., Pedersen, M.O., Alfredsen, J.A.: A novel doppler based considerations. All authors have provided feedback and contributed to writing speed measurement technique for individual free-ranging fish. In: 2019 the manuscript. All authors read and approved the final manuscript. IEEE SENSORS, pp. 1–4 (2019). IEEE 10. Baktoft H, Gjelland KØ, Økland F, Thygesen UH. Positioning of aquatic Funding animals based on time-of-arrival and random walk models using yaps This study was funded by the Research Council of Norway (RCN Project Num- (yet another positioning solver). Sci Rep. 2017;7(1):1–10. ber 280864) and through in-kind from SINTEF Ocean. 11. Hassan W, Føre M, Urke HA, Kristensen T, Ulvund JB, Alfredsen JA, et al. System for real-time positioning and monitoring of fish in commercial Availability of data and materials marine farms based on acoustic telemetry and internet of fish (iof ). In: The data sets used and/or analysed during the current study are available The 29th International Ocean and Polar Engineering Conference (2019). from the corresponding author on reasonable request. International Society of Oshor ff e and Polar Engineers 12. Thorstad EB, Rikardsen AH, Alp A, Økland F. The use of electronic tags in fish research–an overview of fish telemetry methods. Turk J Fish Aquat Declarations Sci. 2013;13(5):881–96. 13. Deng ZD, Weiland MA, Fu T, Seim TA, LaMarche BL, Choi EY, Carlson Ethics approval and consent to participate TJ, Eppard MB. A cabled acoustic telemetry system for detecting and The animal experiment described herein was conducted with permission tracking juvenile salmon: Part 2. Three-dimensional tracking and passage given by the Norwegian Food Safety Authority (permit number 19/104024). outcomes. Sensors. 2011;11(6):5661–76. Doi: https:// doi. org/ 10. 3390/ Permission to use up to, but not exceeding, 15 individual Atlantic salmon was s1106 05661 given under the condition that data collection was stopped when 5 complete 14. Weiland MA, Deng ZD, Seim TA, LaMarche BL, Choi EY, Fu T, Carlson TJ, data sets were obtained as a reduction measure (3R) [49], a point that was Thronas AI, Eppard MB. A cabled acoustic telemetry system for detecting reached after 7 fish. The experiment was carried out at NINA Ims Research and tracking juvenile salmon: Part 1. Engineering design and instrumen- Station close to Stavanger, Norway, on August 28th, 2019. tation. Sensors. 2011;11(6):5645–60. Doi: https:// doi. org/ 10. 3390/ s1106 Consent for publication 15. Hussey NE, Kessel ST, Aarestrup K, Cooke SJ, Cowley PD, Fisk AT, Not applicable. Harcourt RG, Holland KN, Iverson SJ, Kocik JF et al. Aquatic animal telemetry: a panoramic window into the underwater world. Science. Competing interests 2015;348(6240):1255642 The authors declare that they have no competing interests. 16. Muñoz L, Aspillaga E, Palmer M, Saraiva JL, Arechavala-Lopez P. Acoustic telemetry: a tool to monitor fish swimming behavior in sea-cage aqua- Author details culture. Front Marine Sci. 2020;7: 645 Department of Engineering Cybernetics, NTNU, O.S. Bragstads Plass 2D, 17. Føre M, Frank K, Dempster T, Alfredsen JA, Høy E. Biomonitoring using 7034 Trondheim, Norway. Department of Aquaculture Technology, SINTEF tagged sentinel fish and acoustic telemetry in commercial salmon aqua- Ocean, Brattørkaia 17C, 7010 Trondheim, Norway. Department of Electronic culture: a feasibility study. Aquacult Eng. 2017;78:163–72. Systems, O.S. Bragstads Plass 2B, 7034 Trondheim, Norway. Nor wegian 18. Wendelaar Bonga SE. The stress response in fish. Physiol Rev. Institute for Nature Research, Høgskoleringen 9, 7485 Trondheim, Norway. 1997;77(3):591–625. Department of Biology, NTNU, Brattørkaia 17B, 7010 Trondheim, Norway. 19. Hvas M, Folkedal O, Oppedal F. Heart rate bio-loggers as welfare indicators in Atlantic salmon (Salmo salar) aquaculture. Aquaculture. Received: 6 May 2021 Accepted: 13 September 2021 2020;529:735630. 20. Svendsen E, Føre M, Økland F, Gräns A, Hedger RD, Alfredsen JA, Uglem I, Rosten C, Frank K, Erikson U et al. Heart rate and swimming activity as stress indicators for Atlantic salmon (Salmo salar). Aquaculture. 2020;531: References 21. Brijs J, Sandblom E, Rosengren M, Sundell K, Berg C, Axelsson M, Gräns A. 1. FAO: Food and Agriculture Organization of the United Nations, Fisheries Prospects and pitfalls of using heart rate bio-loggers to assess the welfare and Aquaculture department (2021). http:// www. fao. org/ fishe ry/ aquac of rainbow trout (Oncorhynchus mykiss) in aquaculture. Aquaculture. ulture/ en Accessed 2 May 2021 2019;509:188–197 Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 12 of 12 22. Charlton PH, Birrenkott DA, Bonnici T, Pimentel MA, Johnson AE, Alastruey 34. Coyle SD, Durborow RM, Tidwell JH, et al. Anesthetics in aquaculture vol. J, Tarassenko L, Watkinson PJ, Beale R, Clifton DA. Breathing rate estima- 3900. Southern Regional Aquaculture Center, Stoneville (2004) tion from the electrocardiogram and photoplethysmogram: a review. 35. Elgendi M. Optimal signal quality index for photoplethysmogram signals. IEEE Rev Biomed Eng. 2017;11:2–20. Bioengineering. 2016;3(4):21. https:// doi. org/ 10. 3390/ bioen ginee ring3 23. Severinghaus JW. Takuo aoyagi: discovery of pulse oximetry. Anesth 040021. Analg. 2007;105(6):1–4. 36. Doving K, Reimers E. Fiskens Fysiologi. Bergen: John Grieg; 1992 24. Chan ED, Chan MM, Chan MM. Pulse oximetry: understanding its 37. Lucas M. Heart rate as an indicator of metabolic rate and activity in adult basic principles facilitates appreciation of its limitations. Respir Med. Atlantic salmon, Salmo salar. J Fish Biol. 1994;44(5):889–903. 2013;107(6):789–99. 38. Butterworth C. Filter approximation theory. Engineer. 1930;7:536–41. 25. Islam MT, Zabir I, Ahamed ST, Yasar MT, Shahnaz C, Fattah SA. A time- 39. Savitzky A, Golay MJ. Smoothing and differentiation of data by simplified frequency domain approach of heart rate estimation from photoplethys- least squares procedures. Anal Chem. 1964;36(8):1627–39. mographic (ppg) signal. Biomed Signal Process Control. 2017;36:146–54. 40. Farrell AP. Encyclopedia of fish physiology: from genome to environment. 26. Erhardt W, Lendl C, Hipp R, von Hegel G, Wiesner G, Wiesner H. The use London: Academic press; 2011. of pulse oximetry in clinical veterinary anaesthesia. J Assoc Vet Anaesth 41. Johns M, Giller CA, German DC, Liu H. Determination of reduced scatter- Great Br Irel. 1990;17(1):30–1. https:// doi. org/ 10. 1111/j. 1467- 2995. 1990. ing coefficient of biological tissue from a needle-like probe. Opt Express. tb003 85.x. 2005;13(13):4828–42. 27. Theodor M, Ruh D, Subramanian S, Förster, K, Heilmann C, Beyersdorf F, 42. Bigio, I.J., Fantini, S.: Quantitative Biomedical Optics: Theory, Methods, Plachta D, Manoli Y, Zappe H, Seifert A. Implantable pulse oximetry on and Applications. Cambridge University Press, University Printing Press, subcutaneous tissue. In: 2014 36th Annual International Conference of Cambridge CB2 8BS, United Kingdom (2016) the IEEE Engineering in Medicine and Biology Society, pp. 2089–2092 43. Nitzan M, Khanokh B, Slovik Y. The difference in pulse transit time to (2014). IEEE the toe and finger measured by photoplethysmography. Physiol Meas. 28. Reynolds J, Ahmmed P, Bozkurt A. An injectable system for subcutaneous 2001;23(1):85. photoplethysmography, accelerometry, and thermometry in animals. 44. Golyandina N, Zhigljavsky A. Singular spectrum analysis for time series, IEEE Trans Biomed Circuits Syst. 2019;13(5):825–34. vol. 120. Heidelberg New York Dordrech London: Springer; 2013. 29. Machikhin AS, Burlakov AB, Volkov MV, Khokhlov DD. Imaging pho- 45. Meyer Y. Wavelets: algorithms & applications. Philadelphia: SIAM (Society toplethysmography and videocapillaroscopy enable noninvasive for Industrial and Applied Mathematics; 1993. study of zebrafish cardiovascular system functioning. J Biophotonics. 46. Cooley JW, Tukey JW. An algorithm for the machine calculation of com- 2020;13(7):202000061. plex fourier series. Math Comput. 1965;19(90):297–301. 30. Seo E, Sazi T, Togawa M, Nagata O, Murakami M, Kojima S, Seo Y. A port- 47. Svendsen E, Føre M, Randeberg L, JA,. Design of a novel biosensor able infrared photoplethysmograph: heartbeat of Mytilus galloprovincialis implant for farmed Atlantic salmon (Salmo salar). In: 2021 IEEE SENSORS, analyzed by mri and application to bathymodiolus septemdierum. Biol Accepted for Publication (2021). IEEE Open. 2016;5(11):1752–7. 48. Zaretskiy A, Mityagin K, Tarasov V, Moroz D, Kuraleva A. Robust heart 31. Tomlinson S, Behrmann S, Cranford J, Louie M, Hashikawa A. Accuracy rate estimation using combined ecg and ppg signal processing. In: IOP of smartphone-based pulse oximetry compared with hospital-grade Conference Series: Materials Science and Engineering, vol. 537, p. 042077 pulse oximetry in healthy children. Telemed e-Health. 2018;24(7):527–35. (2019). IOP Publishing https:// doi. org/ 10. 1089/ tmj. 2017. 0166. 49. Norecopa: Norecopa, Norway’s National Consensus Platform (2021). 32. Hahnen C, Freeman CG, Haldar N, Hamati JN, Bard DM, Murali V, Merli GJ, https:// norec opa. no/ Accessed 10 Mar 2021 Joseph JI, van Helmond N. Accuracy of vital signs measurements by a smartwatch and a portable health device: validation study. JMIR Mhealth Publisher’s Note Uhealth. 2020;8(2):16811. Springer Nature remains neutral with regard to jurisdictional claims in pub- 33. Maxim Integrated: Maxim Integrated, Integrated Photoplethysmogram lished maps and institutional affiliations. and Electrocardiogram Bio-sensor Module for Mobile Health (2021). https:// datas heets. maxim integ rated. com/ en/ ds/ MAX86 150. pdf Accessed 9 Aug 2021. Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Animal Biotelemetry Springer Journals

Optical measurement of tissue perfusion changes as an alternative to electrocardiography for heart rate monitoring in Atlantic salmon (Salmo salar)

Loading next page...
 
/lp/springer-journals/optical-measurement-of-tissue-perfusion-changes-as-an-alternative-to-di11Vfl0Yy

References (34)

Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2021
eISSN
2050-3385
DOI
10.1186/s40317-021-00264-w
Publisher site
See Article on Publisher Site

Abstract

Background: Welfare challenges in salmon farming highlights the need to improve understanding of the fish’s response to its environment and rearing operations. This can be achieved by monitoring physiological responses such as heart rate (HR) for individual fish. Existing solutions for heart rate monitoring are typically based on Electrocardiog- raphy (ECG) which is sensitive to placement and electrode orientation. These factors are difficult to control and affects the reliability of the principle, prompting the desire to find an alternative to ECG for heart rate monitoring in fish. This study was aimed at adapting an optical photoplethysmography (PPG) sensor for this purpose. An embedded sensor unit measuring both PPG and ECG was developed and tested using anesthetized Atlantic salmon in a series of in-vivo experiments. HR was derived from PPG and compared to the ECG baseline to evaluate its efficacy in estimating heart rate. Results: The results show that PPG HR was estimated with an accuracy of 0.7 ± 1.0% for 660 nm and 1.1 ± 1.2% for 880 nm wavelengths, respectively, relative to the ECG HR baseline. The results also indicate that PPG should be meas- ured in the anterior part of the peritoneal cavity in the direction of the heart. Conclusion: A PPG/ECG module was successfully adapted to measure both ECG and PPG in-vivo for anesthetized Atlantic salmon. Using ECG as baseline, PPG analysis results show that that HR can be accurately estimated from PPG. Thus, PPG has the potential to become an alternative to ECG HR measurements in fish. Keywords: Salmo salar, Heart rate, Implant, Biosensors, Photoplethysmography management and operation of such salmon farms entails Background a broad range of interrelated operations exerting convo- Atlantic salmon (Salmo salar) is one of the most impor- luted effects on the fish. To ensure acceptable fish health tant species in fish farming with more than 2.6 Mt pro - and welfare conditions during production, relevant data duced globally in 2019 [1]. A typical salmon production to describe these must be collected and evaluated in site consists of 8–10 flexible sea cages usually 50  m in conjunction with operational data. Such evaluations are diameter that holds a volume of around 40,000 m . The largely subjective and experience based, and are carried out as part of the daily inspection and feeding routines. *Correspondence: eirik.svendsen@ntnu.no 1 About 15% of all farmed salmon are lost during pro- Department of Engineering Cybernetics, NTNU, O.S. Bragstads Plass 2D, 7034 Trondheim, Norway duction, a loss partly attributed to the lack of objective Full list of author information is available at the end of the article input to production control [2]. To address this loss, © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 2 of 12 recent research efforts have focused on determining Changes in treatment and/or living conditions chal- what constitutes good fish welfare and which indicators lenges an animal’s homeostasis and may be expressed as one should quantify (i.e., measure) to obtain a more various stress responses [18]. Recently, heart rate meas- objective foundation for making welfare critical deci- ured using DSTs was demonstrated as being linked to sions [3], and safeguard ethically sound production. welfare [19] and stress [20] in Atlantic salmon and in Sensing such indicators should be done using as unob- other salmonids such as rainbow trout [21]. Heart rate trusive measures as possible to avoid compromising may, therefore, serve as a proxy for stress, thereby provid- safe and efficient operations. However, this is in conflict ing information on the fish’s welfare in aquaculture pro - with the challenge of collecting data from a biomass vided that heart rate can be obtained in real time. Heart contained in a volume of roughly 40,000 m of water, rate estimates for fish are most commonly derived from as this calls for distributed or mobile data collection an electrocardiogram (ECG) measured by electrodes systems which are based on, e.g., environmental sens- integrated in the tag’s encapsulation. However, the reli- ing networks, sonars/echo sounders, passive acoustic ability of this method has been shown to be sensitive to monitoring and different types of optical methods such tag deployment as the derived heart rate depends on the as underwater cameras [4]. Such systems are likely to lateral placement in the fish as well as the tag’s orienta - be intrusive on the production, and may thus have to be tion [21]. Exploring other potentially more robust meas- removed from the aquaculture cage before operations urement principles for their ability to sense heart rate are carried out to reduce the likelihood of, e.g., equip- is, therefore, desirable, as it might contribute to making ment damage. Furthermore, such technologies provide heart rate data more reliable and feasible to obtain for data with different spatial and temporal resolution and free swimming fish in aquaculture. provide data on a population level since the recorded Photoplethysmography is an optical sensing technique data are obtained from a sub-volume in the sea cage quantifying changes in tissue perfusion (i.e., blood vol- rather than from a specific group of fish. Sonars and ume) by optical absorption. The photoplethysmogram cameras can provide data on a group level if smaller (PPG) is a convolution of many components. In humans, parts of the biomass are present within the sensor’s slow-varying components may arise from breathing [22], field of view, a feature that has been used to estimate, while the superimposed pulsatile component changes e.g., fish size and cage biomass [5 ]. Although group with the cardiac cycle [23]. Using the pulsatile compo- and population level data can provide a certain insight nent from several wavelengths of light enables estima- into the dynamics in animal husbandry operations, the tion of arterial blood oxygen saturation (SpO2) provided sensing technologies’ intrusiveness on operations and a mapping function compensating for tissue scattering limited understanding of the measurements’ link to effects is known [24]. Independent from this, the pulsatile welfare imply that alternative or supplemental solutions component can be used to obtain a heart rate estimate for welfare evaluations are needed. as shown for both humans and other mammals [25, 26]. Individual level data provides the highest possible data u Th s, PPG is considered a likely candidate to address the resolution with respect to biomass, a feature recognized challenges associated with ECG obtained using implants in precision farming both on land [6] and at sea [4]. In provided a suitable sensing solution can be designed. aquaculture, the main method for obtaining individual Implantable PPG solutions have been developed for, level data is using miniaturized, encapsulated electronic e.g., sheep [27] and other mammals [28]. Although systems commonly referred to as “tags”. Typical parame- potentially relevant, such solutions are proprietary and ters that can be measured in an operational setting today either require additional, external interfaces for power are activity and swimming depth [7], heart rate (HR) [8], and data collection/processing, or depend on radio- swimming speed [9] and position [10, 11]. Tags are avail- based data transfer, making them infeasible for use in able in two main types: Data storage tags (DSTs) and fish in seawater. Measurement of PPG has been dem - telemetry transmitter tags [12], and are usually surgically onstrated for aquatic animals such as zebrafish (Danio implanted into the fish’s peritoneal cavity. rerio) using imaging PPG [29]. This is a suitable tech - Tagging has been used successfully to study fish in dif - nique when tissues with low opacity can be remotely ferent situations ranging from tracking fish in rivers and imaged in a controlled environment. This is the case sea [13–15] to shedding light on fish behaviour in sea for zebrafish in the larval stage, but not for Atlantic cages during aquaculture operations [7, 16]. Thus, tagged salmon in a context where using implants is consid- individuals may act as representatives (i.e., “sentinel fish”) ered relevant. Correspondingly, obtaining heart rate for the rest of the biomass in an aquaculture cage [17] from PPG has been demonstrated by drilling holes and to facilitate safer and less operationally intrusive welfare inserting PPG sensors through the shell of the Medi- evaluations. terranean mussel (Mytilus galloprovincialis) [30]. This S vendsen et al. Anim Biotelemetry (2021) 9:41 Page 3 of 12 approach requires stationary subjects as the sensors are Methods wired to external hardware for power and data logging. Sensing and logging equipment These approaches are, thus, not directly suited for fish The MAX86150 optical biosensor module [33] from implants, but demonstrate the feasibility of applying Maxim Integrated (San Jose, USA) that offers both PPG PPG for heart rate sensing for different aquatic animal and ECG output in a miniature package (3.3 × 5.6 × mm) groups. with ultra-low power consumption, was selected for the In recent years a market desire for mobile medi- measurements in this study. The physical and electrical cal equipment and consumer products containing properties of the module make it suitable for integration such sensors (e.g., smart phones and sports watches as the biosensor element of a prospective DST or acoustic [31, 32]) has driven development and miniaturiza- tag with such sensing capability. A printed circuit board tion of PPG sensors. Such innovations have increased (PCB) was designed as a platform for miniaturizing the the potential to include such off-the-shelf sensors in reference circuit design for the MAX86150. The PCB was implants for fish, as this is an application with simi- subsequently cast in a cylindrical epoxy casing such that lar requirements with respect to size and power con- the sensing direction was perpendicular to the cylinder’s sumption. However, the anatomical and optical tissue longitudinal axis. Seven thin electrical wires protruded property differences between mammals and fishes, the rear of the cast and was used to extend the ECG raise the question of whether PPG captured in Atlan- electrode connections of the MAX86150, and to enable tic salmon using the common intraperitoneal implan- power connection and data transfer from the sensor. tation approach results in useful data. The purpose A battery powered MAX32630FTHR microcontroller of this study was, therefore, to adapt an off-the-shelf board was used to relay data to a PC using Bluetooth. PPG/ECG biosensing module for in-vivo testing in The cast also included a threaded aluminium end piece Atlantic salmon to investigate its potential to provide for attachment to a fixture rod. The fixture rod allowed an alternative heart rate estimate from Atlantic salmon horizontal and vertical sensor orientations as well as that does not suffer from the same limitations as ECG, rotation of the sensor around its longitudinal axis. The while facilitating a future low-cost implant based on complete assembly is shown in Fig. 1. Maxim DeviceStu- standard electronic components. dio (V 5.3.03289.0) was installed on a Dell Latitude 7490 Fig. 1 A (left): 1. Fixture rod, 2. Wire extensions for MAX86150 ECG electrodes, power connection and data transfer, 3. Threaded aluminium end piece, 4. MAX86150 biosensor module. B (right): 1. Epoxy cast with MAX86150 biosensor module, 2. ECG electrode extension leads, 3. Data and power supply extension leads, 4. MAX32630FTHR microcontroller board for wireless data transfer to PC Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 4 of 12 laptop computer and used for real-time data inspection and logging. To determine the biosensor’s feasibility when applied to Atlantic salmon, all sensor settings were kept the same for all data sets (sampling frequency fs = 200  Hz, RED (660  nm) and IR (880  nm) LED pulse amplitudes A = A = 20 mA, LED pulse width PW = 400  µs, and 660 880 no sample averaging), thus enabling signal quality com- parisons across data sets. Note that when using PPG to estimate heart rate only, no reference measurements apart from the ECG baseline used for HR comparisons are required. Experimental procedures Individual fish were captured from a holding tank using a knotless dip net and immediately transferred to a tank containing water from the same water supply and anes- −1 thetic in a knock-out solution (70  mg  l Benzoak Vet). The tank was covered using a Styrofoam plate after the fish was placed inside. Using a wheeled trolley, the tank was then moved to the indoor experimental location. When deemed to have reached level 3 anesthesia [34], the fish was placed in a specialized surgical table. By Fig. 2 A (top): Rotation angles for the horizontal sensor orientation. inserting a hose connected to a small pump into the fish’ B (bottom): Rotation angles for the vertical sensor orientation. buccal cavity, continuously aerated water with mainte- Illustration by Mats Mulelid, SINTEF Ocean AS −1 nance anesthetic (35  mg  l Benzoak Vet) was used to irrigate the gills and keep the fish sedated. The wire elec - trodes for ECG measurement attached to the MAX86150 were inserted into the muscle on the left and right lateral sides close to the heart. A 1–2  cm incision was made in the abdominal wall along the sagittal plane just anterior of the pelvic fins (position 1, Fig.  2A). For position 1, the sensor was inserted horizontally into the peritoneal cav- ity and one data set collected for orientations of 0° (ven- tral), 90° (right lateral), 180° (dorsal), 270° (left lateral) and 0° (ventral repeated) degrees rotation, respectively (Fig.  2A). Following data collection in position 1, a sec- ond 1 cm incision was made in the abdominal wall along the sagittal plane just posterior from the transverse sep- tum (position 2, Fig.  2B). For position 2, the sensor was inserted vertically into the peritoneal cavity and one data set collected for orientations of 0° (anterior), 90° (right lateral), 180 degree (posterior), 270 degree (left lateral) Fig. 3 Data collection and 0° (anterior repeated) degrees, respectively (see Figs.  2B and 3). Data was collected for 1  min for both positions and all orientations for each individual (10 data sets per fish / 50 data sets in total), after which the fish preventing valid ECG signals. Data collection was suc- −1 was euthanized by an anesthetic overdose (> 70  mg  l cessful from the remaining 6 fish (from now on referred Benzoak Vet) and exsanguination. to as Fish 1 through 6), resulting in 10 data sets for fish A total of 7 fish were used for the experiment. No 1, 2, 4, 5 and 6, and 6 data sets for fish 3 which suffered valid data was obtained for the first fish due to electri - cardiac arrest after 6 data sets. The analyses were thus cal interference likely caused by a nearby water pump based on 56 data sets in total. S vendsen et al. Anim Biotelemetry (2021) 9:41 Page 5 of 12 Data processing QRS peaks were 40% higher than neighbouring peaks. Valid 20  s data subsets were manually selected and HR was then determined by the average time difference labelled based on the data evaluation criteria presented (∆t ) between peaks using: avg by Elgendi [35]: HR = · 60 BPM (1) avg • Salient ECG baseline. • Salient PPG signal for one or both wavelengths. Because raw PPG data containing both a stochastic • Similar PPG waveform morphology throughout the trend and measurement noise was logged, a preprocess- entire subset. ing step to remove these was necessary. The slow-varying (i.e., low-frequency) trend was removed using a third Data analysis using Python 3.8 (Anaconda Inc., Aus- order Butterworth highpass filter [38] with a cutoff fre - tin, Texas, USA) could then be conducted for valid data quency of 0.25  Hz. Measurement noise was reduced sets labelled”GOOD” (accept) or”FAIR” (accept). Data using a second-order Savitzky-Golay filter [39] with a 50 sets labelled “BAD” (reject) were judged to not contain sample (i.e., 0.25 s) window. This window size was chosen the required information for signal quality calculations to retain as much information in the signal as possible and were, therefore, omitted from further analysis. A based on the lowest expected HR of 15 BPM for Atlan- complete overview of data sets and corresponding wave- tic Salmon [37] and comparable species [21]. An example lengths considered valid based on these criteria is given illustrating the result of these processing steps is given in Table  1, while representative samples of PPG data in Fig. 5. HR from PPG was then calculated from the de- sets labelled “GOOD”, “FAIR” and “BAD” can be seen in trended and noise suppressed signal using autocorrela- Fig. 4. tion defined by: The default filtered ECG signal output from the MAX86150 was used to find HR by first scaling the fil - y(m) = X(n)X(n + m) tered output to [0, 1] and then running a peak detection (2) n=1 algorithm. Scaling was done so the same peak detec- tion parameters for all ECG time series could be used. where N is the number of samples in the signal X, and Peaks were identified using the “find peaks()” method in m ∈ [0, N − 1]. By applying the same peak detection Python’s statistics module utilizing both a time window method as that used for ECG, the index of the domi- (i.e., the minimum required distance between peaks) and nating peak (i.e., the first peak) in the autocorrelation a prominence limit to identify the QRS peaks (i.e., ventri- series was identified. Because the autocorrelation series cle depolarization) [36]. The time window was set to 160 featured softer (but not flat/diffuse) peaks compared samples (i.e., 0.8 s) based on a maximum expected HR of to the ECG signal, the prominence setting in the peak 80 BPM for Atlantic salmon [37] and comparable species detection algorithm was relaxed to 0.3 to avoid too [21]. Prominence was set to 0.4, thus demanding that the Table 1 Summary of data set evaluations for both orientations, “Or.”, (Hor = horizontal and Ver = vertical) and all rotations, “Rot.”, (see Fig. 2) for all fish Or. Rot. Fish1 Fish2 Fish3 Fish4 Fish5 Fish6 Hor 1 0 NVD NVD λ2 NVD λ2 NVD Hor 90 NVD NVD λ2 NVD NVD NVD Hor 180 NVD NVD λ1, λ2 NVD λ2 NVD Hor 270 NVD NVD λ1, λ2 NVD λ1, λ2 NVD Hor 2 0 NVD NVD λ1, λ2 NVD NVD NVD Ver 1 0 λ1, λ2 λ1, λ2 λ1, λ2 λ1, λ2 λ1, λ2 NVD Ver 90 λ1, λ2 λ1, λ2 N/A λ1, λ2 NVD λ1, λ2 Ver 180 λ1, λ2 NVD N/A NVD NVD NVD Ver 270 λ1, λ2 λ1, λ2 N/A λ1, λ2 λ1, λ2 λ1, λ2 Ver 2 0 λ1, λ2 λ1, λ2 N/A λ1, λ2 λ1, λ2 λ1, λ2 NVD signifies “No Valid Data”, i.e., no part of the time series fulfils all selection requirements. λ1 and λ2 represents valid data for 660 nm and 880 nm, respectively “N/A” (Not Applicable) denotes the data series which could not be collected for Fish 3 due to cardiac arrest. Note that the 0 degrees measurement has been repeated for both orientations Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 6 of 12 Fig. 4 Signal examples Fig. 5 Top: Raw data. Middle: De-trended and smoothed data. The red and blue dots illustrate the detected peaks and valleys, respectively. Bottom: Scaled ECG. The red dots illustrate the detected ECG peaks aggressive peak rejection. As for ECG HR, the time calculate PPG HR in beats per minute (HRBPM) using window was set to 160 samples. The index of the identi - Eq. 1. fied peak, therefore, represented the average time delay To evaluate which sensor orientation and rotation gave between peaks in the PPG time series, and was used to the highest quality data, the signal quality index (SQI) for S vendsen et al. Anim Biotelemetry (2021) 9:41 Page 7 of 12 Fig. 6 Signal processing flow the noise suppressed (but not de-trended) PPG signals, was Discussion calculated. SQI is defined for individual PPG pulses by [35]: The high accuracy for HR computed from PPG com - pared to the ECG baseline from our experiments indi- PPG pa cates that PPG is a viable alternative sensor principle for P = · 100 SQI (3) PPG avg monitoring heart rate in Atlantic salmon. Moreover, the MAX86150 unit appears to be a suitable sensor module where PPG is the peak-to-peak PPG pulse amplitude and pa for implementation in future DSTs or transmitter tags PPG its mean value. The PPG signal where only noise avg aiming to measure and report HR over time. Although the suppression had been applied was used to retain access sensor is designed for integration in optical HR measure- to the pulses’ mean value which was removed during de- ment applications for humans and, thus, with human tis- trending. A pulse was defined as the valley to valley noise sues and blood in mind, the HR estimate relies solely on suppressed data subsets. Valley indices (i.e., positions) were the time-varying tissue perfusion. While human and s fi h identified by inverting the de-trended (i.e., mean-centered) blood is different with respect to composition and cell signal and reapplying the peak detection algorithm with the morphology, the essential functionality is the same, as same window and prominence settings as those for auto- is the hemoglobin [40]. The wavelengths emitted by the correlation peak detection. Using the valley indices, the MAX86150 sensor are chosen due to their absorption pulses were extracted from the noise suppressed (but not sensitivity to oxy- and deoxyhemoglobin which deter- de-trended) data sets, and the average SQI (SQI ) for all avg mines the blood’s oxygenation and thus its colour. Because pulses calculated to represent the subset SQI. An example tissue colour is affected by perfusion, processing either or illustrating the peak detection result is given in Fig. 5. both wavelength will be a feasible approach to obtain HR To evaluate the accuracy of the PPG HR estimates, the in both humans and fishes. difference between the ECG baseline and both wave - The data used in this study originates from 6 different lengths separately were calculated in percent using: fish, yielding 28 data sets. Although up to 10 data sets were collected from each fish, the data sets are consid - HR − HR PPG ECG = · 100 (4) HR ered independent because each data set is separate in HR ECG either time, relates to different tissues, or both. When The signal processing flow is illustrated in Fig. 6. reviewing Table  1, most data available for processing came from the anterior part of the fish. This is prob - Results ably because the measurements in this region coincide ECG was labeled “GOOD” in all data sets, while one with locations with a high blood supply such as the or both PPG wavelengths were judged to qualify as liver and gut (thus implying high perfusion) compared “GOOD” or “FAIR” in 28 data sets. Thus, 28 (of 56) to the posterior region where tissues with lower perfu- data sets were omitted from analysis. The results from sion (mostly white muscle tissue) are present. This is the data processing are given in Table  2. The results supported by that the average SQI (Table 2) was highest show an average accuracy of 0.7 ± 1.0% for 660 nm and for locations and orientations associated with the ante- 1.1 ± 1.2% for 880  nm wavelengths relative to the ECG rior part of the fish, especially for the 0 degree rotation HR baseline. The average SQI (SQI column in Table 2) which is towards the heart. This indicates that future avg indicates that the vertical orientation and 0 degrees implementations of PPG sensors for Atlantic salmon rotation (i.e., measuring towards the heart) resulted in should facilitate data collection in this area. the best data quality. Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 8 of 12 Table 2 Results sorted by average SQI where the columns named Fish is the fish number, Or. is the orientation, Rot. is the rotation, HR is the heart rate in BPM calculated from ECG, HR is the calculated heart rate in BPM for the 660 nm time series, HR is the ECG 660 880 calculated heart rate in BPM for the 880 nm time series, ∆HR is the difference in HR estimate between the 660 nm time series and ECG in percent, ∆HR is the difference in HR estimate between the 880 nm time series and ECG in percent, SQI is the SQI for 880 660 660 nm, SQ I is the SQI for 880 nm and SQI is the average SQI, respectively 880 avg Fish Or. Rot. HR HR HR ∆HR ∆HR SQI SQI SQI Likely tissues ECG 660 880 660 880 660 880 avg 5 Ver 270 50.4 50.2 50.4 0.40 0.00 0.03 0.02 0.02 M, F 3 Hor 2 0 53.1 53.3 53.3 0.38 0.38 0.04 0.05 0.05 F 5 Hor 270 50.6 50.4 50.6 0.40 0.00 0.02 0.08 0.05 M, F 5 Hor 1 0 35.9 N/A 35.2 N/A 1.95 N/A 0.08 0.08 F 5 Hor 180 46.0 N/A 46.0 N/A 0.00 N/A 0.10 0.10 F, I 2 Ver 270 43.0 42.7 42.4 0.70 1.40 0.16 0.14 0.15 M, F 5 Ver 1 0 49.0 48.6 48.2 0.82 1.63 0.28 0.06 0.17 L, G 3 Hor 90 41.1 N/A 40.7 N/A 0.97 N/A 0.18 0.18 M, F 3 Hor 1 0 41.1 N/A 40.7 N/A 0.97 N/A 0.18 0.18 F 3 Hor 180 41.7 41.2 41.0 1.20 1.68 0.21 0.21 0.21 F, I 6 Ver 90 27.5 27.3 26.8 0.73 2.55 0.36 0.17 0.26 M, F 3 Hor 270 42.9 42.9 43.5 0.00 1.40 0.48 0.24 0.36 M, F 5 Ver 2 0 52.6 52.6 52.6 0.00 0.00 0.48 0.25 0.36 L, G 4 Ver 270 37.7 37.7 37.9 0.00 0.53 0.81 0.51 0.66 M, F 2 Ver 90 35.0 35.1 35.2 0.29 0.57 0.95 0.74 0.84 M, F 6 Ver 270 36.0 35.8 35.7 0.56 0.83 1.51 0.55 1.03 M, F 2 Ver 2 0 48.6 48.8 48.8 0.41 0.41 0.87 1.59 1.23 L, G 1 Ver 270 18.0 18.2 18.2 1.11 1.11 2.35 0.63 1.49 M, F 1 Ver 90 13.9 13.8 13.8 0.72 0.72 1.95 1.24 1.60 M, F 1 Ver 180 19.0 18.9 18.9 0.53 0.53 2.55 0.69 1.62 G, F 6 Ver 2 0 33.6 33.6 33.6 0.00 0.00 3.27 0.66 1.96 L, G 2 Ver 1 0 32.8 32.0 32.0 2.44 2.44 2.66 1.84 2.25 L, G 4 Ver 2 0 70.2 70.2 70.2 0.00 0.00 2.96 1.61 2.28 L, G 3 Ver 1 0 59.4 59.4 59.4 0.00 0.00 3.75 1.32 2.54 L, G 4 Ver 1 0 32.4 30.7 31.3 5.25 3.40 4.69 1.38 3.04 L, G 1 Ver 1 0 19.4 19.3 19.3 0.52 0.52 13.2 2.90 8.05 L, G 1 Ver 2 0 26.7 26.5 26.5 0.75 0.75 14.59 6.20 10.40 L, G 4 Ver 90 37.2 37.6 39.2 1.08 5.38 24.68 7.92 16.30 M, F The table entry “N/A” (Not Applicable) is for the cases where no valid data could be identified for the respective wavelength in a particular data set. Entries in the Likely tissues column are: M: Muscle, F: Fat, I: Intestines, L: Liver and G: Gut, and denote tissues likely present in the sensing volume for the associated sensor orientation and rotation In our results,  HR was reported using one decimal The range between the lowest (13.9 BPM) and high - because additional decimal points are likely inaccurate. est (70.2 BMP) heart rates may be explained by that This conclusion stems from considering the HR-depend different fish individuals would have had different phys - - iological baselines (e.g., stress levels and health) prior ent quantification error for our highest HR (70.2 BPM). to the experiments. These differences would, thus, have When sampling at 200  Hz, the change in BPM resulting yielded different individual responses to handling and from what is considered the maximum timely offset from anesthesia. Differences in heart rates were, therefore, be true peak position can be calculated. This is achieved by expected. Reported heart rates for Atlantic salmon and first finding beats per second (BPS) which in this case is comparable species are between 15 and 80 BPM [21, 70.2/60 = 1.17 BPS. The time between HR peaks will in 37]. With the exception of one individual (13.9 BPM) this case be ∆t = 1/1.17 = 0.855  s. A timely offset in true all measured heart rates fell within this expected range. peak placement exceeding 50% of the sampling interval The individual having a heart rate below 15 could have implies that a peak will be associated with the previous or been more susceptible to sedation thus explaining this next sampling point. Hence, for our quantification error result, or the reported HR range is conservative. we get Eq = 0.855 + (1/200) · 0.5 = 0.8575 s. This peak to S vendsen et al. Anim Biotelemetry (2021) 9:41 Page 9 of 12 peak distance which includes the 50% offset, then gives a component. This is partly remedied using high pass filter - shifted BPM of BPM = 60/0.8575 = 69.995. The BPM dif - ing in the analyses, as this effectually reduces or removes ference can then be calculated as ∆BPM = 70.2 − 69.995 long term trends and any changes therein. Moreover, an ≈ 0.2 BPM. The corresponding number for our lowest estimate of HR relies solely on the frequency content in baseline HR (13.9 BPM) is 0.008 BPM. the PPG signal measurement signal, and not the ampli- The quantification error considerations are closely tude. Based on these observations, movement of tissues related to our measurements’ sensitivity. By first accept - relative to the sensor during data collection were unlikely ing that a peak cannot be placed “between” two sampling to have had impact on the results. points, the sensitivity can then be evaluated in the same Motion artefacts may also have been caused by tissue way as the quantification error, only using the whole sam - contraction (e.g., the heart) if it was within the sens- pling interval. Thus, the sensitivity can be considered to ing volume during data collection. This is particularly be twice that of the quantification error, i.e., 0.4 BPM per relevant for the vertical orientation with 0° rotation sample offset for 70.2 BPM, and 0.016 BPM per sample (i.e., when the sensor was pointing towards the heart). offset for 13.9 BPM. Because this is of particular concern, a subsequent The deviation from the baseline of 0.7 ± 1.0% for post mortem dissection of Atlantic salmon was done to 660  nm and 1.1 ± 1.2% for 880  nm can be explained by assess the sensing volume for this orientation and rota- different factors. One potentially important source tion. The dissection revealed that the tissue observed of error is that PPG is sensitive to motion artefacts. with the present method was likely dominated by low- Such artefacts can be divided into two types: Perfusion perfusion fatty tissues surrounding larger blood ves- changes in tissue caused by motion and relative motion sels such as the hepatic arteries and veins (Fig.  7) [40]. between the sensor and the sensing volume. The former Fat has a high optical scattering [41] coefficient due to is not considered relevant when evaluating the results lipid droplets inside the fat cells. Due to the size of the because all fish were in level 3 anesthesia (surgical) and, scatters, this scattering is highly forward directed and thus, motionless. However, such artefacts are likely to be almost independent of the wavelength of the light. This important when the method is applied to free-swimming implies that the light from the sensor is strongly scat- non-sedated fish. A logical next step on the path towards tered by the tissue while the intensity decays exponen- an operational measurement method would, therefore, tially with distance from the light source in accordance be to apply the sensor to fish exhibiting normal swim - with the (modified) Beer–Lambert law [42], thus lim - ming behaviour, and collect concurrent PPG and motion iting the distance light travels. This is indicated by the data to assess the potential impact of specific motion fact that the SQI for both wavelengths was generally low patterns. due to the big difference between the pulsatile PPG com - To minimize the effect of the latter, the setup was ponent’s amplitude and the signal mean. A large mean designed to be as rigid as possible to ensure a stable value implies that a lot of light is scattered back to the sensing environment during data collection (Figs.  1 and receiver without having penetrated far into the tissue. It 3). However, although the setup was mechanically sta- is, therefore, likely that the data originates from tissues ble and the fish in level 3 anesthesia, motion artifacts close to the light source and that the heart is not part of caused by potential tissue movement such as peristalsis the sensing volume. [40] may have caused transient changes in the trend and Accuracy may also have been affected by the physiolog - potential changes in the amplitude of the PPG’s pulsatile ical state of the fish. The fish used in the experiment were Fig. 7 A (left): 1. Epigastric vein/artery, 2. Pylorus caeca/lipid deposits, 3. Liver. B (right) 1. Septum transversum, 2. Hepatic arteries/veins Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 10 of 12 lab grown and showed no signs of deteriorated health. Autocorrelation was chosen for PPG analysis because The anesthesia had both an analgesic and a paralyzing its low computational demand and robustness against effect. It must, therefore, be expected that the secondary potential transient motion artifacts makes it a likely can- circulation system driven by the caudal heart and move- didate for implementation in a microcontroller suitable ment was impaired during data collection. In addition, for integration in a size and energy constrained fish tag. the fish underwent a surgical procedure and the incisions However, it cannot be ruled out that other, more compu- as well as the sensor insertion may have further disturbed tationally demanding approaches such as singular spec- parts of the circulation system. When reviewing Table  2 trum analysis [44], wavelet analysis [45] or a fast Fourier for fish where data for both iterations of the horizontal transform [46] approach could have given better results. and vertical 0° are available, similar SQIs for both itera- The perfusion index (Eq.  3) was used to evaluate sig- tions appear to be the trend. Although this indicates that nal quality because this is considered the “gold standard” the physiological state of the fish remained stable during for PPG signal quality evaluation, even though alterna- data collection, it is likely that this state differs from that tive methods (e.g., skewness, kurtosis and entropy) [35] a fully awake and moving fish would exhibit. This under - which might lead to better results, exist. Such methods, lines the necessity of conducting further experiments however, are derived using reference data readily avail- with fish exhibiting more normal behaviours and physi - able for humans. To the authors’ knowledge, no reference ological function. PPG data for fish exist for comparison. Such data would The peak detection procedure may have affected the be a very useful resource in developing new methods for accuracy of the method because the detrended PPG sig- quality evaluations of PPG data from fish, particularly if nals consisted of an oscillating curve with wide peaks aspiring to quantify SpO2. compared to the ECG peaks. The accuracy in PPG peak Overall, valid data was identified for all orientations detection was, therefore, lower, thus resulting in slight and rotations. Although certain combinations of orien- differences in the intervals between the detected PPG tation and rotation yielded fewer data sets fulfilling the peaks and their corresponding ECG peaks. For long subset selection criteria than others, this does not nec- time series containing many peaks, such differences are essarily mean that data are harder to obtain for these expected to cancel out, but this may not have been the orientations and rotations. The same low-energy output case for 20 s data sets. This may have been further exac - settings were used for both orientations and all rota- erbated by the fact that the various orientations and rota- tions, thus implying that more valid data could have been tions would have illuminated tissues and capillary beds obtained if sensor settings, such as output power, had supplied by haemal arches connected to different points been increased. PPG, therefore, has the potential of being along the dorsal artery. The resulting differences in pulse a robust alternative to ECG for HR measurement in fish. transit times (PTT, i.e., times between the heart beat and when it is observable in the sensing volume), may have Conclusions shifted the PPG in relation to the ECG [43]. Further- A PPG/ECG module has been successfully adapted to more, if different capillary beds with different PTTs were measure both ECG and PPG in-vivo for anesthetized present in the sensing volume it could have distorted Atlantic salmon. Using ECG as baseline, results from the PPG pulses by dragging them out in time, thereby an analysis of the PPG signals show that that HR can be explaining the differences in morphology seen between accurately estimated from the PPG measurement, thus the “GOOD” and “FAIR” pulse examples in Fig. 4. having the potential to become an alternative to ECG HR The criteria used for the selection of valid data subsets measurements in fish. for analyses are subject to interpretation as highlighted Based on the encouraging results from this experiment, by Elgendi et al. [35], meaning that other evaluators could the MAX86150 has been integrated with an inertial motion have included some rejected data sets and vice versa. unit, a temperature and a magnetic field sensor in a stand- Although this is an inherent weakness in this method for alone cylindrical implant measuring 13 × 40 mm [47]. This assessing data quality, it is of greatest importance when implant will undergo testing in swim tunnel trials logging using PPG for determination of SpO2 where the PPG motion data, ECG and PPG. Data from both wavelengths shape is paramount for the validity of the SpO2 estimate will be logged to evaluate the possibility of deriving SpO2 [24]. When estimating HR only, PPG morphology is of from the data. These studies will enable the evaluation of lesser concern since only the frequency content of the how motion artefacts due to swimming motion impacts pulsatile PPG component is required. The PPG based HR the data, and potentially how such artefacts can be rem- estimate is, therefore, considered robust against varia- edied in post-processing. tions in interpretation of the subset selection criteria. S vendsen et al. Anim Biotelemetry (2021) 9:41 Page 11 of 12 2. Sommerset I, Walde C, Bang Jensen B, Bornø B, Haukaas A, Brun E. Fiske- Because this implant measures both ECG and PPG, co- helserapporten 2019. report 5a/2020 (Norwegian Veterinary Institute, processing of ECG and PPG for an even more robust HR Oslo, Norway, 2019) (2020) estimate is made possible. This is already a topic in human 3. Noble C, Nilsson J, Stien LH, Iversen MH, Kolarevic J, Gismervik K. Velferdsindikatorer for oppdrettslaks: Hvordan vurdere og dokumentere medicine to improve accuracy and robustness for heart fiskevelferd. 2. utgave (2018) rate variabilty estimates beyond what is currently possible 4. Føre M, Frank K, Norton T, Svendsen E, Alfredsen JA, Dempster T, Eguiraun using ECG alone [48]. H, Watson W, Stahl A, Sunde LM, et al. Precision fish farming: a new frame - work to improve production in aquaculture. Biosys Eng. 2017. https:// doi. Acknowledgements org/ 10. 1016/j. biosy stems eng. 2017. 10. 014. The authors thank the Research Council of Norway for funding the work (see 5. Knudsen F, Fosseidengen J, Oppedal F, Karlsen Ø, Ona E. Hydroacoustic “Funding”) as well as SINTEF Ocean AS for supporting the PhD project through monitoring of fish in sea cages: target strength (ts) measurements on their in-kind contribution. We also thank the personnel at NINA’s research sta- Atlantic salmon (Salmo salar). Fish Res. 2004;69(2):205–9. tion at Ims for facilitating the experiment. 6. Berckmans D. General introduction to precision livestock farming. Anim Front. 2017;7(1):6–11. Authors’ contributions 7. Føre M, Svendsen E, Alfredsen JA, Uglem I, Bloecher N, Sveier H, Sunde ES and JAA conceived the idea for the setup and testing the sensing principle LM, Frank K. Using acoustic telemetry to monitor the effects of crowding in fish. ES is responsible for the completion of the PhD in the funding project, and delousing procedures on farmed Atlantic salmon (Salmo salar). designed the required equipment, planned and executed the experiment, Aquaculture. 2018;495:757–65. prepared and processed data, and did the main job in preparing the manu- 8. Brijs J, Sandblom E, Axelsson M, Sundell K, Sundh H, Huyben D, Broström script. FØ participated in the planning and execution of the experiments with R, Kiessling A, Berg C, Gräns A. The final countdown: continuous physi- emphasis on anesthesia and surgery. MF, JAA assisted in equipment design ological welfare evaluation of farmed fish during common aquaculture and planning of the experiment. LLR has participated in data processing, practices before and during harvest. Aquaculture. 2018;495:903–911 while BF and REO have contributed with the anatomical and physiological 9. Hassan, W., Føre, M., Pedersen, M.O., Alfredsen, J.A.: A novel doppler based considerations. All authors have provided feedback and contributed to writing speed measurement technique for individual free-ranging fish. In: 2019 the manuscript. All authors read and approved the final manuscript. IEEE SENSORS, pp. 1–4 (2019). IEEE 10. Baktoft H, Gjelland KØ, Økland F, Thygesen UH. Positioning of aquatic Funding animals based on time-of-arrival and random walk models using yaps This study was funded by the Research Council of Norway (RCN Project Num- (yet another positioning solver). Sci Rep. 2017;7(1):1–10. ber 280864) and through in-kind from SINTEF Ocean. 11. Hassan W, Føre M, Urke HA, Kristensen T, Ulvund JB, Alfredsen JA, et al. System for real-time positioning and monitoring of fish in commercial Availability of data and materials marine farms based on acoustic telemetry and internet of fish (iof ). In: The data sets used and/or analysed during the current study are available The 29th International Ocean and Polar Engineering Conference (2019). from the corresponding author on reasonable request. International Society of Oshor ff e and Polar Engineers 12. Thorstad EB, Rikardsen AH, Alp A, Økland F. The use of electronic tags in fish research–an overview of fish telemetry methods. Turk J Fish Aquat Declarations Sci. 2013;13(5):881–96. 13. Deng ZD, Weiland MA, Fu T, Seim TA, LaMarche BL, Choi EY, Carlson Ethics approval and consent to participate TJ, Eppard MB. A cabled acoustic telemetry system for detecting and The animal experiment described herein was conducted with permission tracking juvenile salmon: Part 2. Three-dimensional tracking and passage given by the Norwegian Food Safety Authority (permit number 19/104024). outcomes. Sensors. 2011;11(6):5661–76. Doi: https:// doi. org/ 10. 3390/ Permission to use up to, but not exceeding, 15 individual Atlantic salmon was s1106 05661 given under the condition that data collection was stopped when 5 complete 14. Weiland MA, Deng ZD, Seim TA, LaMarche BL, Choi EY, Fu T, Carlson TJ, data sets were obtained as a reduction measure (3R) [49], a point that was Thronas AI, Eppard MB. A cabled acoustic telemetry system for detecting reached after 7 fish. The experiment was carried out at NINA Ims Research and tracking juvenile salmon: Part 1. Engineering design and instrumen- Station close to Stavanger, Norway, on August 28th, 2019. tation. Sensors. 2011;11(6):5645–60. Doi: https:// doi. org/ 10. 3390/ s1106 Consent for publication 15. Hussey NE, Kessel ST, Aarestrup K, Cooke SJ, Cowley PD, Fisk AT, Not applicable. Harcourt RG, Holland KN, Iverson SJ, Kocik JF et al. Aquatic animal telemetry: a panoramic window into the underwater world. Science. Competing interests 2015;348(6240):1255642 The authors declare that they have no competing interests. 16. Muñoz L, Aspillaga E, Palmer M, Saraiva JL, Arechavala-Lopez P. Acoustic telemetry: a tool to monitor fish swimming behavior in sea-cage aqua- Author details culture. Front Marine Sci. 2020;7: 645 Department of Engineering Cybernetics, NTNU, O.S. Bragstads Plass 2D, 17. Føre M, Frank K, Dempster T, Alfredsen JA, Høy E. Biomonitoring using 7034 Trondheim, Norway. Department of Aquaculture Technology, SINTEF tagged sentinel fish and acoustic telemetry in commercial salmon aqua- Ocean, Brattørkaia 17C, 7010 Trondheim, Norway. Department of Electronic culture: a feasibility study. Aquacult Eng. 2017;78:163–72. Systems, O.S. Bragstads Plass 2B, 7034 Trondheim, Norway. Nor wegian 18. Wendelaar Bonga SE. The stress response in fish. Physiol Rev. Institute for Nature Research, Høgskoleringen 9, 7485 Trondheim, Norway. 1997;77(3):591–625. Department of Biology, NTNU, Brattørkaia 17B, 7010 Trondheim, Norway. 19. Hvas M, Folkedal O, Oppedal F. Heart rate bio-loggers as welfare indicators in Atlantic salmon (Salmo salar) aquaculture. Aquaculture. Received: 6 May 2021 Accepted: 13 September 2021 2020;529:735630. 20. Svendsen E, Føre M, Økland F, Gräns A, Hedger RD, Alfredsen JA, Uglem I, Rosten C, Frank K, Erikson U et al. Heart rate and swimming activity as stress indicators for Atlantic salmon (Salmo salar). Aquaculture. 2020;531: References 21. Brijs J, Sandblom E, Rosengren M, Sundell K, Berg C, Axelsson M, Gräns A. 1. FAO: Food and Agriculture Organization of the United Nations, Fisheries Prospects and pitfalls of using heart rate bio-loggers to assess the welfare and Aquaculture department (2021). http:// www. fao. org/ fishe ry/ aquac of rainbow trout (Oncorhynchus mykiss) in aquaculture. Aquaculture. ulture/ en Accessed 2 May 2021 2019;509:188–197 Svendsen et al. Anim Biotelemetry (2021) 9:41 Page 12 of 12 22. Charlton PH, Birrenkott DA, Bonnici T, Pimentel MA, Johnson AE, Alastruey 34. Coyle SD, Durborow RM, Tidwell JH, et al. Anesthetics in aquaculture vol. J, Tarassenko L, Watkinson PJ, Beale R, Clifton DA. Breathing rate estima- 3900. Southern Regional Aquaculture Center, Stoneville (2004) tion from the electrocardiogram and photoplethysmogram: a review. 35. Elgendi M. Optimal signal quality index for photoplethysmogram signals. IEEE Rev Biomed Eng. 2017;11:2–20. Bioengineering. 2016;3(4):21. https:// doi. org/ 10. 3390/ bioen ginee ring3 23. Severinghaus JW. Takuo aoyagi: discovery of pulse oximetry. Anesth 040021. Analg. 2007;105(6):1–4. 36. Doving K, Reimers E. Fiskens Fysiologi. Bergen: John Grieg; 1992 24. Chan ED, Chan MM, Chan MM. Pulse oximetry: understanding its 37. Lucas M. Heart rate as an indicator of metabolic rate and activity in adult basic principles facilitates appreciation of its limitations. Respir Med. Atlantic salmon, Salmo salar. J Fish Biol. 1994;44(5):889–903. 2013;107(6):789–99. 38. Butterworth C. Filter approximation theory. Engineer. 1930;7:536–41. 25. Islam MT, Zabir I, Ahamed ST, Yasar MT, Shahnaz C, Fattah SA. A time- 39. Savitzky A, Golay MJ. Smoothing and differentiation of data by simplified frequency domain approach of heart rate estimation from photoplethys- least squares procedures. Anal Chem. 1964;36(8):1627–39. mographic (ppg) signal. Biomed Signal Process Control. 2017;36:146–54. 40. Farrell AP. Encyclopedia of fish physiology: from genome to environment. 26. Erhardt W, Lendl C, Hipp R, von Hegel G, Wiesner G, Wiesner H. The use London: Academic press; 2011. of pulse oximetry in clinical veterinary anaesthesia. J Assoc Vet Anaesth 41. Johns M, Giller CA, German DC, Liu H. Determination of reduced scatter- Great Br Irel. 1990;17(1):30–1. https:// doi. org/ 10. 1111/j. 1467- 2995. 1990. ing coefficient of biological tissue from a needle-like probe. Opt Express. tb003 85.x. 2005;13(13):4828–42. 27. Theodor M, Ruh D, Subramanian S, Förster, K, Heilmann C, Beyersdorf F, 42. Bigio, I.J., Fantini, S.: Quantitative Biomedical Optics: Theory, Methods, Plachta D, Manoli Y, Zappe H, Seifert A. Implantable pulse oximetry on and Applications. Cambridge University Press, University Printing Press, subcutaneous tissue. In: 2014 36th Annual International Conference of Cambridge CB2 8BS, United Kingdom (2016) the IEEE Engineering in Medicine and Biology Society, pp. 2089–2092 43. Nitzan M, Khanokh B, Slovik Y. The difference in pulse transit time to (2014). IEEE the toe and finger measured by photoplethysmography. Physiol Meas. 28. Reynolds J, Ahmmed P, Bozkurt A. An injectable system for subcutaneous 2001;23(1):85. photoplethysmography, accelerometry, and thermometry in animals. 44. Golyandina N, Zhigljavsky A. Singular spectrum analysis for time series, IEEE Trans Biomed Circuits Syst. 2019;13(5):825–34. vol. 120. Heidelberg New York Dordrech London: Springer; 2013. 29. Machikhin AS, Burlakov AB, Volkov MV, Khokhlov DD. Imaging pho- 45. Meyer Y. Wavelets: algorithms & applications. Philadelphia: SIAM (Society toplethysmography and videocapillaroscopy enable noninvasive for Industrial and Applied Mathematics; 1993. study of zebrafish cardiovascular system functioning. J Biophotonics. 46. Cooley JW, Tukey JW. An algorithm for the machine calculation of com- 2020;13(7):202000061. plex fourier series. Math Comput. 1965;19(90):297–301. 30. Seo E, Sazi T, Togawa M, Nagata O, Murakami M, Kojima S, Seo Y. A port- 47. Svendsen E, Føre M, Randeberg L, JA,. Design of a novel biosensor able infrared photoplethysmograph: heartbeat of Mytilus galloprovincialis implant for farmed Atlantic salmon (Salmo salar). In: 2021 IEEE SENSORS, analyzed by mri and application to bathymodiolus septemdierum. Biol Accepted for Publication (2021). IEEE Open. 2016;5(11):1752–7. 48. Zaretskiy A, Mityagin K, Tarasov V, Moroz D, Kuraleva A. Robust heart 31. Tomlinson S, Behrmann S, Cranford J, Louie M, Hashikawa A. Accuracy rate estimation using combined ecg and ppg signal processing. In: IOP of smartphone-based pulse oximetry compared with hospital-grade Conference Series: Materials Science and Engineering, vol. 537, p. 042077 pulse oximetry in healthy children. Telemed e-Health. 2018;24(7):527–35. (2019). IOP Publishing https:// doi. org/ 10. 1089/ tmj. 2017. 0166. 49. Norecopa: Norecopa, Norway’s National Consensus Platform (2021). 32. Hahnen C, Freeman CG, Haldar N, Hamati JN, Bard DM, Murali V, Merli GJ, https:// norec opa. no/ Accessed 10 Mar 2021 Joseph JI, van Helmond N. Accuracy of vital signs measurements by a smartwatch and a portable health device: validation study. JMIR Mhealth Publisher’s Note Uhealth. 2020;8(2):16811. Springer Nature remains neutral with regard to jurisdictional claims in pub- 33. Maxim Integrated: Maxim Integrated, Integrated Photoplethysmogram lished maps and institutional affiliations. and Electrocardiogram Bio-sensor Module for Mobile Health (2021). https:// datas heets. maxim integ rated. com/ en/ ds/ MAX86 150. pdf Accessed 9 Aug 2021. Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions

Journal

Animal BiotelemetrySpringer Journals

Published: Sep 22, 2021

Keywords: Salmo salar; Heart rate; Implant; Biosensors; Photoplethysmography

There are no references for this article.