Prediction Model of Adverse Effects on Liver Functions of COVID-19 ICU Patients
Prediction Model of Adverse Effects on Liver Functions of COVID-19 ICU Patients
Mashraqi, Aisha;Halawani, Hanan;Alelyani, Turki;Mashraqi, Mutaib;Makkawi, Mohammed;Alasmari, Sultan;Shaikh, Asadullah;Alshehri, Ahmad
2022-04-25 00:00:00
Hindawi Journal of Healthcare Engineering Volume 2022, Article ID 4584965, 11 pages https://doi.org/10.1155/2022/4584965 Research Article Prediction Model of Adverse Effects on Liver Functions of COVID-19 ICU Patients 1 1 1 2 Aisha Mashraqi , Hanan Halawani , Turki Alelyani , Mutaib Mashraqi , 3 3 1 Mohammed Makkawi , Sultan Alasmari , Asadullah Shaikh , and Ahmad Alshehri College of Computer Science and Information Systems, Najran University, Najran, Saudi Arabia Department of Clinical Laboratory Sciences, College of Applied Medical Sciences, Najran University, Najran, Saudi Arabia Faculty of Applied Medical Sciences, King Khalid University, Abha, Saudi Arabia Correspondence should be addressed to Hanan Halawani; hthalawani@nu.edu.sa Received 3 November 2021; Accepted 9 April 2022; Published 25 April 2022 Academic Editor: Kuruva Lakshmanna Copyright © 2022 Aisha Mashraqi et al. �is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. SARS-CoV-2 is a recently discovered virus that poses an urgent threat to global health. �e disease caused by this virus is termed COVID-19. Death tolls in di‰erent countries remain to rise, leading to continuous social distancing and lockdowns. Patients of di‰erent ages are susceptible to severe disease, in particular those who have been admitted to an ICU. Machine learning (ML) predictive models based on medical data patterns are an emerging topic in areas such as the prediction of liver diseases. Prediction models that combine several variables or features to estimate the risk of people being infected or experiencing a poor outcome from infection could assist medical sta‰ in the treatment of patients, especially those that develop organ failure such as that of the liver. In this paper, we propose a model called the detecting model for liver damage (DMLD) that predicts the risk of liver damage in COVID-19 ICU patients. �e DMLD model applies machine learning algorithms in order to assess the risk of liver failure based on patient data. To assess the DMLD model, collected data were preprocessed and used as input for several classi”ers. SVM, decision tree (DT), Na¨ıve Bayes (NB), KNN, and ANN classi”ers were tested for performance. SVM and DTperformed the best in terms of predicting illness severity based on laboratory testing. already su‰er from liver diseases, such as cirrhosis, are at a 1. Introduction higher risk of decompensation and death during COVID-19 �e COVID-19 pandemic was declared a health emergency infection [7]. Organ failure is serious; therefore, managing in 2020. Many people have died during the pandemic, infection is of interest. particularly in the early stages, due to a lack of under- �e liver is a vital organ, and its failure could be fatal. standing of the virus. COVID-19 has led to over 3.5 million COVID-19 patients can have mild to severe symptoms and deaths worldwide [1–3]. Patients infected with COVID-19 may develop acute hepatic failure [6]. According to the may experience no symptoms or severe illness that can lead proposed mechanism, hepatic failure occurs due to multiple to death [4]. �e virus continues to evolve, with concerning factors. �ese include angiotensin-converting enzyme 2 mutants emerging all over the world [5]. �is is an alarming (ACE2), a SARS-CoV-2 receptor found in multiple organs situation and requires a better understanding of the disease including the liver, and cytokine storm, which occurs as a in order to save more lives. Critical cases of COVID-19 could result of in¥ammatory mediators, endothelial dysfunction, coagulation abnormalities, and in¥ammatory cell in”ltration result in organ failure and death. Lung failure is the most common complication, but other organs can also be a‰ected into the organs [6]. Direct cytotoxicity caused by active virus by the virus. In fact, multiorgan failure involving the lungs, replication in the liver could result in liver cell damage. kidneys, liver, cardiovascular system, and gastrointestinal Furthermore, hypoxic liver damage is exacerbated by severe tract (GIT) can also occur [6]. Additionally, people who lung failure and disease. Cardiac congestion as a result of 2 Journal of Healthcare Engineering Machine learning (ML) is being introduced to medicine SARS-CoV-2 disease-induced right-sided heart failure can also result in liver damage. Furthermore, people with pre- and used as artificial intelligence (AI) to create predictive models based on data patterns. Machine learning can also be existing liver disease, as well as drug-induced liver injury, experience exacerbation [8]. To avoid COVID-19 disease used to create a predictive model of liver involvement [21]. complications, it is critical to detect liver damage early and Machine learning (ML) is currently being used to predict the understand its extent. possibility of fatty liver disease [22], the success of liver *e exact molecular mechanism of the above-mentioned transplants [23],andotherhepaticconditions. However,there hepatic injury is unknown. However, SARS-CoV-2 viral isstillnofirmagreementonwhichmachinelearningalgorithm RNA has been detected in liver tissue using qRT-PCR, in- is best to use as an illness-prediction method. *e outcome of patients with raised liver enzymes admitted to the ICU with dicating that the virus can affect liver cells [9]. It is still unclear where virus replication occurs in the liver, but an COVID-19 disease should be predicted using machine learning (ML), which could be useful in disease management. intact virus was found in the cytoplasm of COVID-19 pa- tients with abnormal liver function tests [10]. Viral receptors Millions of people have died as a result of the SARS- CoV-2 virus, and more people are becoming infected every have been found on the surface of host cells, which could explain the viral tropism towards the specific tissue. SARS- day. Elevated liver enzymes are linked to the severity of the CoV-2 enters the cell via the virus’s S protein, which binds to illness, which can be fatal. Early detection of disease warning host cell receptors such as ACE2 and TMPRSS2 [11]. *e signs, on the other hand, can be beneficial. During COVID- expression of ACE2 and TMPRSS2 receptors is low but still 19 disease, elevated liver enzymes are seen, and their level is presents in the hepatic cells [12]. Moreover, it is a note- related to the severity of the disease and the extent of liver worthy finding that the expression of ACE2 receptors is damage. *erefore, monitoring of liver enzymes in ICU SARS-CoV-2 patients can be used to improve their health. increased in both humans and mice with liver fibrosis [13]. Interestingly, hypoxic cases were found to be associated with Moreover, with the progress of machine learning toward improved screening methods for the severity of COVID-19 increased expression of ACE2 receptors, which could ex- plain the mechanism of ACE2 receptor upregulation in infection, the numbers of infected individuals have de- creased significantly, motivating artificial intelligence (AI) COVID-19 patients due to lung damage [13]. A variety of factors in SARS-CoV-2 infection can result scientists and medical physicians to employ this subject in hypoxia-induced liver damage. Heart failure, lung failure, more thoroughly in the health sector. Algorithms in ma- and sepsis are the three most serious of these. *ese factors chine learning are developed to allow computers to learn. account for 90% of all cases of hypoxic damage in COVID- ML algorithms can be used for classification problems, 19 cases. Moreover, right-sided heart failure causes liver which have been applied in the medical field to help in the congestion due to raised central venous pressure (CVP). early diagnosis of several diseases. However, there are specific difficulties with these computational methods, in- Hypoxia and liver congestion cause centrilobular necrosis over time [14]. Many known hepatotoxic agents have been cluding the feature-selection step in prediction models. Other studies have used a different methodology for feature used to treat COVID-19 disease. *ese drugs include cor- ticosteroids and antivirals. Corticosteroids have been found selection, such as a pivot table in [24] and a P-value in [25]. to cause steatosis, and hepatotoxicity is caused by antivirals In this paper, we propose a model to predict liver damage such as ritonavir and remdesivir [8]. based on data patterns using supervised learning techniques. Liver enzymes, which were found to be elevated in a *e model is named detecting model for liver damage number of COVID-19 cases, can be used to detect liver (DMLD), and it employs machine learning algorithms to damage. Although the incidence of liverinvolvement has been assist in the early detection of the risk of liver damage. It will reported in several COVID-19 cases, the extent of the prev- support healthcare professionals to diagnose the disease at alence of hepatic damage remains unknown [15]. Elevated its early stages. Data from blood tests of COVID-19 patients admitted to the ICU were collected, cleaned, and prepared to liver enzymes, particularly alanine aminotransferase (ALT) and aspartate aminotransferase (AST), have been reported in be used as input for the model. Secondly, we designed the DMLD model that prepares the data set in the preprocessing 14% to 53% of patients [16]. *ere is a strong correlation between the severity of the disease and the extent of liver phase by addressing the missing values and applying the involvement [16]. According to research, mild COVID-19 normalization approach. *en, the DMLD model identifies disease causes a mild elevation of liver enzymes, whereas the most relevant features in the feature-selection phase by severe disease causes a significantly higher level of liver en- applying a filtering method. Consequently, five machine zymes [16, 17]. In a study of 222 COVID-19 patients, 28.2% learning classifiers were examined in order to find the best- had elevated liver enzymes. *e reason for this elevation, performing algorithms; which are support vector machine (SVM), decision tree, Na¨ıve Bayes (NB), K-nearest neigh- however, was not specified, and it could have been preexisting [18]. Furthermore, a study of 417 COVID-19 patients dis- bors (KNN), and artificial neural network (ANN). *ese methods have certain drawbacks; for example, NB is simple covered that 76.3% of the total sample had abnormal liver function tests. During their hospital stay,21.5% suffereda liver and suitable for large data sets. However, it assumes that numeric properties have a normal distribution. Data injury. *eir levels of liver enzymes significantly increased within two weeks of hospitalization. According to the findings preparation is easier with the DT but is dependent on the of the study, patients with significantly elevated liver enzymes sequence of the characteristics. KNN, SVM, and ANN are are at a higher risk of developing severe disease [19, 20]. computationally expensive [26]. In our study, the Journal of Healthcare Engineering 3 performance of the DMLD model was evaluated on the collected data set, and the results show that the accuracy, precision, and recall of the SVM and DTclassifiers are better RESULTS than others. *erefore, we considered SVM and DT the Liver Damage COVID-19 Labeled DMLD likely best algorithms for detecting the risk of liver damage. No Liver Damage Observation Patinets Figure 1 illustrates the study framework. Figure 1: Study framework. *e rest of the paper is structured as follows. First, we present the related work. *en, we explain the DMLD prediction model in detail, describing the data set details and Deep learning has exploded significantly in scientific the DMLD stages with the classification algorithms. *en, computing, with its techniques being utilized by a variety of we present the results and discuss the performance of the fields to solve complicated problems. To perform certain DMLD model, including the measurement of classification tasks, all deep learning algorithms employ various forms of techniques. Finally, we provide the conclusions and identify neural networks. Neural networks are used in deep learning the future directions. to perform complex computations on massive amounts of data. It is a form of machine learning that is based on the 2. Related Work human brain’s structure and function. *e performance of classification is improved the most when the machine Machine learning approaches have attracted the attention of learning algorithm is updated with a deep learning algo- many researchers and have been applied in different dis- rithm. Over the last few years, there has been a lot of de- ciplines such as medicine, the economy, and education. velopment in the use of neural networks for feature Moreover, machine learning plays an essential role in the extraction in object identification problems. For example, medical field, contributing to various health sectors such as Zhang et al. created Deep-IRTarget, a unique backbone the early diagnosis of disease and treatment. Liver disease is a network composed of a frequency feature extractor, a spatial common health issue. *erefore, early diagnosis of the risk feature extractor, and a dual-domain feature resource al- factors will help medical physicians predict the development location model, to cope with challenges in feature extraction of the disease [27]. [40]. Moreover, the deep learning algorithm is employed in Ayeldeen et al. [28] highlighted that the positive pre- burnt area mapping with the use of Sentinel-12 data [41]. diction of different stages of liver fibrosis can be predicted by Zhang et al. present a Siamese self-attention (SSA) classi- biochemical markers. *e decision tree algorithm has been fication approach for multisensor burnt area mapping, and a considered to predict the risk of liver fibrosis, and the model multisource data set is created at the object level for training has been tested using a data set that includes laboratory tests and testing. Zhang et al. implement a robust, multicamera, and fibrosis markers. Another study [29] compared the multiplayer tracking framework. *ey used a deep learning performance of different algorithms (logistic regression, algorithm in their system to understand the impact of player KNN, ANN, and SVM) to assess liver disease detection. identification and the most distinguishing data [42]. Fur- Additionally, Sontakke et al. [30] utilized backpropagation thermore, deep learning algorithms have been used to and SVM algorithms to predict liver disease. *ir- identify COVID-19 using X-ray processing. For example, unavukkarasu et al. [24] applied logistic regression, SVM, several studies [43–45] present a rapid, robust, and practical and KNN for predicting liver disease based on the evaluation method for detecting COVID-19 from chest X-ray images. of accuracy, sensitivity, and specificity (recall). Moreover, According to experiments by Mahajan et al. [43], DenseNet Venkata Ramana et al. [31] studied the performance of is the best classifier to utilize as a base network with SSD512, various machine learning algorithms using different metrics especially for the problem of identifying COVID-19 infec- (accuracy, precision, sensitivity, and specificity). tion in chest X-ray images. Mahajan et al. [44] developed a A support vector machine (SVM) is considered a model for detecting COVID-19 from chest X-ray images. promising machine learning algorithm for classification *ey used ResNet101 as the basic network and implemented problems. In addition, there are many studies that apply the transposed convolution, prediction modules, and informa- SVM algorithm to text classification, face recognition, and tion injection into the DSSD network. *e artificial intel- bioinformatics. *e performance of the SVM algorithm is ligence-based detection models can significantly contribute often good compared to other techniques [32–34]. Another to the attainment of massive and high-performing screening machine learning algorithm is the Na¨ıve Bayes classifier, programs in various medical sectors. which is a simple probabilistic classifier applying Bayes’ theorem. In addition, the Na¨ıve Bayes classifier estimates the 3. Proposed Method means and variances of the variables for classification using a small amount of training data [35]. Moreover, decision tree *e main contribution of this study is the design of a (DT) and K-nearest neighbors (KNN) are supervised prediction model to detect the risk of liver damage, called the learning algorithms considered suitable for addressing both detecting model for liver damage (DMLD). classification and regression problems [36–38]. Another popular machine learning method is the artificial neural networks (ANN) that are inspired by the neural networks of 3.1. Detecting Model for Liver Damage (DMLD). In this the human brain [39]. study, we design a prediction model for adverse effects on 4 Journal of Healthcare Engineering ward, a common scale or range can be used. A popular and liver functionality of COVID-19 ICU patients called detecting model for liver damage (DMLD). *e method- widely used normalization technique is min-max normali- zation, which is applied in this study. *e min-max nor- ology of this study involves five stages, which are data collection, data preprocessing, feature selection, classifiers, malization technique transforms and rescales the data and evaluation and then result collection. Figure 2 illustrates between the range [0, 1] by the following equation: the system architecture of the DMLD prediction model. x − min x′ Moreover, a detailed explanation of the DMLD model will be � , (1) max − min F F presented in the following subsections. where min and max are the minimum and the maximum F F values of the feature F, respectively. *e original and the 3.1.1. Material. *e data set used in this research was ob- normalized value of the attributes, F, are represented by x tained from two main hospitals in the southern region of and x , respectively [48]. Saudi Arabia (Asir Central Hospital (ACH) in Asir and King Khalid Hospital in Najran). A total of 140 patients were included in the data set. *e study was limited to patients 3.1.3. Features Selection. *e data collected from the blood with positive COVID-19 infection who were admitted to the test will have plenty of different features with different in- intensive care unit (ICU). Ethical approval (REC No.: REC- formation. *erefore, the feature-selection step is applied to 11-1O-2020) for this study was obtained from the Regional reduce the number of relevant features in the data set, and Committee for Research Ethics, Directorate of Health Af- consequently, the size of the problem will be reduced, and we fairs, Asir Region, Ministry of Health, Saudi Arabia, and can obtain a better prediction for the risk of liver damage. In ethical approval (IRB Log Number: 2020-24E) for this study this research, the filter method has been followed in order to was obtained from the Regional Committee for Research rank the importance of k features in the data set based on the Ethics, Directorate of Health Affairs Najran, Ministry of relationship between the features and the target variable Health, Saudi Arabia. [49]. In addition, the correlation between the selected fea- *e data set has recent laboratory results and missing tures was examined in order to understand the data set and values are very minimal. *e laboratory results contain 20 the relationship between the features. numeric attributes as follows: creatinine, glucose, sodium, potassium, calcium, phosphorus, magnesium, chloride, uric acid, urea, total protein, TG, AST, ALT, cholesterol-VLDL, 3.1.4. Classifiers. In the DMLD model, five machine learning cholesterol-LDL, cholesterol-HDL, and LDH. *e class classifiers have been used, which are support vector machine presented in this data set is binary, which refers to whether a (SVM), decision tree (DT), Na¨ıve Bayes (NB), K-nearest patient has damage in the liver functionality or not based on neighbors (KNN), and artificial neural network (ANN). abnormal liver enzymes. Prediction of liver damage is very *ese classifiers were used to determine the risk of liver likely based on elevated liver enzymes, which are released damage and the selection of these classifiers is based on the from the liver as a result of liver injury. SARS-CoV-2 has following characteristics. been reported to cause infection of the liver via binding to Support vector machine (SVM): Support vector ma- angiotensin-converting enzyme 2 (ACE2) on chol- chines (SVM) are extensively used in medical appli- angiocytes, which are a population of liver cells [46]. *e cations. *e SVM algorithm with class labels of binding of SARS-CoV-2 to ACE2 will facilitate viral entry unknown data is used to develop an effective model for into the liver, causing damage to liver cells (hepatocytes) predicting disease. SVM is used [50] in both classifi- [46, 47]. Levels of ALT and AST in our data, which are cation and regression [51]. *e data points in the SVM specific liver enzymes, were significantly increased indi- model are represented in space and divided into cating liver injury. We identified liver damage based on normal values of liver enzymes. Table 1 shows the liver groups, with all points with comparable qualities falling into the same group. *e given data set is treated as a p- enzymes along with their normal and disturbing values. Any patient with increased liver enzymes levels is considered at dimensional vector in linear SVM, which can be split by a maximum of p-1 planes termed hyperplanes. As risk of liver damage. In the study data set, the percentage of shown in Figure 3, these planes divide the data space or possible liver damage is 50%. Table 2 shows the data set define the boundaries between data groups for classi- attributes and the obtained results from the laboratory, fication or regression issues. On the basis of the dis- which were used to examine the DMLD prediction model. tance between the two classes it separates, the optimal hyperplane can be chosen among a large number of 3.1.2. Data Preprocessing. *e aim of the data preprocessing hyperplanes. *e maximum-margin hyperplane is the phase is to clean the data set in order to use it as input for plane with the largest margin between the two classes classifier algorithms and then to provide more accurate [52, 53]. For n data points, the formula is observation. One of the significant issues in the collected real �→ �→ x , y ), . . . , x , y ), (2) data is missing values. *ese missing values are very rare, at 1 1 n n 4%; therefore, they were excluded from the data set. Another important aspect of data preprocessing is normalization, in where x is a real vector and y is the class to which x 1 1 1 which all attributes should have equal weight. In a simple belongs and is either 1 or −1. *e distance between the margin Journal of Healthcare Engineering 5 Featsures Preprocessing Input Classfiers Extraction Missing values ICU Patients Normalization Laboratory tests Results Evaluation SVM – DT – Measurement Interoperations KNN-ANN Metrics Figure 2: System architecture of DMLD prediction model. Table 1: Speci”c liver enzymes with reference ranges. Liver enzymes Normal range Disturbed range Number of patients with disturbed range AST 0–0–40 <40 109 ALT 0–37 <37 109 Table 2: Data set attributes. Attribute no. Attribute Variable type Reference range A1 Creatinine Real 0.5–1.3 A2 Glucose Real 70–110 A3 Sodium Real 135–153 A4 Potassium Real 3.5–5.3 A5 Calcium Real 8.8–10.2 A6 Phosphorus Real 2.7–5 A7 Magnesium Real 1.5–2.6 A8 Chloride Real 98–105 A9 Uric acid Real 3.4–7 A10 Urea Real 10–50 A11 Albumin Real 3.4–4.8 A12 Total protein Real 6.4–8.3 A13 Cholesterol – total Real 50–200 A14 TG Real 23–56 A15 ALT Real 0–37 A16 AST Real 0–41 A17 Cholesterol – VLDL Real 10–40 A18 Cholesterol – LDL Real 50–190 A 19 Cholesterol – HDL Real 30–70 A 20 LDH Real 135–225 Class Liver damage or not Binary 0 or 1 0 healthy liver 1 possible liver damage Positive examples Negative examples Support vector Figure 3: Classi”cation of data by support vector machine (SVM). Separating Hyperplane 6 Journal of Healthcare Engineering two classes y � 1 and y � −1 can be maximized by K-Nearest Neighbors (KNN): In machine learning, KNN constructing a hyperplane, which is defined as follows: is one of the most fundamental classification algorithms, and it produces excellent results [36]. KNN is a non- → → w · x − b � 0, (3) parametric, instance-based learning algorithm and can be used to solve problems involving classification and re- → → where w is the normal vector and b/‖ w ‖ is the hy- gression. In classification, KNN is used to determine which class a new unlabeled item belongs to. In any case, perplane’s offset along w. the KNN makes a shot at the assumption that comparable In an SVM model, tuning parameters help optimize the samples are close fits [38]. KNN sorts a sample into the classification results based on the specific data points most decided class among K neighbors. K is usually odd provided [54]. One of them may be the kernel, a and is restricted by how the classification algorithms can mathematical function that accepts data as input and be adjusted [56]. *is will be achieved by computing the transforms it into the required format. *ese functions distance between the data points that are nearest to the return the inner combination between two points in a samples by using methods such as Euclidean distance, sufficient space, which might be linear, nonlinear, ra- Manhattan distance, Hamming distance, or Minkowski dial base function (RBF), polynomial, or sigmoid. distance. In this study, the Euclidean distance metric was Decision Tree (DT): *e decision tree classifier is used in the final model for calculating the distance be- considered a supervised learning algorithm [36]. tween data points. Following the calculation of the dis- Compared with other supervised learning algorithms, a tance, the K closest neighbors are chosen, and the decision tree algorithm can be used for dealing with resultant class of the new object is determined using the both classification and regression problems. *e overall votes of the neighbors [51, 57]. perspective of using a DT is to create a preparation Artificial neural network (ANN): *e functionality of an model that can predict class or assessment of target artificial neural network (ANN) is similar to that of the factors by taking decision standards derived from human brain [39]. It resembles a network of nodes known training data. *e decision tree classifier can be a fast as artificial neurons. All of these nodes communicate with learner when constructing a decision/regression tree each other to transmit information. *e neurons in the utilizing acquired information as the splitting criterion, ANN can be represented by a state (0 or 1), and each node and it prunes the tree by minimizing error pruning [37]. might have a weight attached to it that determines its Na¨ıve Bayes (NB): A Na¨ıve Bayes classifier is a classical relevance or strength in the system. *e ANN structure is probabilistic classifier dependent on performing Bayes’ separated into layers with many nodes; data flow from the theorem within a highly independent assumption [35]. first layer (input layer) to the output layer after passing *e fundamental probability model would be as de- through intermediary levels (hidden layers). Every layer scriptive as the self-determining feature model. *e basic turns the data into relevant information before delivering assumption in the Na¨ıve Bayes classifier is that the the target output [58]. *e processes of transfer and presence of a specific feature of a class is unassociated activation are crucial in the functioning of neurons. *e with the presence of other features [55]. Even if the sum of all the weighted inputs is calculated using the assumption is not accurate, the Naıve Bayes classifier transfer function: performs reasonably well. *e Na¨ıve Bayes classifier has another advantage, which is that it only requires a small data set for the training stage in order to compute the z � w x + w b, (5) i i b means and variances of the essential variables for x�1 classification. For each label, only the variances of the variables need to be computed, not the whole covariance where b is the bias value, which in most cases is 1. Fur- matrix, because unassociated variables are unspecified. thermore, the activation function essentially flattens the *e kernel of the Na¨ıve Bayes operator can be formu- transfer function’s output into a specified range. *e acti- lated on numerical attributes. *is is clearly achieved by vation function could be linear or nonlinear and can be applying Bayes’ theorem and kernel density estimation. expressed simply as follows: f(z) � z, (6) π f x j j 0 (4) P y � j|x � , k Since no data restrictions are provided by the activation π f x k�1 k k 0 function, the sigmoid function is employed [51], which is written as follows: where π is an estimate of the prior probability of class j, and normally, π is the sample proportion falling into (7) a � σ(z) � . −z th 1 + e the j classification. f is the predictable density at x j 0 depending on a kernel density fit, including only th perceptions from the j class. *is is essentially similar to discriminant analysis, only instead of assuming 3.1.5. Evaluation. *e proposed model’s (DMLD) perfor- normality, it estimates the probability density of the mance was evaluated using the measurement performance classes utilizing a nonparametric method, Patrick. of several classification algorithms. Various evaluation Journal of Healthcare Engineering 7 methodologies, such as accuracy, precision, and recall, are Score used. *e following is a list of their definition. 1.20E+03 Accuracy: *e percentage of accurate and valid classi- 1.00E+03 fications is known as the accuracy [59]. To calculate the 8.00E+02 accuracy, the true positive (TP), false positive (FP), true 6.00E+02 negative (TN), and false negative (FN) values are required. 4.00E+02 TP + TN 2.00E+02 Accuracy � . (8) TP + FP + TN + FN 0.00E+00 Precision: Positive predictive value is another term for precision. It shows the percentage of positive outcomes successfully predicted by classifier algorithms. TP (9) Precision � . FP + TN Figure 4: Top selected features. Recall: Recall is also referred to as sensitivity or true positive rate because it mostly displays the method’s positive outcomes [60]. *e affectability evaluation determines the data because the data set variables (e.g., ALT, AST, and patient’s ability to be identified by their liver condition. LDH) have different ranges of values. For example, LDH for TP a single patient is 499 U/L, and ALTand ASTare 90 U/L and Recall � , (10) 34 U/L, respectively. *erefore, we applied different nor- TP + FN malization algorithms such as min-max and mean, but the *e evaluation variables that are used in the performance results did not show any difference. After applying the measurement, which is the confusion matrix, are deter- feature-selection step in the DMLD model, the results mined as follows. True positive (TP): *e outcome of the revealed that the three highest-scoring features were AST, prediction properly identifies the presence of the risk of liver ALT, and LDH, as shown in Figure 4. *ese selected features damage in a patient. False positive (FP): *e outcome of the agreed with clinically reported features related to liver in- prediction mistakenly identifies a patient as having the risk jury. ALT and AST are specific liver enzymes, and hence, of liver damage. True negative (TN): *e outcome of the they are considered markers for liver injury and failure prediction properly rejects the possibility of a patient being [63, 64]. Moreover, increased LDH levels have been reported at risk of liver damage. False negative (FN): *e outcome of in patients with acute liver failure [65, 66]. Correlation the prediction mistakenly rejects the possibility of a patient coefficients of selected features were applied to screen for being at risk of liver damage. possible correlation. *e linear relationship among selected Tenfold cross-validation is used to avoid the problems of features was defined as follows: positive correlation for over- and underfitting [61]. *en, the previous measurement r � 0.01 to 1.0 (where 1.0 was considered strong). As illus- performance is used to evaluate the classification systems’ trated in Figure 5, a heat map was used to present our results, performance. Accuracy reflects how accurate our classifier is in which ALT and AST showed a significant positive cor- in determining whether or not a patient is at risk of liver relation with r � 0.96. *is correlation between ALTand AST damage. Precision also has been applied to measure the is not surprising, since they are already approved scientif- classifier’s ability to make an accurate, positive prediction of ically as liver function markers. However, in agreement with the risk of liver damage. Additionally, sensitivity or recall is our selection of LDH as an important feature, the heat map employed in our research to determine the percentage of results interestingly revealed a very strong correlation be- actual positive cases of risk of liver damage that the classifier tween LDH and both specific liver enzymes ALT and AST, properly detects. with r � 0.94 and r � 0.97, respectively. Figure 6 describes the performance of the different classifiers used in the DMLD model, which are support vector 4. Results and Discussion machine (SVM), DT, Naıve Bayes (NB), K-nearest neighbors In this study, the DMLD model is proposed to contribute to (KNN), and artificial neural network (ANN). In the validation phase, the model was tested in two different methods, namely the prediction of the risk of liver damage using laboratory blood tests. *e DMLD model was implemented and ex- train-test split and tenfold cross-validation. In the train-test split approach, the data set was divided into two parts, amined in the Python 3.8 programming language via An- aconda Navigator [62]. In addition, different measurement training and testing. *e DMLD model was trained with 80% metrics (accuracy, precision, and recall) were considered to of the data set, and the remaining data were used for testing assess the performance of the DMLD model. *is was the DMLD model, by which preliminary results were gained. conducted using different machine learning classifiers to In addition, the tenfold cross-validation was applied in order predict the risk of liver damage. Tenfold cross-validation was to avoid overfitting, as shown in Table 3 and Figure 6. considered in order to validate the results. *e data set in this Table 3 and Figure 6 show that the accuracy of SVM is 0.87 and that of DT is 0.85, while for the Na¨ıve Bayes, KNN, study includes 140 COVID-19 ICU patients with 20 features, as shown in Table 1. Normalization is used for scaling the and ANN, it is 0.71. *erefore, SVM and DTachieved higher Creatinine Glucose sodium potassium calcium phoshours Magnesium Chloride Uric Acid Urea Albumin Total Protein Cholestrol-Total TG ALT Cholestrol-VLDL Cholestrol-LDL Cholestrol-HDL LDH AST Precision 8 Journal of Healthcare Engineering 1.00 accuracy than other classifiers (Na¨ıve Bayes, KNN, and ANN). In addition, we tried to study the impact of different 1 0.96 0.97 0.99 layers on the ANN performance by measuring the accuracy 0.98 of the ANN algorithm, but the results showed no effect on the algorithm performance, as presented in Table 4. Re- 0.97 0.96 1 0.94 garding precision, SVM achieved the highest score, with 0.96 0.95, and the score was 0.93 for DT. For Na¨ıve Bayes, KNN, and ANN classifiers, the precision values were found to be 0.95 0.97 0.94 0.5, 0.5, and 0.49, respectively. *e recall score of SVM was 0.94 the highest, at 0.95, and this score was 0.93 for DT. For Na¨ıve Bayes, KNN, and ANN classifiers, recall scores were 0.5, 0.5, AST ALT LDH and 0.49, respectively. Figure 5: Heat map for checking the correlation between selected *e performances of five classifiers in the DMLD model features. have been examined. *erefore, from the above results, it can be noted that SVM and DT are the most sufficient classifiers in the DMLD model for predicting the risk of liver damage in COVID-19 patients. In agreement with our study, 1.0 performances of the SVM [30] and DT [28] algorithms have been utilized to predict liver disease. SVM has shown the 0.8 best performance. *is is perhaps due to its ability to classify classes and generate a hyperplane that segregates classes after 0.6 data transformation. *erefore, early diagnosis of risk fac- tors by machine learning models such as SVM could assist in 0.4 planning medical decisions and treatment. 0.2 5. Conclusions and Future Work 0.0 *e effects of COVID-19 on the body are widespread. *e early diagnosis of liver damage due to COVID-19 can contribute to making medical decisions. *erefore, this study suggests that the DMLD model can help in the prediction of the risk of liver damage during SARS-CoV-2 infection. To evaluate the DMLD model, data on COVID-19 and ICU patients were collected, Accurecy preprocessed, and then used as an input for different classifiers. Recall *e performances of SVM, DT, Na¨ıve Bayes, KNN, and ANN classifiers were evaluated. SVM and DT showed the best Figure 6: Results of the classifier’s performance on the DMLD performance for predicting the diagnosis of disease severity model. based on laboratory tests. *erefore, this model could be ap- plied for the prediction of other diseases. *e further study of our work can be considered from two directions. Firstly, the Table 3: Evaluation parameters of different classifiers in the DMLD model. prediction of different risk levels of liver diseases could be extended, as the current work is limited to the DMLD model. Predictive models Accuracy Precision Recall Secondly, our data were limited to laboratory tests, and SVM 0.857 0.95 0.95 therefore future work could consider CT scan images. DT 0.85 0.93 0.93 NB 0.71 0.5 0.5 KNN 0.71 0.5 0.5 Data Availability ANN 0.7 0.49 0.49 For the privacy of individuals (patients’ laboratory results involved in the study), data cannot be made available publicly. Table 4: *e impact of different layers on the ANN performance. Ethical Approval Number of layers Accuracy Ethical approval (REC No.: REC-11-1O-2020) for this study 1 0.7 was obtained from the Regional Committee for Research 3 0.7 Ethics, Directorate of Health Affairs, Asir Region, Ministry 4 0.7 5 0.7 of Health, Saudi Arabia, and ethical approval (IRB Log 10 0.7 Number: 2020-24E) for this study was obtained from the 15 0.7 Regional Committee for Research Ethics, Directorate of 20 0.7 Health Affairs Najran, Ministry of Health, Saudi Arabia. LDH ALT AST SVM Decision Tree Naive Bayes KNN ANN Journal of Healthcare Engineering 9 [15] A. Ghoda and M. Ghoda, “Liver injury in covid-19 infection: a Conflicts of Interest systematic review,” Cureus, vol. 12, no. 7, Article ID e9487, *e authors declare that they have no conflicts of interest. [16] C. Zhang, L. Shi, and F.-S. Wang, “Liver injury in covid-19: management and challenges,” ;e lancet Gastroenterology & Acknowledgments hepatology, vol. 5, no. 5, pp. 428–430, 2020. [17] A. Mantovani, G. Beatrice, and A. Dalbeni, “Coronavirus *e authors would like to express their gratitude to the disease 2019 and prevalence of chronic liver disease: a meta- Ministry of Education and the Deanship of Scientific analysis,” Liver International, vol. 40, no. 6, pp. 1316–1320, Research—Najran University—Kingdom of Saudi Arabia for their financial and Technical support under code [18] H. Shao-Rui, S.-Y. Zhang, and J.-S. Lian, “Liver Enzyme number: NU/-/SERC/10/623. Elevation in Coronavirus Disease 2019: A Multicenter, Ret- rospective, Cross-Sectional Study,” ;e American journal of References gastroenterology, vol. 115, no. 7, 2020. [19] Q. Cai, D. Huang, H. Yu et al., “Covid-19: abnormal liver [1] World Health Organization Organization, “Who coronavirus function tests,” Journal of Hepatology, vol. 73, no. 3, (covid-19) dashboard,” Last checked on, 2021. pp. 566–574, 2020. [2] A. Shaikh Mana Saleh AlReshan, Y. Asiri, A. Sulaiman, [20] H. Shokri Afra, N. Amiri-Dashatan, F. Ghorbani, I. Maleki, H. Alshahrani, and H. Alshahrani, “Tele-COVID: a tele- and M. Rezaei-Tavirani, “Positive association between severity medicine SOA-based architectural design for COVID-9 pa- of covid-19 infection and liver damage: a systematic review tients,” Computers, Materials & Continua, vol. 67, no. 1, and meta-analysis,” Gastroenterology and hepatology from bed pp. 549–576, 2021. to bench, vol. 13, no. 4, p. 292, 2020. [3] A. Shaikh, M. S. Al Reshan, A. Sulaiman, H. Alshahrani, and [21] A. Spann, A. Yasodhara, J. Kang et al., “Applying machine Y. Asiri, “Secure telemedicine system design for covid-19 learning in liver disease and transplantation: a comprehensive patients treatment using service oriented architecture,” Sen- review,” Hepatology, vol. 71, no. 3, pp. 1093–1105, 2020. sors, vol. 22, no. 3, p. 952, 2022. [22] C.-C. Wu, W.-C. Yeh, and W.-D. Hsu, “Prediction of fatty [4] H. Esakandari, M. Nabi-Afjadi, J. Fakkari-Afjadi, liver disease using machine learning algorithms,” Computer N Farahmandian, S. M Miresmaeili, and E Bahreini, “A Methods and Programs in Biomedicine, vol. 170, pp. 23–29, comprehensive review of covid-19 characteristics,” Biological Procedures Online, vol. 22, pp. 19–10, 2020. [23] C. L. Liu, R. S. Soong, W. C. Lee, G. W. Jiang, and Y. C. Lin, [5] C. E. Gomez, B. Perdiguero, and M. Esteban, “Emerging sars- “Predicting short-term survival after liver transplantation cov-2 variants and impact in global vaccination programs against using machine learning,” Scientific Reports, vol. 10, no. 1, sars-cov-2/covid-19,” Vaccines, vol. 9, no. 3, p. 243, 2021. pp. 5654–5710, 2020. [6] T.. Mokhtari, F.. Hassani, and N.. Ghaffari, “Covid-19 and [24] K. *irunavukkarasu, A. S. Singh, M. d. Irfan, and multiorgan failure: a narrative review on potential mecha- A. Chowdhury, “Prediction of liver disease using classification nisms,” Journal of Molecular Histology, vol. 51, pp. 1–16, 2020. algorithms,” in Proceedings of the 2018 4th international [7] T. Marjot, G. J. Webb, and A. S Barritt, “Covid-19 and liver conference on computing communication and automation disease: mechanistic and clinical perspectives,” Nature Re- (ICCCA), pp. 1–3, IEEE, Greater Noida, INDIA, December views Gastroenterology & Hepatology, vol. 18, pp. 1–17, 2021. [8] A. D. Nardo, M. Schneeweiss-Gleixner, M. Bakail, [25] S. Hashem, G. Esmat, W. Elakel et al., “Accurate prediction of E. D. Dixon, S. F. Lax, and M. Trauner, “Pathophysiological advanced liver fibrosis using the decision tree learning al- mechanisms of liver injury in COVID-19,” Liver Interna- gorithm in chronic hepatitis c egyptian patients,” Gastroen- tional: Official Journal of the International Association for the terology research and practice, vol. 2016, Article ID 2636390, Study of the Liver, vol. 41, no. 1, pp. 20–32, 2021. 7 pages, 2016. [9] V. G. Puelles, M. Lutgehetmann, ¨ M. T. Lindenmeyer et al., [26] S. Uddin, A. Khan, M. E. Hossain, and M. A. Moni, “Multiorgan and renal tropism of sars-cov-2,” New England “Comparing different supervised machine learning algo- Journal of Medicine, vol. 383, no. 6, pp. 590–592, 2020. rithms for disease prediction,” BMC Medical Informatics and [10] Y. Wang, S. Liu, H. Liu et al., “Sars-cov-2 infection of the liver Decision Making, vol. 19, no. 1, pp. 281–316, 2019. directly contributes to hepatic impairment in patients with [27] B. Shahzad, I. Javed, A. Shaikh, A. Sulaiman, A. Abro, and covid-19,” Journal of Hepatology, vol. 73, no. 4, pp. 807–816, M. Ali Memon, “Reliable requirements engineering practices for covid-19 using blockchain,” Sustainability, vol. 13, no. 12, [11] K. A. Ratnakaran and H. S. Sahoo, “Risk of airborne infection p. 6748, 2021. and transmission-based precautions in dental setting,” Jour- [28] H. Ayeldeen, O. Shaker, G. Ayeldeen, and K. M. Anwar, nal of Operative Dentistry and Endodontics, vol. 5, no. 1, p. 19, “Prediction of liver fibrosis stages by machine learning model: a decision tree approach,” in Proceedings of the 2015 ;ird [12] C. J. Pirola and S. Sookoian, “SARS-CoV-2 virus and liver expression of host receptors: p,” Liver International, vol. 40, World Conference on Complex Systems (WCCS), pp. 1–6, IEEE, Morocco, November 2015. no. 8, pp. 2038–2040, 2020. [13] G. Paizis, C. Tikellis, and M. E. Cooper, “Chronic liver injury [29] J. Jacob, J. Chakkalakal Mathew, and J. Mathew and Elizabeth Issac, “Diagnosis of liver disease using machine learning in rats and humans upregulates the novel enzyme angiotensin converting enzyme 2,” Gut, vol. 54, no. 12, pp. 1790–1796, techniques,” Int Res J Eng Technol, vol. 5, no. 04, 2018. [30] S. Sontakke, J. Lohokare, and R. Dani, “Diagnosis of liver [14] T. Horvatits, M. Trauner, and V. Fuhrmann, “Hypoxic liver diseases using machine learning,” in Proceedings of the 2017 International Conference on Emerging Trends & Innovation in injury and cholestasis in critically ill patients,” Current Opinion in Critical Care, vol. 19, no. 2, pp. 128–132, 2013. ICT (ICEI), pp. 129–133, IEEE, Pune, India, February 2017. 10 Journal of Healthcare Engineering [31] B. Venkata Ramana, M. S. P. Babu, and N. B. Venkateswarlu, [48] “What is data mining. “Data mining: concepts and tech- “A critical study of selected classification algorithms for liver niques”,” Morgan Kaufinann, vol. 10, pp. 559–569, 2006. [49] G. Chandrashekar and F. Sahin, “A survey on feature selection disease diagnosis,” International Journal of Database Man- methods,” Computers & Electrical Engineering, vol. 40, no. 1, agement Systems, vol. 3, no. 2, pp. 101–114, 2011. [32] C. J. C. Burges, “A tutorial on support vector machines for pp. 16–28, 2014. pattern recognition,” Data Mining and Knowledge Discovery, [50] M. A. Hearst, S. T. Dumais, E. Osuna, J. Platt, and B. Scholkopf, “Support vector machines,” IEEE Intelligent vol. 2, no. 2, pp. 121–167, 1998. [33] N. Cristianini and J. Shawe-Taylor, Support Vector Machines Systems and ;eir Applications, vol. 13, no. 4, pp. 18–28, and Other Kernel-Based Learning Methods, Cambridge, En- 1998. gland, UK, 2004. [51] H. Kaur and V. Kumari, “Predictive modelling and analytics [34] K. P. Bennett and C. Campbell, “Support vector machines: for diabetes using a machine learning approach,” Applied hype or hallelujah?” Acm Sigkdd Explorations Newsletter, Computing and Informatics, vol. 16, 2020. vol. 2, no. 2, pp. 1–13, 2000. [52] R. A. McKinnon, D. A. Winkler, F. R. Burden, [35] X. Xiao Fang, “Inference-based na¨ıve bayes: turning na¨ıve P. l. A. Smith, M. J. Sorich, and J. O. Miners, “Comparison bayes cost-sensitive,” IEEE Transactions on Knowledge and of linear and nonlinear classification algorithms for the Data Engineering, vol. 25, no. 10, pp. 2302–2313, 2013. prediction of drug and chemical metabolism by human [36] A. S. Rahman, F. M. Javed Mehedi Shamrat, Z. Tasnim, J. Roy, udp-glucurono syltransferase isoforms,” Journal of and S. A. Hossain, “A comparative study on liver disease Chemical Information and Computer Sciences, vol. 43, prediction using supervised machine learning algorithms,” no. 6, pp. 2019–2024, 2003. International Journal of Scientific & Technology Research, [53] S. M. Mai and E. M. Hashem, “A study of support vector vol. 8, pp. 419–422, 2019. machine algorithm for liver disease diagnosis,” American [37] F.-J. Yang, “An extended idea about decision trees,” in Pro- Journal of Intelligent Systems, vol. 04, no. 9–14, 2014. ceedings of the 2019 International Conference on Computa- [54] S. Mahajan, A. Raina, X.-Z. Gao, and A. K. Pandit, “Plant tional Science and Computational Intelligence (CSCI), recognition using morphological feature extraction and pp. 349–354, IEEE, USA, December 2019. transfer learning over svm and adaboost,” Symmetry, vol. 13, [38] Z. Z. M. Ml-knn, “A lazy learning approach to multi-label no. 2, 2021. learning,” Pattern Recognition, vol. 40, no. 7, pp. 2038–2048, [55] S. Vijayarani and S. Dhayanand, “Liver disease prediction 2007. using SVM and naive bayes algorithms,” vol. 4, pp.216–220, [39] A. Mustafa and M. Ceylan, “Comparison of artificial neural network and extreme learning machine in benign liver lesions [56] D. Bell, Y. Bi K, Greer, G. Guo, and H. Wang, Knn Model- classification,” in Proceedings of the 2015 Medical Technologies Based Approach in Classification, vol. 2888, pp. 986–996, National Conference (TIPTEKNO), pp. 1–4, Bodrum, Turkey, Springer, Berlin, Heidelberg, 2003. [57] I. Gazalba and I. R. Nurul Gayatri, “Comparative analysis of October 2015. [40] R. Zhang, L. Xu, Z. Yu, Y. Shi, C. Mu, and M. Xu, “Deep- k-nearest neighbor and modified k-nearest neighbor algo- irtarget: an automatic target detector in infrared imagery rithm for data classification,” in Proceedings of the 2017 2nd using dual-domain feature extraction and allocation,” IEEE International conferences on Information Technology, Transactions on Multimedia, vol. 24, pp. 1735–1749, 2021. pp. 294–298, Information Systems and Electrical Engi- [41] Q. Zhang, L. Ge, R. Zhang et al., “Deep-learning-based burned neering (ICITISEE), Yogyakarta, Indonesia, November area mapping using the synergy of sentinel-1&2 data,” Remote 2017. Sensing of Environment, vol. 264, Article ID 112575, 2021. [58] T. Kohonen, “An introduction to neural computing,” Neural [42] R. Zhang, L. Wu, Y. Yang, W. Wu, Y. Chen, and M. Xu, Networks, vol. 1, no. 1, pp. 3–16, 1988. “Multi-camera multi-player tracking with deep player iden- [59] S.WanhuaandZ.Peng,“Statisticalinferenceonrecall,precision tification in sports video,” Pattern Recognition, vol. 102, and average precision under random selection,” in Proceedings Article ID 107260, 2020. of the 9th International Conference on Fuzzy Systems and [43] S. Mahajan, A. Raina, X.-Z. Gao, and A. K. Pandit, “Covid-19 Knowledge Discovery, pp. 1348–1352, IEEE, China, May 2012. detection using hybrid deep learning model in chest x-rays [60] M. Junker, R. Hoch, and A. Dengel, “On the evaluation of images,” Concurrency and Computation: Practice and Expe- document analysis components by recall, precision, and ac- rience, vol. 34, no. 5, Article ID e6747, 2022. curacy,” in Proceedings of the Fifth International Conference [44] S. Mahajan, A. Raina, M. Abouhawwash, X.-Z. Gao, and on Document Analysis and Recognition. ICDAR ’99 (Cat. A. K. Pandit, “Covid-19 detection from chest x-ray images No.PR00318), pp. 713–716, IEEE, Banganlore, India, Sep- using advanced deep learning techniques,” Computers, Ma- tember 1999. terials & Continua, vol. 70, pp. 1541–1556, 2021. [61] S. Yadav and S. Shukla, “Analysis of k-fold cross-validation [45] A. Raina, S. Mahajan, C. Vanipriya, A. Bhardwaj, and over hold-out validation on colossal datasets for quality A. K. Pandit, “Covid-19 detection: an approach using x-ray classification,” in Proceedings of the 2016 IEEE 6th Interna- images and deep learning techniques,” in Lecture Notes in tional Conference on Advanced Computing (IACC), pp. 78–83, Networks and Systemsvol. 6, , pp. 7–16, Springer, 2021. IEEE, Bhimavaram, India, February 2016. [46] X. Chai, L. Hu, and Y. Zhang, Specific Ace2 Expression in [62] Inc Anaconda, “Anaconda navigator,” 2021. Cholangiocytes May Cause Liver Damage after 2019-ncov [63] K. Ram, N. Saini, and S. Ram, “Covid-19 associated variations Infection, Berlin, Germany, 2020. in liver function parameters: a retrospective study,” Post- [47] P. Kumar, M. Sharma, A. Kulkarni, and P. N. Rao, “Patho- graduate Medical Journal, vol. 98, 2020. genesis of liver injury in coronavirus disease 2019,” Journal of [64] Y. Yang, Q. Lu, and M. Liu, Epidemiological and Clinical Clinical and Experimental Hepatology, vol. 10, no. 6, Features of the 2019 Novel Coronavirus Outbreak in china”, pp. 641-642, 2020. medrxiv, Berlin, Germany, 2020. Journal of Healthcare Engineering 11 [65] K. Kotoh, M. Enjoji, M. Kato, M Kohjima, M Nakamuta, and R Takayanagi, “A new parameter using serum lactate dehydrogenase and alanine aminotransferase level is useful for predicting the prognosis of patients at an early stage of acute liver injury: a retrospective study,” Com- parative Hepatology, vol. 7, no. 1, 2008. [66] K. Kotoh, M. Kato, and M. Kohjima, “Lactate dehydrogenase production in hepatocytes is increased at an early stage of acute liver failure,” Experimental and ;erapeutic Medicine, vol. 2, no. 2, pp. 195–199, 2011.
http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png
Journal of Healthcare Engineering
Hindawi Publishing Corporation
http://www.deepdyve.com/lp/hindawi-publishing-corporation/prediction-model-of-adverse-effects-on-liver-functions-of-covid-19-icu-IF64O9hHa8