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

Learn More →

Computed fiber evaluation of SEM images using DiameterJ

Computed fiber evaluation of SEM images using DiameterJ Fiber materials offer a high potential for improving the surface characteristics of medical implants. For quality 1 Introduction assurance of nano- and microfiber structures the morphology is inspected by Scanning Electron Microscopy (SEM) as a Electrospinning is a widely established method of producing standard method. Vast quantities of image data have to be micro- to nanofibers. The statistical distribution of fiber evaluated. Usual practice for obtaining the fiber diameters is diameters comprises information about process stability and the manually setting of measurement points. The software the effect of influencing factors. Scanning Electron DiameterJ which runs as plugin in ImageJ automatically Microscopy is the standard technique for monitoring the fiber computes fiber diameters. Here we investigated its morphology of nonwoven structures. State of the art for the capabilities and limitations by comparing the evaluation of evaluation of fiber diameters is the manual determination of selected sample SEM images of electrospun fibers. In this an adequate number of measurement points in the image [1, study the fibers of three examplary images specified by 2]. That practice reveals two shortcomings. The first one is different contrast and fiber morphology were analyzed by the personal influence: an operator selects measuring points using varied segmentation algorithms. The results are based on feeling and experience. Thus, different persons may displayed in bar charts of frequency distribution. determine different fiber diameters. The second shortcoming Additionally the computed fiber diameters were compared to of that time consuming method is the limited number of manual measurements. Depending on various image points. For reliable statistics high quantities of values are properties the segmentation process works more or less necessary. reliable, and fault data of incomplete segmented fibers are A promising alternative are computerized evaluation computed. Often the results are eligible, but frequently procedures, which automatically generate plentiful data. DiameterJ generates data resembling to thin fibers, which are DiameterJ as plugin for the extensively used software ImageJ not present in the image. In some cases the peaks of fault data is such a tool [3]. It is specially designed for measuring fibers are much higher than peaks of real fibers. In consequence in SEM images and was validated on special test images as misinterpretation of data cannot be avoided. DiameterJ is a well as SEM images of steel wires and PLGA fibers [4]. That validated tool with the ability to generate reliable results. software tool uses different algorithms and gray scale values Future work on improving the segmentation algorithms can to binarize the fiber structure and consequently separates refine computed evaluation. fibers from background. The user has to choose one of Keywords: automated, calculation, fiber diameter, ImageJ, different results of segmentation provided by the software. In nanofiber, microfiber, electrospinning further processing the fiber, structures are skeletonized and the fiber radii, pores, angles and other data are calculated. To study the capabilities and limitations of DiameterJ for fiber evaluation we investigated numerous images. Three ______ examples are shown here: one image with reliable results *Corresponding author: Andreas Götz: Institute for Biomedical compared to two images which reveal the limits of automated Engineering, Rostock University Medical Center, Friedrich- Barnewitz-Str. 4, 18119 Rostock, Germany, andreas.goetz@uni- computation. rostock.de Volkmar Senz, Sabine Illner, Niels Grabow: Institute for Biomedical Engineering, University Medical Center Rostock, Rostock, Germany Open Access. © 2020 Andreas Götz et al., published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0 License. Andreas Götz et al., Computed fiber evaluation of SEM images using DiameterJ — 2 generated which appear as fibers with small diameters. It is 2 Material and Methods not possible for the user to distinguish between fault data and real small fibers. We see the manual measurement as reliable Three groups of representative SEM images of clearly visible reference for data comparison. For image A the segmentation electrospun fibers and corresponding segmented images were algorithms M and S generated a result which fits good to the evaluated: A: gray fibers and dark background, B: gray fibers manually measured values, whereas algorithm T generated on gray background, and C: gray fibers of different some thin white structures in the segmentation image thicknesses on dark background. All image processing was resulting as a half of maximum sized peak at around 0.25 µm done using ImageJ version 1.52a, the computed evaluation fiber diameter. Image B clearly reveals the limits of was done using the plugin DiameterJ version 1-018. Three DiameterJ. Algorithm M detected half of the fibers different segmentation algorithms were performed on each incompletely resulting in a peak for thin fibers. That peak of image to compare the results: M (Mixed), S (Stat. Region fault data is 28% higher than the peak of the actual fiber Merged), and T (Traditional). DiameterJ generates eight diameter. The algorithms S and T show inferior results. Due different separation images of each algorithm. The operator to a high fraction of faulty segmented fibers the fault peaks has to choose “the best one” for further computation. For a are 7 times and 18 times higher than those of actual fibers, trusty comparableness segmented images of the same order respectively. The clearest difference between manually and of each algorithm were chosen, image A: M3, S3, T3, image automated evaluation reveals image C. Only thin fibers (< 1 B: M1, S1, T1, image C: M7, S7, T7. As reference, 50 µm) are detected by DiameterJ. Existing thick fibers (1.5-3.5 manually set measurement points on each image were µm) remained undetected because they were not segmented statistically processed. in full diameter. Remarkably the shape of the peaks of thin fibers and fault data seem to resemble each other: high-angle steepness at the left flank and a less inclined steepness 3 Results towards increasing diameters, hence it is impossible to discriminate between real thin fibers and fault data. In most The results of the evaluation of three SEM images are cases automated evaluations of daily routine images generate displayed in Figure 1. In the left row images and frequency peaks with different magnitudes of nonexistent fiber distribution of manually chosen fiber diameters are shown as diameters. Thus, the rate of false positive results is relatively red colored bar charts. The segmented images using the high. In case there are no thin fibers in the image, the algorithms M, S, and T as well as the corresponding corresponding peaks can be ignored. For evaluation of frequency distributions are shown in the other rows as blue numerous images DiameterJ then is advantageous. In Table 1 colored bar charts. Average fiber diameters in image A were the main characteristics of both methods, manual and found as 0.81 ± 0.23 µm for manually measurement and 0.79 automatical evaluation, are compared. ± 0.25 µm, 0.77 ± 0.26 µm and 0.69 ± 0.35 µm for algorithms M, S and T, respectively. In image B the manually Table 1: Evaluation methods, comparison of core characteristics measured fiber diameters are 1.55 ± 0.14 µm, whereas in image C the thin fibers are about 0.2 µm and the thick fibers Methods Advantages Disadvantages are in a range from 1.5 to 3.5 µm. A remarkable difference between manual counts (50 in total) and automatically manually measured exact and easy, subjective, standard software limited number of generated counts (hundreds or thousands) is clearly visible. measuring points automatically high amount of data fault data, computed using fully computable, misinterpretation 4 Discussion DiameterJ user-independent We compared three segmentation algorithms in ImageJ for As the separation of thin fibers located next to thicker ones is three characteristic SEM images of electrospun fibers. No not executed properly, the use of DiameterJ is not doubt, DiameterJ is a supportive and validated tool for the recommended in those cases. If only thin fibers are present in evaluation of SEM images and generates beneficial data. the image, the separated image has to be reviewed Because of the circumstance that in several SEM images thoroughly, including a check for plausibility of the final fibers are not mapped faultlessly, the segmentation process results. does not work correctly. In consequence fault data are A Andreas Götz et al., Computed fiber evaluation of SEM images using DiameterJ — 3 SEM Images Algorithm M Algorithm S Algorithm T 1200 1200 1200 1000 1000 1000 6 800 800 800 600 600 600 3 400 400 400 200 200 200 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] 2000 2000 700 1800 1800 1600 1600 8 600 1400 1400 1200 1200 1000 1000 800 800 4 300 600 600 400 400 200 200 0 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] 10 7000 7000 7000 6000 6000 6000 5000 5000 5000 4000 4000 4000 3000 3000 3000 2000 2000 2000 1000 1000 1000 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] Counts manually Counts manually Counts manually Counts M Counts M Counts M Counts S Counts S Counts S Counts T Counts T Counts T Computed fiber evaluation of SEM images using DiameterJ — 4 Figure 1: SEM images of electrospun fibers, scale bar 10 µm; Images computed by DiameterJ performing the segmentation algorithms Mixed (M), Stat. Region Merged (S), and Traditional (T); The images represent specific properties - A: good evaluable image for reliable results, B: image with less background contrast but good visible fibers, C: image with highly different fibers on dark background. In rows images and corresponding bar charts of diameter distribution are displayed. The SEM image row shows the SEM images with 50 manually set measurement points. The algorithm rows show the computed results of segmented images using algorithm M, S and T, respectively. Author Statement 5 Summary and Outlook Research funding: Partial financial support by the Federal Ministry of Education and Research (BMBF) within We compared the manual evaluation method and the RESPONSE “Partnership for Innovation in Implant software tool DiameterJ to obtain diameters of electrospun Technology” and by the European Social Fund (ESF) within fibers in SEM images. Using three exemplary images we the excellence research program of the state Mecklenburg- showed the capability and limitations of DiameterJ. No Vorpommern Card-ii-Omics is gratefully acknowledged. segmentation algorithm revealed as the “best case”. Due to Conflict of interest: Authors state no conflict of interest. faulty segmentation, bar charts of fiber diameter distribution Informed consent: Informed consent is not applicable. Ethical showed peaks of nonexistent fibers at one hand and non- approval: The conducted research is not related to either detected fibers at the other hand. A good image contrast is no human or animal use. surety for reliable results, the fiber surface seems to strongly influence the results. Performing DiameterJ on numerous SEM images of daily routine revealed the necessity for References improving automated software tools. [1] Wang, Bo; Cai, Qing; Zhang, Shen; Yang, Xiaoping; Deng, Certainly, no software can chose measurement points Xuliang: The effect of poly (L-lactic acid) nanofiber with a similar intention like an individual person, fault data orientation on osteogenic responses of human osteoblast- are unavoidable. The main advantage of automated like MG63 cells, Bd. 4, S. 600–609. evaluation is the processing of vast amounts of image data [2] Zhang, Kuihua; Zheng, Honghao; Liang, Su; Gao, Changyou: Aligned PLLA nanofibrous scaffolds coated with graphene less influenced by the user. To improve the correctness of oxide for promoting neural cell growth, Bd. 37, S. 131–142. automatically generated data the focus is to set on better [3] Hotaling, Nathan A.; Bharti, Kapil; Kriel, Haydn; Simon, Carl performing segmentation algorithms. Further effort can be G. (2015): DiameterJ: A validated open source nanofiber directed at thresholds of gray scale values or on fiber surface. diameter measurement tool. In: Biomaterials 61, S. 327–338. DOI: 10.1016/j.biomaterials.2015.05.015. However, refining the segmentation algorithms is a [4] Hotaling, Nathan A.; Bharti, Kapil; Kriel, Haydn; Simon, Carl promising approach for future works. G. (2015): Dataset for the validation and use of DiameterJ an open source nanofiber diameter measurement tool. In: Biomaterials 61, S. 327–338. DOI: 10.1016/j.biomaterials.2015.05.015. Acknowledgment The authors would like to thank Jonathan Ortelt, Manfred Strotmeier, Babette Hummel and Katja Hahn for their skillful work. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Current Directions in Biomedical Engineering de Gruyter

Computed fiber evaluation of SEM images using DiameterJ

Download PDF
4 pages

Loading next page...
 
/lp/de-gruyter/computed-fiber-evaluation-of-sem-images-using-diameterj-C3c5b0UbTQ

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
de Gruyter
Copyright
© 2020 by Walter de Gruyter Berlin/Boston
eISSN
2364-5504
DOI
10.1515/cdbme-2020-3113
Publisher site
See Article on Publisher Site

Abstract

Fiber materials offer a high potential for improving the surface characteristics of medical implants. For quality 1 Introduction assurance of nano- and microfiber structures the morphology is inspected by Scanning Electron Microscopy (SEM) as a Electrospinning is a widely established method of producing standard method. Vast quantities of image data have to be micro- to nanofibers. The statistical distribution of fiber evaluated. Usual practice for obtaining the fiber diameters is diameters comprises information about process stability and the manually setting of measurement points. The software the effect of influencing factors. Scanning Electron DiameterJ which runs as plugin in ImageJ automatically Microscopy is the standard technique for monitoring the fiber computes fiber diameters. Here we investigated its morphology of nonwoven structures. State of the art for the capabilities and limitations by comparing the evaluation of evaluation of fiber diameters is the manual determination of selected sample SEM images of electrospun fibers. In this an adequate number of measurement points in the image [1, study the fibers of three examplary images specified by 2]. That practice reveals two shortcomings. The first one is different contrast and fiber morphology were analyzed by the personal influence: an operator selects measuring points using varied segmentation algorithms. The results are based on feeling and experience. Thus, different persons may displayed in bar charts of frequency distribution. determine different fiber diameters. The second shortcoming Additionally the computed fiber diameters were compared to of that time consuming method is the limited number of manual measurements. Depending on various image points. For reliable statistics high quantities of values are properties the segmentation process works more or less necessary. reliable, and fault data of incomplete segmented fibers are A promising alternative are computerized evaluation computed. Often the results are eligible, but frequently procedures, which automatically generate plentiful data. DiameterJ generates data resembling to thin fibers, which are DiameterJ as plugin for the extensively used software ImageJ not present in the image. In some cases the peaks of fault data is such a tool [3]. It is specially designed for measuring fibers are much higher than peaks of real fibers. In consequence in SEM images and was validated on special test images as misinterpretation of data cannot be avoided. DiameterJ is a well as SEM images of steel wires and PLGA fibers [4]. That validated tool with the ability to generate reliable results. software tool uses different algorithms and gray scale values Future work on improving the segmentation algorithms can to binarize the fiber structure and consequently separates refine computed evaluation. fibers from background. The user has to choose one of Keywords: automated, calculation, fiber diameter, ImageJ, different results of segmentation provided by the software. In nanofiber, microfiber, electrospinning further processing the fiber, structures are skeletonized and the fiber radii, pores, angles and other data are calculated. To study the capabilities and limitations of DiameterJ for fiber evaluation we investigated numerous images. Three ______ examples are shown here: one image with reliable results *Corresponding author: Andreas Götz: Institute for Biomedical compared to two images which reveal the limits of automated Engineering, Rostock University Medical Center, Friedrich- Barnewitz-Str. 4, 18119 Rostock, Germany, andreas.goetz@uni- computation. rostock.de Volkmar Senz, Sabine Illner, Niels Grabow: Institute for Biomedical Engineering, University Medical Center Rostock, Rostock, Germany Open Access. © 2020 Andreas Götz et al., published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0 License. Andreas Götz et al., Computed fiber evaluation of SEM images using DiameterJ — 2 generated which appear as fibers with small diameters. It is 2 Material and Methods not possible for the user to distinguish between fault data and real small fibers. We see the manual measurement as reliable Three groups of representative SEM images of clearly visible reference for data comparison. For image A the segmentation electrospun fibers and corresponding segmented images were algorithms M and S generated a result which fits good to the evaluated: A: gray fibers and dark background, B: gray fibers manually measured values, whereas algorithm T generated on gray background, and C: gray fibers of different some thin white structures in the segmentation image thicknesses on dark background. All image processing was resulting as a half of maximum sized peak at around 0.25 µm done using ImageJ version 1.52a, the computed evaluation fiber diameter. Image B clearly reveals the limits of was done using the plugin DiameterJ version 1-018. Three DiameterJ. Algorithm M detected half of the fibers different segmentation algorithms were performed on each incompletely resulting in a peak for thin fibers. That peak of image to compare the results: M (Mixed), S (Stat. Region fault data is 28% higher than the peak of the actual fiber Merged), and T (Traditional). DiameterJ generates eight diameter. The algorithms S and T show inferior results. Due different separation images of each algorithm. The operator to a high fraction of faulty segmented fibers the fault peaks has to choose “the best one” for further computation. For a are 7 times and 18 times higher than those of actual fibers, trusty comparableness segmented images of the same order respectively. The clearest difference between manually and of each algorithm were chosen, image A: M3, S3, T3, image automated evaluation reveals image C. Only thin fibers (< 1 B: M1, S1, T1, image C: M7, S7, T7. As reference, 50 µm) are detected by DiameterJ. Existing thick fibers (1.5-3.5 manually set measurement points on each image were µm) remained undetected because they were not segmented statistically processed. in full diameter. Remarkably the shape of the peaks of thin fibers and fault data seem to resemble each other: high-angle steepness at the left flank and a less inclined steepness 3 Results towards increasing diameters, hence it is impossible to discriminate between real thin fibers and fault data. In most The results of the evaluation of three SEM images are cases automated evaluations of daily routine images generate displayed in Figure 1. In the left row images and frequency peaks with different magnitudes of nonexistent fiber distribution of manually chosen fiber diameters are shown as diameters. Thus, the rate of false positive results is relatively red colored bar charts. The segmented images using the high. In case there are no thin fibers in the image, the algorithms M, S, and T as well as the corresponding corresponding peaks can be ignored. For evaluation of frequency distributions are shown in the other rows as blue numerous images DiameterJ then is advantageous. In Table 1 colored bar charts. Average fiber diameters in image A were the main characteristics of both methods, manual and found as 0.81 ± 0.23 µm for manually measurement and 0.79 automatical evaluation, are compared. ± 0.25 µm, 0.77 ± 0.26 µm and 0.69 ± 0.35 µm for algorithms M, S and T, respectively. In image B the manually Table 1: Evaluation methods, comparison of core characteristics measured fiber diameters are 1.55 ± 0.14 µm, whereas in image C the thin fibers are about 0.2 µm and the thick fibers Methods Advantages Disadvantages are in a range from 1.5 to 3.5 µm. A remarkable difference between manual counts (50 in total) and automatically manually measured exact and easy, subjective, standard software limited number of generated counts (hundreds or thousands) is clearly visible. measuring points automatically high amount of data fault data, computed using fully computable, misinterpretation 4 Discussion DiameterJ user-independent We compared three segmentation algorithms in ImageJ for As the separation of thin fibers located next to thicker ones is three characteristic SEM images of electrospun fibers. No not executed properly, the use of DiameterJ is not doubt, DiameterJ is a supportive and validated tool for the recommended in those cases. If only thin fibers are present in evaluation of SEM images and generates beneficial data. the image, the separated image has to be reviewed Because of the circumstance that in several SEM images thoroughly, including a check for plausibility of the final fibers are not mapped faultlessly, the segmentation process results. does not work correctly. In consequence fault data are A Andreas Götz et al., Computed fiber evaluation of SEM images using DiameterJ — 3 SEM Images Algorithm M Algorithm S Algorithm T 1200 1200 1200 1000 1000 1000 6 800 800 800 600 600 600 3 400 400 400 200 200 200 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] 2000 2000 700 1800 1800 1600 1600 8 600 1400 1400 1200 1200 1000 1000 800 800 4 300 600 600 400 400 200 200 0 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] 10 7000 7000 7000 6000 6000 6000 5000 5000 5000 4000 4000 4000 3000 3000 3000 2000 2000 2000 1000 1000 1000 0 0 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] fiber diameter [µm] Counts manually Counts manually Counts manually Counts M Counts M Counts M Counts S Counts S Counts S Counts T Counts T Counts T Computed fiber evaluation of SEM images using DiameterJ — 4 Figure 1: SEM images of electrospun fibers, scale bar 10 µm; Images computed by DiameterJ performing the segmentation algorithms Mixed (M), Stat. Region Merged (S), and Traditional (T); The images represent specific properties - A: good evaluable image for reliable results, B: image with less background contrast but good visible fibers, C: image with highly different fibers on dark background. In rows images and corresponding bar charts of diameter distribution are displayed. The SEM image row shows the SEM images with 50 manually set measurement points. The algorithm rows show the computed results of segmented images using algorithm M, S and T, respectively. Author Statement 5 Summary and Outlook Research funding: Partial financial support by the Federal Ministry of Education and Research (BMBF) within We compared the manual evaluation method and the RESPONSE “Partnership for Innovation in Implant software tool DiameterJ to obtain diameters of electrospun Technology” and by the European Social Fund (ESF) within fibers in SEM images. Using three exemplary images we the excellence research program of the state Mecklenburg- showed the capability and limitations of DiameterJ. No Vorpommern Card-ii-Omics is gratefully acknowledged. segmentation algorithm revealed as the “best case”. Due to Conflict of interest: Authors state no conflict of interest. faulty segmentation, bar charts of fiber diameter distribution Informed consent: Informed consent is not applicable. Ethical showed peaks of nonexistent fibers at one hand and non- approval: The conducted research is not related to either detected fibers at the other hand. A good image contrast is no human or animal use. surety for reliable results, the fiber surface seems to strongly influence the results. Performing DiameterJ on numerous SEM images of daily routine revealed the necessity for References improving automated software tools. [1] Wang, Bo; Cai, Qing; Zhang, Shen; Yang, Xiaoping; Deng, Certainly, no software can chose measurement points Xuliang: The effect of poly (L-lactic acid) nanofiber with a similar intention like an individual person, fault data orientation on osteogenic responses of human osteoblast- are unavoidable. The main advantage of automated like MG63 cells, Bd. 4, S. 600–609. evaluation is the processing of vast amounts of image data [2] Zhang, Kuihua; Zheng, Honghao; Liang, Su; Gao, Changyou: Aligned PLLA nanofibrous scaffolds coated with graphene less influenced by the user. To improve the correctness of oxide for promoting neural cell growth, Bd. 37, S. 131–142. automatically generated data the focus is to set on better [3] Hotaling, Nathan A.; Bharti, Kapil; Kriel, Haydn; Simon, Carl performing segmentation algorithms. Further effort can be G. (2015): DiameterJ: A validated open source nanofiber directed at thresholds of gray scale values or on fiber surface. diameter measurement tool. In: Biomaterials 61, S. 327–338. DOI: 10.1016/j.biomaterials.2015.05.015. However, refining the segmentation algorithms is a [4] Hotaling, Nathan A.; Bharti, Kapil; Kriel, Haydn; Simon, Carl promising approach for future works. G. (2015): Dataset for the validation and use of DiameterJ an open source nanofiber diameter measurement tool. In: Biomaterials 61, S. 327–338. DOI: 10.1016/j.biomaterials.2015.05.015. Acknowledgment The authors would like to thank Jonathan Ortelt, Manfred Strotmeier, Babette Hummel and Katja Hahn for their skillful work.

Journal

Current Directions in Biomedical Engineering – de Gruyter

Published: Sep 1, 2020

Keywords: automated; calculation; fiber diameter; ImageJ; nanofiber; microfiber; electrospinning

There are no references for this article.