1
|
Forouzandehmehr M, Paci M, Hyttinen J, Koivumäki JT. In silico study of the mechanisms of hypoxia and contractile dysfunction during ischemia and reperfusion of hiPSC cardiomyocytes. Dis Model Mech 2024; 17:dmm050365. [PMID: 38516812 PMCID: PMC11073514 DOI: 10.1242/dmm.050365] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/21/2023] [Accepted: 03/15/2024] [Indexed: 03/23/2024] Open
Abstract
Interconnected mechanisms of ischemia and reperfusion (IR) has increased the interest in IR in vitro experiments using human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs). We developed a whole-cell computational model of hiPSC-CMs including the electromechanics, a metabolite-sensitive sarcoplasmic reticulum Ca2+-ATPase (SERCA) and an oxygen dynamics formulation to investigate IR mechanisms. Moreover, we simulated the effect and action mechanism of levosimendan, which recently showed promising anti-arrhythmic effects in hiPSC-CMs in hypoxia. The model was validated using hiPSC-CM and in vitro animal data. The role of SERCA in causing relaxation dysfunction in IR was anticipated to be comparable to its function in sepsis-induced heart failure. Drug simulations showed that levosimendan counteracts the relaxation dysfunction by utilizing a particular Ca2+-sensitizing mechanism involving Ca2+-bound troponin C and Ca2+ flux to the myofilament, rather than inhibiting SERCA phosphorylation. The model demonstrates extensive characterization and promise for drug development, making it suitable for evaluating IR therapy strategies based on the changing levels of cardiac metabolites, oxygen and molecular pathways.
Collapse
Affiliation(s)
| | - Michelangelo Paci
- Department of Electrical, Electronic, and Information Engineering ‘Guglielmo Marconi’, University of Bologna, 47522 Cesena, Italy
| | - Jari Hyttinen
- Faculty of Medicine and Health Technology, Tampere University, 33520 Tampere, Finland
| | - Jussi T. Koivumäki
- Faculty of Medicine and Health Technology, Tampere University, 33520 Tampere, Finland
| |
Collapse
|
2
|
PulDi-COVID: Chronic obstructive pulmonary (lung) diseases with COVID-19 classification using ensemble deep convolutional neural network from chest X-ray images to minimize severity and mortality rates. Biomed Signal Process Control 2023; 81:104445. [PMID: 36466567 PMCID: PMC9708623 DOI: 10.1016/j.bspc.2022.104445] [Citation(s) in RCA: 19] [Impact Index Per Article: 19.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/17/2022] [Revised: 10/10/2022] [Accepted: 11/20/2022] [Indexed: 12/05/2022]
Abstract
Background and Objective In the current COVID-19 outbreak, efficient testing of COVID-19 individuals has proven vital to limiting and arresting the disease's accelerated spread globally. It has been observed that the severity and mortality ratio of COVID-19 affected patients is at greater risk because of chronic pulmonary diseases. This study looks at radiographic examinations exploiting chest X-ray images (CXI), which have become one of the utmost feasible assessment approaches for pulmonary disorders, including COVID-19. Deep Learning(DL) remains an excellent image classification method and framework; research has been conducted to predict pulmonary diseases with COVID-19 instances by developing DL classifiers with nine class CXI. However, a few claim to have strong prediction results; because of noisy and small data, their recommended DL strategies may suffer from significant deviation and generality failures. Methods Therefore, a unique CNN model(PulDi-COVID) for detecting nine diseases (atelectasis, bacterial-pneumonia, cardiomegaly, covid19, effusion, infiltration, no-finding, pneumothorax, viral-Pneumonia) using CXI has been proposed using the SSE algorithm. Several transfer-learning models: VGG16, ResNet50, VGG19, DenseNet201, MobileNetV2, NASNetMobile, ResNet152V2, DenseNet169 are trained on CXI of chronic lung diseases and COVID-19 instances. Given that the proposed thirteen SSE ensemble models solved DL's constraints by making predictions with different classifiers rather than a single, we present PulDi-COVID, an ensemble DL model that combines DL with ensemble learning. The PulDi-COVID framework is created by incorporating various snapshots of DL models, which have spearheaded chronic lung diseases with COVID-19 cases identification process with a deep neural network produced CXI by applying a suggested SSE method. That is familiar with the idea of various DL perceptions on different classes. Results PulDi-COVID findings were compared to thirteen existing studies for nine-class classification using COVID-19. Test results reveal that PulDi-COVID offers impressive outcomes for chronic diseases with COVID-19 identification with a 99.70% accuracy, 98.68% precision, 98.67% recall, 98.67% F1 score, lowest 12 CXIs zero-one loss, 99.24% AUC-ROC score, and lowest 1.33% error rate. Overall test results are superior to the existing Convolutional Neural Network(CNN). To the best of our knowledge, the observed results for nine-class classification are significantly superior to the state-of-the-art approaches employed for COVID-19 detection. Furthermore, the CXI that we used to assess our algorithm is one of the larger datasets for COVID detection with pulmonary diseases. Conclusion The empirical findings of our suggested approach PulDi-COVID show that it outperforms previously developed methods. The suggested SSE method with PulDi-COVID can effectively fulfill the COVID-19 speedy detection needs with different lung diseases for physicians to minimize patient severity and mortality.
Collapse
|
3
|
Que W, Han C, Zhao X, Shi L. An ECG generative model of myocardial infarction. COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE 2022; 225:107062. [PMID: 35994870 DOI: 10.1016/j.cmpb.2022.107062] [Citation(s) in RCA: 2] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Received: 04/21/2022] [Revised: 08/02/2022] [Accepted: 08/06/2022] [Indexed: 06/15/2023]
Abstract
Background and Objective Computer-aided diagnosis (CAD) of Myocardial Infarction (MI) using machine learning depends on a large amount of clinical Electrocardiogram (ECG) data. Existing infarct ECG databases face the problem of class imbalance. Data augmentation using generative simulation models is a new approach to effectively address this problem. Methods A multiscale ECG generative model was established for ECG data augmentation. In the cellular layer, an ischemic Action Potential (AP) model was established to generate APs in cardiomyocytes with different transmural regions of infraction or different ischemic durations. In the tissue layer, a probability-driven cellular automata excitation propagation model was established to simulate the propagation speed and direction of excitation. An infarct tissue model and a coronary artery model were established to describe the spatiotemporal diversity of MI. A ventricle model, a human torso model, and a computational model of surface ECG based on field source theory were established in the heart-torso layer. Results The model generated pathological 12-lead ECGs of MI with different topography and different extent. When simulating different ventricular wall infarction, the lesions appear in the same leads as the clinical 12-lead ECG. The ST-segment decreases and the T-wave amplitude decreases, similar to the clinical ECG features when simulating subendocardial ischemia. The average fidelity of the 12-lead ECG the model generated is 95.6%, according to the designed DTW-GRA distance algorithm. Conclusions The generative model considers the electrophysiological properties of the natural heart, the pathology of myocardial infarction, and the diversity of clinical ECGs. The model can provide many reliable samples for machine learning of MI.
Collapse
Affiliation(s)
- Wenge Que
- Department of Automation, Tsinghua University, Beijing 100084, China.
| | - Chuang Han
- School of Computer and Communication Engineering, Zhengzhou University of Light Industry, Zhengzhou 450000, China
| | - Xiliang Zhao
- Center for Coronary Artery Disease, Division of Cardiology, Beijing Anzhen Hospital, Capital Medical University, Beijing 100029, China.
| | - Li Shi
- Department of Automation, Tsinghua University, Beijing 100084, China; Beijing National Research Center for Information Science and Technology, Beijing 100084, China.
| |
Collapse
|
4
|
Marzog HA, Abd HJ. Machine Learning ECG Classification Using Wavelet Scattering of Feature Extraction. APPLIED COMPUTATIONAL INTELLIGENCE AND SOFT COMPUTING 2022; 2022:1-8. [DOI: 10.1155/2022/9884076] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 09/02/2023] Open
Abstract
The heart’s electrical activity is registered by an electrocardiogram (ECG), which consists of a wealth of pathological data on heart diseases such as arrhythmia. However, with increasing complexity and nonlinearity, direct observation of ECG signals and analysis is very tough. The highest accuracy of classification performance for machine learning approaches are 99.7 for neural network with wavelet scattering features extraction and 99.92 for SVM also with wavelet scattering features extraction. Through wavelet cascades with a neural network, the wavelet scattering transform can yield a translation invariant and deflection depictions of ECG signals. We suggested a new wavelet scattering transform-based method for automatically classifying three types of ECG heart diseases as follows: arrhythmia (ARR), congestive heart failure (CHF), and normal sinus rhythm (NSR). The bandwidth of the scaling function is used to critically downsample the wavelet scattering transform in time. As a result, each of the scattering paths has 16-time windows. Beat classification performance is classified by utilizing the MIT-BIH arrhythmia dataset. The suggested method is able to conduct high accuracy arrhythmia classification, with a 99.7% and 99.92% accuracy rate of the neural network (NN) and support vector machine (SVM), respectively, and will aid physicians in ECG explanation.
Collapse
Affiliation(s)
- Heyam A. Marzog
- Electrical Engineering Department, College of Engineering, University of Babylon, Hilla, Babil, Iraq
- Engineering Technical College/Najaf, Al-Furat Al-Awsat Technical University, Al Najaf 31001, Iraq
| | - Haider. J. Abd
- Electrical Engineering Department, College of Engineering, University of Babylon, Hilla, Babil, Iraq
| |
Collapse
|
5
|
Cheng T, Jiang F, Li Q, Zeng J, Zhang B. Quantitative Analysis Using Consecutive Time Window for Unobtrusive Atrial Fibrillation Detection Based on Ballistocardiogram Signal. SENSORS (BASEL, SWITZERLAND) 2022; 22:5516. [PMID: 35898020 PMCID: PMC9331962 DOI: 10.3390/s22155516] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Figures] [Subscribe] [Scholar Register] [Received: 06/19/2022] [Revised: 07/19/2022] [Accepted: 07/19/2022] [Indexed: 06/15/2023]
Abstract
Atrial fibrillation (AF) is the most common clinically significant arrhythmia; therefore, AF detection is crucial. Here, we propose a novel feature extraction method to improve AF detection performance using a ballistocardiogram (BCG), which is a weak vibration signal on the body surface transmitted by the cardiogenic force. In this paper, continuous time windows (CTWs) are added to each BCG segment and recurrence quantification analysis (RQA) features are extracted from each time window. Then, the number of CTWs is discussed and the combined features from multiple time windows are ranked, which finally constitute the CTW-RQA features. As validation, the CTW-RQA features are extracted from 4000 BCG segments of 59 subjects, which are compared with classical time and time-frequency features and up-to-date energy features. The accuracy of the proposed feature is superior, and three types of features are fused to obtain the highest accuracy of 95.63%. To evaluate the importance of the proposed feature, the fusion features are ranked using a chi-square test. CTW-RQA features account for 60% of the first 10 fusion features and 65% of the first 17 fusion features. It follows that the proposed CTW-RQA features effectively supplement the existing BCG features for AF detection.
Collapse
Affiliation(s)
- Tianqing Cheng
- College of Medicine and Biological Information Engineering, Northeastern University, Shenyang 110819, China; (T.C.); (Q.L.); (J.Z.)
| | - Fangfang Jiang
- College of Medicine and Biological Information Engineering, Northeastern University, Shenyang 110819, China; (T.C.); (Q.L.); (J.Z.)
| | - Qing Li
- College of Medicine and Biological Information Engineering, Northeastern University, Shenyang 110819, China; (T.C.); (Q.L.); (J.Z.)
| | - Jitao Zeng
- College of Medicine and Biological Information Engineering, Northeastern University, Shenyang 110819, China; (T.C.); (Q.L.); (J.Z.)
| | - Biyong Zhang
- College of Medicine and Biological Information Engineering, Eindhoven University of Technology, 5600 MB Eindhoven, The Netherlands;
- BOBO Technology, Hangzhou 310000, China
| |
Collapse
|
6
|
Vigier M, Vigier B, Andritsch E, Schwerdtfeger AR. Cancer classification using machine learning and HRV analysis: preliminary evidence from a pilot study. Sci Rep 2021; 11:22292. [PMID: 34785733 PMCID: PMC8595703 DOI: 10.1038/s41598-021-01779-1] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.3] [Reference Citation Analysis] [Abstract] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/12/2021] [Accepted: 11/01/2021] [Indexed: 12/16/2022] Open
Abstract
Most cancer patients exhibit autonomic dysfunction with attenuated heart rate variability (HRV) levels compared to healthy controls. This research aimed to create and evaluate a machine learning (ML) model enabling discrimination between cancer patients and healthy controls based on 5-min-ECG recordings. We selected 12 HRV features based on previous research and compared the results between cancer patients and healthy individuals using Wilcoxon sum-rank test. Recursive Feature Elimination (RFE) identified the top five features, averaged over 5 min and employed them as input to three different ML. Next, we created an ensemble model based on a stacking method that aggregated the predictions from all three base classifiers. All HRV features were significantly different between the two groups. SDNN, RMSSD, pNN50%, HRV triangular index, and SD1 were selected by RFE and used as an input to three different ML. All three base-classifiers performed above chance level, RF being the most efficient with a testing accuracy of 83%. The ensemble model showed a classification accuracy of 86% and an AUC of 0.95. The results obtained by ML algorithms suggest HRV parameters could be a reliable input for differentiating between cancer patients and healthy controls. Results should be interpreted in light of some limitations that call for replication studies with larger sample sizes.
Collapse
Affiliation(s)
- Marta Vigier
- Division of Oncology, Medical University of Graz, Auenbruggerplatz 15, 8036, Graz, Austria. .,Institute of Psychology, University of Graz, Graz, Austria.
| | | | - Elisabeth Andritsch
- Division of Oncology, Medical University of Graz, Auenbruggerplatz 15, 8036, Graz, Austria
| | | |
Collapse
|
7
|
Abstract
Computer modeling of the electrophysiology of the heart has undergone significant progress. A healthy heart can be modeled starting from the ion channels via the spread of a depolarization wave on a realistic geometry of the human heart up to the potentials on the body surface and the ECG. Research is advancing regarding modeling diseases of the heart. This article reviews progress in calculating and analyzing the corresponding electrocardiogram (ECG) from simulated depolarization and repolarization waves. First, we describe modeling of the P-wave, the QRS complex and the T-wave of a healthy heart. Then, both the modeling and the corresponding ECGs of several important diseases and arrhythmias are delineated: ischemia and infarction, ectopic beats and extrasystoles, ventricular tachycardia, bundle branch blocks, atrial tachycardia, flutter and fibrillation, genetic diseases and channelopathies, imbalance of electrolytes and drug-induced changes. Finally, we outline the potential impact of computer modeling on ECG interpretation. Computer modeling can contribute to a better comprehension of the relation between features in the ECG and the underlying cardiac condition and disease. It can pave the way for a quantitative analysis of the ECG and can support the cardiologist in identifying events or non-invasively localizing diseased areas. Finally, it can deliver very large databases of reliably labeled ECGs as training data for machine learning.
Collapse
|
8
|
Montnach J, Baró I, Charpentier F, De Waard M, Loussouarn G. Modelling sudden cardiac death risks factors in patients with coronavirus disease of 2019: the hydroxychloroquine and azithromycin case. Europace 2021; 23:1124-1133. [PMID: 34009333 PMCID: PMC8135857 DOI: 10.1093/europace/euab043] [Citation(s) in RCA: 8] [Impact Index Per Article: 2.7] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/18/2020] [Accepted: 02/16/2021] [Indexed: 12/23/2022] Open
Abstract
AIMS Coronavirus disease of 2019 (COVID-19) has rapidly become a worldwide pandemic. Many clinical trials have been initiated to fight the disease. Among those, hydroxychloroquine and azithromycin had initially been suggested to improve clinical outcomes. Despite any demonstrated beneficial effects, they are still in use in some countries but have been reported to prolong the QT interval and induce life-threatening arrhythmia. Since a significant proportion of the world population may be treated with such COVID-19 therapies, evaluation of the arrhythmogenic risk of any candidate drug is needed. METHODS AND RESULTS Using the O'Hara-Rudy computer model of human ventricular wedge, we evaluate the arrhythmogenic potential of clinical factors that can further alter repolarization in COVID-19 patients in addition to hydroxychloroquine (HCQ) and azithromycin (AZM) such as tachycardia, hypokalaemia, and subclinical to mild long QT syndrome. Hydroxychloroquine and AZM drugs have little impact on QT duration and do not induce any substrate prone to arrhythmia in COVID-19 patients with normal cardiac repolarization reserve. Nevertheless, in every tested condition in which this reserve is reduced, the model predicts larger electrocardiogram impairments, as with dofetilide. In subclinical conditions, the model suggests that mexiletine limits the deleterious effects of AZM and HCQ. CONCLUSION By studying the HCQ and AZM co-administration case, we show that the easy-to-use O'Hara-Rudy model can be applied to assess the QT-prolongation potential of off-label drugs, beyond HCQ and AZM, in different conditions representative of COVID-19 patients and to evaluate the potential impact of additional drug used to limit the arrhythmogenic risk.
Collapse
Affiliation(s)
- Jérôme Montnach
- Université de Nantes, CNRS, INSERM, l’institut du thorax, Nantes F-44000, France
| | - Isabelle Baró
- Université de Nantes, CNRS, INSERM, l’institut du thorax, Nantes F-44000, France
| | - Flavien Charpentier
- Université de Nantes, CNRS, INSERM, l’institut du thorax, Nantes F-44000, France
| | - Michel De Waard
- Université de Nantes, CNRS, INSERM, l’institut du thorax, Nantes F-44000, France
- Laboratory of Excellence, Ion Channels, Science & Therapeutics, Valbonne F-06560, France
| | - Gildas Loussouarn
- Université de Nantes, CNRS, INSERM, l’institut du thorax, Nantes F-44000, France
| |
Collapse
|
9
|
Exploration of physiological sensors, features, and machine learning models for pain intensity estimation. PLoS One 2021; 16:e0254108. [PMID: 34242325 PMCID: PMC8270203 DOI: 10.1371/journal.pone.0254108] [Citation(s) in RCA: 16] [Impact Index Per Article: 5.3] [Reference Citation Analysis] [Abstract] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/10/2020] [Accepted: 06/20/2021] [Indexed: 11/19/2022] Open
Abstract
In current clinical settings, typically pain is measured by a patient’s self-reported information. This subjective pain assessment results in suboptimal treatment plans, over-prescription of opioids, and drug-seeking behavior among patients. In the present study, we explored automatic objective pain intensity estimation machine learning models using inputs from physiological sensors. This study uses BioVid Heat Pain Dataset. We extracted features from Electrodermal Activity (EDA), Electrocardiogram (ECG), Electromyogram (EMG) signals collected from study participants subjected to heat pain. We built different machine learning models, including Linear Regression, Support Vector Regression (SVR), Neural Networks and Extreme Gradient Boosting for continuous value pain intensity estimation. Then we identified the physiological sensor, feature set and machine learning model that give the best predictive performance. We found that EDA is the most information-rich sensor for continuous pain intensity prediction. A set of only 3 features from EDA signals using SVR model gave an average performance of 0.93 mean absolute error (MAE) and 1.16 root means square error (RMSE) for the subject-independent model and of 0.92 MAE and 1.13 RMSE for subject-dependent. The MAE achieved with signal-feature-model combination is less than 1 unit on 0 to 4 continues pain scale, which is smaller than the MAE achieved by the methods reported in the literature. These results demonstrate that it is possible to estimate pain intensity of a patient using a computationally inexpensive machine learning model with 3 statistical features from EDA signal which can be collected from a wrist biosensor. This method paves a way to developing a wearable pain measurement device.
Collapse
|
10
|
Alakus TB, Turkoglu I. Comparison of deep learning approaches to predict COVID-19 infection. CHAOS, SOLITONS, AND FRACTALS 2020; 140:110120. [PMID: 33519109 PMCID: PMC7833512 DOI: 10.1016/j.chaos.2020.110120] [Citation(s) in RCA: 107] [Impact Index Per Article: 26.8] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Subscribe] [Scholar Register] [Received: 06/22/2020] [Accepted: 07/10/2020] [Indexed: 05/05/2023]
Abstract
The SARS-CoV2 virus, which causes COVID-19 (coronavirus disease) has become a pandemic and has expanded all over the world. Because of increasing number of cases day by day, it takes time to interpret the laboratory findings thus the limitations in terms of both treatment and findings are emerged. Due to such limitations, the need for clinical decisions making system with predictive algorithms has arisen. Predictive algorithms could potentially ease the strain on healthcare systems by identifying the diseases. In this study, we perform clinical predictive models that estimate, using deep learning and laboratory data, which patients are likely to receive a COVID-19 disease. To evaluate the predictive performance of our models, precision, F1-score, recall, AUC, and accuracy scores calculated. Models were tested with 18 laboratory findings from 600 patients and validated with 10 fold cross-validation and train-test split approaches. The experimental results indicate that our predictive models identify patients that have COVID-19 disease at an accuracy of 86.66%, F1-score of 91.89%, precision of 86.75%, recall of 99.42%, and AUC of 62.50%. It is observed that predictive models trained on laboratory findings could be used to predict COVID-19 infection, and can be helpful for medical experts to prioritize the resources correctly. Our models (available at (https://github.com/burakalakuss/COVID-19-Clinical)) can be employed to assists medical experts in validating their initial laboratory findings, and can also be used for clinical prediction studies.
Collapse
Affiliation(s)
- Talha Burak Alakus
- Kirklareli University, Engineering Faculty, Department of Software Engineering, Kirklareli, 39000, Turkey
| | - Ibrahim Turkoglu
- Firat University, Technology Faculty, Department of Software Engineering, Elazig, 23119, Turkey
| |
Collapse
|
11
|
Stoean R, Stoean C, Becerra-García R, García-Bermúdez R, Atencia M, García-Lagos F, Velázquez-Pérez L, Joya G. A hybrid unsupervised-Deep learning tandem for electrooculography time series analysis. PLoS One 2020; 15:e0236401. [PMID: 32692779 PMCID: PMC7373280 DOI: 10.1371/journal.pone.0236401] [Citation(s) in RCA: 4] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [MESH Headings] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/28/2019] [Accepted: 07/06/2020] [Indexed: 11/18/2022] Open
Abstract
Medical data are often tricky to get mined for patterns even by the generally demonstrated successful modern methodologies of deep learning. This paper puts forward such a medical classification task, where patient registers of two of the categories are sometimes hard to be distinguished because of samples showing characteristics of both labels in turn in several repetitions of the screening procedure. To this end, the current research appoints a pre-processing clustering step (through self-organizing maps) to group the data based on shape similarity and relabel it accordingly. Subsequently, a deep learning approach (a tandem of convolutional and long short-term memory networks) performs the training classification phase on the ‘cleaned’ samples. The dual methodology was applied for the computational diagnosis of electrooculography tests within spino-cerebral ataxia of type 2. The accuracy obtained for the discrimination into three classes was of 78.24%. The improvement that this duo brings over the deep learner alone does not stem from significantly higher accuracy results when the performance is considered for all classes. The major finding of this combination is that half of the presymptomatic cases were correctly found, in opposition to the single deep model, where this category was sacrificed by the learner in favor of a good accuracy overall. A high accuracy in general is desirable for any medical task, however the correct identification of cases before the symptoms become evident is more important.
Collapse
Affiliation(s)
| | | | | | | | | | | | - Luis Velázquez-Pérez
- Cuban Academy of Sciences, La Habana, Cuba
- Center for Research and Rehabilitation of Hereditary Ataxias, Holguín, Cuba
| | | |
Collapse
|
12
|
Ranking Information Extracted from Uncertainty Quantification of the Prediction of a Deep Learning Model on Medical Time Series Data. MATHEMATICS 2020. [DOI: 10.3390/math8071078] [Citation(s) in RCA: 8] [Impact Index Per Article: 2.0] [Reference Citation Analysis] [Abstract] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 01/23/2023]
Abstract
Uncertainty quantification in deep learning models is especially important for the medical applications of this complex and successful type of neural architectures. One popular technique is Monte Carlo dropout that gives a sample output for a record, which can be measured statistically in terms of average probability and variance for each diagnostic class of the problem. The current paper puts forward a convolutional–long short-term memory network model with a Monte Carlo dropout layer for obtaining information regarding the model uncertainty for saccadic records of all patients. These are next used in assessing the uncertainty of the learning model at the higher level of sets of multiple records (i.e., registers) that are gathered for one patient case by the examining physician towards an accurate diagnosis. Means and standard deviations are additionally calculated for the Monte Carlo uncertainty estimates of groups of predictions. These serve as a new collection where a random forest model can perform both classification and ranking of variable importance. The approach is validated on a real-world problem of classifying electrooculography time series for an early detection of spinocerebellar ataxia 2 and reaches an accuracy of 88.59% in distinguishing between the three classes of patients.
Collapse
|