Pub Date : 2024-09-03DOI: 10.1016/j.cmpb.2024.108406
Background and objective:
Electrocardiogram (ECG) analysis is crucial in diagnosing cardiovascular diseases (CVDs). It is important to consider both temporal and spatial features in ECG analysis to improve automated CVDs diagnosis. Significant progress has been made in automated CVDs diagnosis based on ECG with the continuous development of deep learning. Current most researches often treat 12-lead ECG signals as synchronous sequences in Euclidean space, focusing primarily on extracting temporal features while overlooking the spatial relationships among the 12-lead. However, the spatial distribution of 12-lead ECG electrodes can be more naturally represented using non-Euclidean data structures, which makes the relationships among leads more consistent with their intrinsic characteristics.
Methods:
This study proposes an innovative method, Convolutional Residual Graph Neural Network (Conv-RGNN), for ECG classification. The first step is to segment the 12-lead ECG into twelve single-lead ECG, which are then mapped to nodes in a graph that captures the relationships between the different leads through spatial connections, resulting in the 12-lead ECG graph. The graph is then used as input for Conv-RGNN. A convolutional neural network with a position attention mechanism is used to extract temporal sequence information and selectively integrate contextual information to enhance semantic features at different positions. The spatial features of the 12-lead ECG graph are extracted using the residual graph neural network.
Results:
The experimental results indicate that Conv-RGNN is highly competitive in two multi-label datasets and one single-label dataset, demonstrating exceptional parameter efficiency, inference speed, model performance, and robustness.
Conclusion:
The Conv-RGNN proposed in this paper offer a promising and feasible approach for intelligent diagnosis in resource-constrained environments.
{"title":"Conv-RGNN: An efficient Convolutional Residual Graph Neural Network for ECG classification","authors":"","doi":"10.1016/j.cmpb.2024.108406","DOIUrl":"10.1016/j.cmpb.2024.108406","url":null,"abstract":"<div><h3>Background and objective:</h3><p>Electrocardiogram (ECG) analysis is crucial in diagnosing cardiovascular diseases (CVDs). It is important to consider both temporal and spatial features in ECG analysis to improve automated CVDs diagnosis. Significant progress has been made in automated CVDs diagnosis based on ECG with the continuous development of deep learning. Current most researches often treat 12-lead ECG signals as synchronous sequences in Euclidean space, focusing primarily on extracting temporal features while overlooking the spatial relationships among the 12-lead. However, the spatial distribution of 12-lead ECG electrodes can be more naturally represented using non-Euclidean data structures, which makes the relationships among leads more consistent with their intrinsic characteristics.</p></div><div><h3>Methods:</h3><p>This study proposes an innovative method, Convolutional Residual Graph Neural Network (Conv-RGNN), for ECG classification. The first step is to segment the 12-lead ECG into twelve single-lead ECG, which are then mapped to nodes in a graph that captures the relationships between the different leads through spatial connections, resulting in the 12-lead ECG graph. The graph is then used as input for Conv-RGNN. A convolutional neural network with a position attention mechanism is used to extract temporal sequence information and selectively integrate contextual information to enhance semantic features at different positions. The spatial features of the 12-lead ECG graph are extracted using the residual graph neural network.</p></div><div><h3>Results:</h3><p>The experimental results indicate that Conv-RGNN is highly competitive in two multi-label datasets and one single-label dataset, demonstrating exceptional parameter efficiency, inference speed, model performance, and robustness.</p></div><div><h3>Conclusion:</h3><p>The Conv-RGNN proposed in this paper offer a promising and feasible approach for intelligent diagnosis in resource-constrained environments.</p></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":null,"pages":null},"PeriodicalIF":4.9,"publicationDate":"2024-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142145347","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-02DOI: 10.1016/j.cmpb.2024.108405
Background and Objective:
Automatic sleep staging is essential for assessing and diagnosing sleep disorders, serving millions of people who suffer from them. Numerous sleep staging models have been proposed recently, but most of them have not fully explored the sleep transition rules that are essential for sleep experts to identify sleep stages. Therefore, one objective of this paper is to develop an automatic sleep staging model to capture the transition rules between sleep stages.
Methods:
In this paper, we propose a novel sleep staging model named SleepGCN. It utilizes the deep features of electroencephalogram (EEG) and electrooculogram (EOG) signals extracted by the sleep representation learning (SRL) module, in conjunction with the transition rules learned by the sleep transition rule learning (STRL) module to identify sleep stages. Specifically, the SRL module utilizes the residual network (ResNet) and Long Short Term Memory (LSTM) structure to capture the deep time-invariant features and temporal information of each sleep stage from the two-channel EEG-EOG, and then applies a feature enhancement block to obtain the refined features. The STRL module employs a Graph Convolutional Network (GCN) and a transition rule matrix to capture transition rules between sleep stages based on the sequence labels of the input signals.
Results:
We evaluate SleepGCN on five public datasets: SleepEDF-20, SleepEDF-78, SHHS, DOD-H and DOD-O. Overall, SleepGCN achieves an accuracy of 89.70%, 87.70%, 86.16%, 82.07%, and 81.20%, alongside a macro-average F1-score of 85.20%, 82.70%, 77.69%, 72.44%, and 72.93% across these datasets, respectively.
Conclusions:
The results achieved by our proposed model are much better than those of all other compared models. The ablation study validates the contributions of the SRL and STRL modules proposed in SleepGCN to the sleep staging tasks. Additionally, it shows that the sleep staging model using two-channel EEG-EOG outperforms the model using single-channel EEG or EOG. Overall, SleepGCN is an effective solution for sleep staging using two-channel EEG-EOG.
{"title":"SleepGCN: A transition rule learning model based on Graph Convolutional Network for sleep staging","authors":"","doi":"10.1016/j.cmpb.2024.108405","DOIUrl":"10.1016/j.cmpb.2024.108405","url":null,"abstract":"<div><h3>Background and Objective:</h3><p>Automatic sleep staging is essential for assessing and diagnosing sleep disorders, serving millions of people who suffer from them. Numerous sleep staging models have been proposed recently, but most of them have not fully explored the sleep transition rules that are essential for sleep experts to identify sleep stages. Therefore, one objective of this paper is to develop an automatic sleep staging model to capture the transition rules between sleep stages.</p></div><div><h3>Methods:</h3><p>In this paper, we propose a novel sleep staging model named SleepGCN. It utilizes the deep features of electroencephalogram (EEG) and electrooculogram (EOG) signals extracted by the sleep representation learning (SRL) module, in conjunction with the transition rules learned by the sleep transition rule learning (STRL) module to identify sleep stages. Specifically, the SRL module utilizes the residual network (ResNet) and Long Short Term Memory (LSTM) structure to capture the deep time-invariant features and temporal information of each sleep stage from the two-channel EEG-EOG, and then applies a feature enhancement block to obtain the refined features. The STRL module employs a Graph Convolutional Network (GCN) and a transition rule matrix to capture transition rules between sleep stages based on the sequence labels of the input signals.</p></div><div><h3>Results:</h3><p>We evaluate SleepGCN on five public datasets: SleepEDF-20, SleepEDF-78, SHHS, DOD-H and DOD-O. Overall, SleepGCN achieves an accuracy of 89.70%, 87.70%, 86.16%, 82.07%, and 81.20%, alongside a macro-average F1-score of 85.20%, 82.70%, 77.69%, 72.44%, and 72.93% across these datasets, respectively.</p></div><div><h3>Conclusions:</h3><p>The results achieved by our proposed model are much better than those of all other compared models. The ablation study validates the contributions of the SRL and STRL modules proposed in SleepGCN to the sleep staging tasks. Additionally, it shows that the sleep staging model using two-channel EEG-EOG outperforms the model using single-channel EEG or EOG. Overall, SleepGCN is an effective solution for sleep staging using two-channel EEG-EOG.</p></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":null,"pages":null},"PeriodicalIF":4.9,"publicationDate":"2024-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142145349","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-01DOI: 10.1016/j.cmpb.2024.108404
Background and Objective
The increasing implementation and use of electronic health records over the last few decades has made a significant volume of clinical data being available. Over the past 20 years, hospitals have also adopted and implemented data warehouse technology to facilitate the reuse of administrative and clinical data for research. However, the implementation of clinical data warehouses encounters a set of barriers: ethical, legislative, technical, human and organizational. This paper proposes an overview of difficulties and barriers encountered during a clinical data warehouse (CDW) development and implementation project.
Methods
We conducted a focus group at the 2023 Medical Informatics Europe Conference and invited professionals involved in the implementation of CDW. These experts described their CDW and the difficulties and barriers they encountered at each phase: (i) launching of the data warehouse project, (ii) implementing the data warehouse and (iii) using a data warehouse in routine operations. They were also asked to propose solutions they were able to implement to address the barriers previously reported.
Results
After synthesis and consensus, a total of 26 barriers were identified, 10 pertained to tasks, 5 to tools and technologies, 4 to persons, 4 to organization, and 3 to the external environment. To address these challenges, a set of 15 practical recommendations was offered, covering essential aspects such as governance, stakeholder engagement, interdisciplinary collaboration, and external expertise utilization.
Conclusions
These recommendations serve as a valuable resource for healthcare institutions seeking to establish and optimize CDWs, offering a roadmap for leveraging clinical data for research, quality enhancement, and improved patient care.
{"title":"Barriers encountered with clinical data warehouses: Recommendations from a focus group","authors":"","doi":"10.1016/j.cmpb.2024.108404","DOIUrl":"10.1016/j.cmpb.2024.108404","url":null,"abstract":"<div><h3>Background and Objective</h3><p>The increasing implementation and use of electronic health records over the last few decades has made a significant volume of clinical data being available. Over the past 20 years, hospitals have also adopted and implemented data warehouse technology to facilitate the reuse of administrative and clinical data for research. However, the implementation of clinical data warehouses encounters a set of barriers: ethical, legislative, technical, human and organizational. This paper proposes an overview of difficulties and barriers encountered during a clinical data warehouse (CDW) development and implementation project.</p></div><div><h3>Methods</h3><p>We conducted a focus group at the 2023 Medical Informatics Europe Conference and invited professionals involved in the implementation of CDW. These experts described their CDW and the difficulties and barriers they encountered at each phase: (i) launching of the data warehouse project, (ii) implementing the data warehouse and (iii) using a data warehouse in routine operations. They were also asked to propose solutions they were able to implement to address the barriers previously reported.</p></div><div><h3>Results</h3><p>After synthesis and consensus, a total of 26 barriers were identified, 10 pertained to tasks, 5 to tools and technologies, 4 to persons, 4 to organization, and 3 to the external environment. To address these challenges, a set of 15 practical recommendations was offered, covering essential aspects such as governance, stakeholder engagement, interdisciplinary collaboration, and external expertise utilization.</p></div><div><h3>Conclusions</h3><p>These recommendations serve as a valuable resource for healthcare institutions seeking to establish and optimize CDWs, offering a roadmap for leveraging clinical data for research, quality enhancement, and improved patient care.</p></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":null,"pages":null},"PeriodicalIF":4.9,"publicationDate":"2024-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142145346","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-30DOI: 10.1016/j.cmpb.2024.108403
Background
Acute heart failure (AHF) in the intensive care unit (ICU) is characterized by its criticality, rapid progression, complex and changeable condition, and its pathophysiological process involves the interaction of multiple organs and systems. This makes it difficult to predict in-hospital mortality events comprehensively and accurately. Traditional analysis methods based on statistics and machine learning suffer from insufficient model performance, poor accuracy caused by prior dependence, and difficulty in adequately considering the complex relationships between multiple risk factors. Therefore, the application of deep neural network (DNN) techniques to the specific scenario, predicting mortality events of patients with AHF under intensive care, has become a research frontier.
Methods
This research utilized the MIMIC-IV critical care database as the primary data source and employed the synthetic minority over-sampling technique (SMOTE) to balance the dataset. Deep neural network models—backpropagation neural network (BPNN) and recurrent neural network (RNN), which are based on electronic medical record data mining, were employed to investigate the in-hospital death event judgment task of patients with AHF under intensive care. Additionally, multiple single machine learning models and ensemble learning models were constructed for comparative experiments. Moreover, we achieved various optimal performance combinations by modifying the classification threshold of deep neural network models to address the diverse real-world requirements in the ICU. Finally, we conducted an interpretable deep model using SHapley Additive exPlanations (SHAP) to uncover the most influential medical record features for each patient from the aspects of global and local interpretation.
Results
In terms of model performance in this scenario, deep neural network models outperform both single machine learning models and ensemble learning models, achieving the highest Accuracy, Precision, Recall, F1 value, and Area under the ROC curve, which can reach 0.949, 0.925, 0.983, 0.953, and 0.987 respectively. SHAP value analysis revealed that the ICU scores (APSIII, OASIS, SOFA) are significantly correlated with the occurrence of in-hospital fatal events.
Conclusions
Our study underscores that DNN-based mortality event classifier offers a novel intelligent approach for forecasting and assessing the prognosis of AHF patients in the ICU. Additionally, the ICU scores stand out as the most predictive features, which implies that in the decision-making process of the models, ICU scores can provide the most crucial information, making the greatest positive or negative contribution to influence the incidence of in-hospital mortality among patients with acute heart failure.
{"title":"Prediction of mortality events of patients with acute heart failure in intensive care unit based on deep neural network","authors":"","doi":"10.1016/j.cmpb.2024.108403","DOIUrl":"10.1016/j.cmpb.2024.108403","url":null,"abstract":"<div><h3>Background</h3><p>Acute heart failure (AHF) in the intensive care unit (ICU) is characterized by its criticality, rapid progression, complex and changeable condition, and its pathophysiological process involves the interaction of multiple organs and systems. This makes it difficult to predict in-hospital mortality events comprehensively and accurately. Traditional analysis methods based on statistics and machine learning suffer from insufficient model performance, poor accuracy caused by prior dependence, and difficulty in adequately considering the complex relationships between multiple risk factors. Therefore, the application of deep neural network (DNN) techniques to the specific scenario, predicting mortality events of patients with AHF under intensive care, has become a research frontier.</p></div><div><h3>Methods</h3><p>This research utilized the MIMIC-IV critical care database as the primary data source and employed the synthetic minority over-sampling technique (SMOTE) to balance the dataset. Deep neural network models—backpropagation neural network (BPNN) and recurrent neural network (RNN), which are based on electronic medical record data mining, were employed to investigate the in-hospital death event judgment task of patients with AHF under intensive care. Additionally, multiple single machine learning models and ensemble learning models were constructed for comparative experiments. Moreover, we achieved various optimal performance combinations by modifying the classification threshold of deep neural network models to address the diverse real-world requirements in the ICU. Finally, we conducted an interpretable deep model using SHapley Additive exPlanations (SHAP) to uncover the most influential medical record features for each patient from the aspects of global and local interpretation.</p></div><div><h3>Results</h3><p>In terms of model performance in this scenario, deep neural network models outperform both single machine learning models and ensemble learning models, achieving the highest Accuracy, Precision, Recall, F1 value, and Area under the ROC curve, which can reach 0.949, 0.925, 0.983, 0.953, and 0.987 respectively. SHAP value analysis revealed that the ICU scores (APSIII, OASIS, SOFA) are significantly correlated with the occurrence of in-hospital fatal events.</p></div><div><h3>Conclusions</h3><p>Our study underscores that DNN-based mortality event classifier offers a novel intelligent approach for forecasting and assessing the prognosis of AHF patients in the ICU. Additionally, the ICU scores stand out as the most predictive features, which implies that in the decision-making process of the models, ICU scores can provide the most crucial information, making the greatest positive or negative contribution to influence the incidence of in-hospital mortality among patients with acute heart failure.</p></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":null,"pages":null},"PeriodicalIF":4.9,"publicationDate":"2024-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S0169260724003961/pdfft?md5=603e2fe58aafff457c4aa49d7025c8c7&pid=1-s2.0-S0169260724003961-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142137113","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-29DOI: 10.1016/j.cmpb.2024.108397
Background and Objectives:
Genome-scale metabolic networks (GEMs) represent a valuable modeling and computational tool in the broad field of systems biology. Their ability to integrate constraints and high-throughput biological data enables the study of intricate metabolic aspects and processes of different cell types and conditions. The past decade has witnessed an increasing number and variety of applications of GEMs for the study of human diseases, along with a huge effort aimed at the reconstruction, integration and analysis of a high number of organisms. This paper presents a systematic review of the scientific literature, to pursue several important questions about the application of constraint-based modeling in the investigation of human diseases. Hopefully, this paper will provide a useful reference for researchers interested in the application of modeling and computational tools for the investigation of metabolic-related human diseases.
Methods:
This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Elsevier Scopus®, National Library of Medicine PubMed® and Clarivate Web of Science™ databases were enquired, resulting in 566 scientific articles. After applying exclusion and eligibility criteria, a total of 169 papers were selected and individually examined.
Results:
The reviewed papers offer a thorough and up-to-date picture of the latest modeling and computational approaches, based on genome-scale metabolic models, that can be leveraged for the investigation of a large variety of human diseases. The numerous studies have been categorized according to the clinical research area involved in the examined disease. Furthermore, the paper discusses the most typical approaches employed to derive clinically-relevant information using the computational models.
Conclusions:
The number of scientific papers, utilizing GEM-based approaches for the investigation of human diseases, suggests an increasing interest in these types of approaches; hopefully, the present review will represent a useful reference for scientists interested in applying computational modeling approaches to investigate the aetiopathology of human diseases; we also hope that this work will foster the development of novel applications and methods for the discovery of clinically-relevant insights on metabolic-related diseases.
背景与目的:基因组尺度代谢网络(GEM)是系统生物学广泛领域中一种宝贵的建模和计算工具。它能够整合约束条件和高通量生物数据,有助于研究不同细胞类型和条件下错综复杂的代谢问题和过程。在过去的十年中,GEMs 在人类疾病研究中的应用越来越多,种类也越来越丰富,同时在重建、整合和分析大量生物体方面也做出了巨大努力。本文对科学文献进行了系统回顾,探讨了基于约束的建模在人类疾病研究中的应用的几个重要问题。希望本文能为那些对应用建模和计算工具研究代谢相关人类疾病感兴趣的研究人员提供有用的参考。方法:本系统综述是根据系统综述和元分析首选报告项目(PRISMA)指南进行的。查询了 Elsevier Scopus®、美国国家医学图书馆 PubMed® 和 Clarivate Web of Science™ 数据库,共获得 566 篇科学文章。结果:综述论文全面介绍了基于基因组尺度代谢模型的最新建模和计算方法,可用于研究多种人类疾病。众多研究根据所研究疾病涉及的临床研究领域进行了分类。结论:利用基于 GEM 的方法研究人类疾病的科学论文数量表明,人们对这类方法的兴趣与日俱增;希望本综述能为有意应用计算建模方法研究人类疾病病因病理学的科学家提供有用的参考;我们还希望这项工作能促进新型应用和方法的开发,以发现代谢相关疾病的临床相关见解。
{"title":"Applications of genome-scale metabolic models to the study of human diseases: A systematic review","authors":"","doi":"10.1016/j.cmpb.2024.108397","DOIUrl":"10.1016/j.cmpb.2024.108397","url":null,"abstract":"<div><h3>Background and Objectives:</h3><p>Genome-scale metabolic networks (GEMs) represent a valuable modeling and computational tool in the broad field of systems biology. Their ability to integrate constraints and high-throughput biological data enables the study of intricate metabolic aspects and processes of different cell types and conditions. The past decade has witnessed an increasing number and variety of applications of GEMs for the study of human diseases, along with a huge effort aimed at the reconstruction, integration and analysis of a high number of organisms. This paper presents a systematic review of the scientific literature, to pursue several important questions about the application of constraint-based modeling in the investigation of human diseases. Hopefully, this paper will provide a useful reference for researchers interested in the application of modeling and computational tools for the investigation of metabolic-related human diseases.</p></div><div><h3>Methods:</h3><p>This systematic review was conducted according to the <em>Preferred Reporting Items for Systematic Reviews and Meta-Analyses</em> (PRISMA) guidelines. Elsevier Scopus®, National Library of Medicine PubMed® and Clarivate Web of Science™ databases were enquired, resulting in 566 scientific articles. After applying exclusion and eligibility criteria, a total of 169 papers were selected and individually examined.</p></div><div><h3>Results:</h3><p>The reviewed papers offer a thorough and up-to-date picture of the latest modeling and computational approaches, based on genome-scale metabolic models, that can be leveraged for the investigation of a large variety of human diseases. The numerous studies have been categorized according to the clinical research area involved in the examined disease. Furthermore, the paper discusses the most typical approaches employed to derive clinically-relevant information using the computational models.</p></div><div><h3>Conclusions:</h3><p>The number of scientific papers, utilizing GEM-based approaches for the investigation of human diseases, suggests an increasing interest in these types of approaches; hopefully, the present review will represent a useful reference for scientists interested in applying computational modeling approaches to investigate the aetiopathology of human diseases; we also hope that this work will foster the development of novel applications and methods for the discovery of clinically-relevant insights on metabolic-related diseases.</p></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":null,"pages":null},"PeriodicalIF":4.9,"publicationDate":"2024-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S0169260724003900/pdfft?md5=84ba91d2da746a463bbd24ee7f1ef111&pid=1-s2.0-S0169260724003900-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142129475","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-28DOI: 10.1016/j.cmpb.2024.108401
Background and Objective
Registration of pulmonary computed tomography (CT) images with radiation-induced lung diseases (RILD) was essential to investigate the voxel-wise relationship between the formation of RILD and the radiation dose received by different tissues. Although various approaches had been developed for the registration of lung CTs, their performances remained clinically unsatisfactory for registration of lung CT images with RILD. The main difficulties arose from the longitudinal change in lung parenchyma, including RILD and volumetric change of lung cancers, after radiation therapy, leading to inaccurate registration and artifacts caused by erroneous matching of the RILD tissues.
Methods
To overcome the influence of the parenchymal changes, a divide-and-conquer approach rooted in the coherent point drift (CPD) paradigm was proposed. The proposed method was based on two kernel ideas. One was the idea of component structure wise registration. Specifically, the proposed method relaxed the intrinsic assumption of equal isotropic covariances in CPD by decomposing a lung and its surrounding tissues into component structures and independently registering the component structures pairwise by CPD. The other was the idea of defining a vascular subtree centered at a matched branch point as a component structure. This idea could not only provide a sufficient number of matched feature points within a parenchyma, but avoid being corrupted by the false feature points resided in the RILD tissues due to globally and indiscriminately sampling using mathematical operators. The overall deformation model was built by using the Thin Plate Spline based on all matched points.
Results
This study recruited 30 pairs of lung CT images with RILD, 15 of which were used for internal validation (leave-one-out cross-validation) and the other 15 for external validation. The experimental results showed that the proposed algorithm achieved a mean and a mean of maximum 1 % of average surface distances <2 and 8 mm, respectively, and a mean and a maximum target registration error <2 mm and 5 mm on both internal and external validation datasets. The paired two-sample t-tests corroborated that the proposed algorithm outperformed a recent method, the Stavropoulou's method, on the external validation dataset (p < 0.05).
Conclusions
The proposed algorithm effectively reduced the influence of parenchymal changes, resulting in a reasonably accurate and artifact-free registration.
{"title":"Longitudinal registration of thoracic CT images with radiation-induced lung diseases: A divide-and-conquer approach based on component structure wise registration using coherent point drift","authors":"","doi":"10.1016/j.cmpb.2024.108401","DOIUrl":"10.1016/j.cmpb.2024.108401","url":null,"abstract":"<div><h3>Background and Objective</h3><p>Registration of pulmonary computed tomography (CT) images with radiation-induced lung diseases (RILD) was essential to investigate the voxel-wise relationship between the formation of RILD and the radiation dose received by different tissues. Although various approaches had been developed for the registration of lung CTs, their performances remained clinically unsatisfactory for registration of lung CT images with RILD. The main difficulties arose from the longitudinal change in lung parenchyma, including RILD and volumetric change of lung cancers, after radiation therapy, leading to inaccurate registration and artifacts caused by erroneous matching of the RILD tissues.</p></div><div><h3>Methods</h3><p>To overcome the influence of the parenchymal changes, a divide-and-conquer approach rooted in the coherent point drift (CPD) paradigm was proposed. The proposed method was based on two kernel ideas. One was the idea of component structure wise registration. Specifically, the proposed method relaxed the intrinsic assumption of equal isotropic covariances in CPD by decomposing a lung and its surrounding tissues into component structures and independently registering the component structures pairwise by CPD. The other was the idea of defining a vascular subtree centered at a matched branch point as a component structure. This idea could not only provide a sufficient number of matched feature points within a parenchyma, but avoid being corrupted by the false feature points resided in the RILD tissues due to globally and indiscriminately sampling using mathematical operators. The overall deformation model was built by using the Thin Plate Spline based on all matched points.</p></div><div><h3>Results</h3><p>This study recruited 30 pairs of lung CT images with RILD, 15 of which were used for internal validation (leave-one-out cross-validation) and the other 15 for external validation. The experimental results showed that the proposed algorithm achieved a mean and a mean of maximum 1 % of average surface distances <2 and 8 mm, respectively, and a mean and a maximum target registration error <2 mm and 5 mm on both internal and external validation datasets. The paired two-sample <em>t</em>-tests corroborated that the proposed algorithm outperformed a recent method, the Stavropoulou's method, on the external validation dataset (<em>p</em> < 0.05).</p></div><div><h3>Conclusions</h3><p>The proposed algorithm effectively reduced the influence of parenchymal changes, resulting in a reasonably accurate and artifact-free registration.</p></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":null,"pages":null},"PeriodicalIF":4.9,"publicationDate":"2024-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142129577","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-28DOI: 10.1016/j.cmpb.2024.108399
<div><p>Magnetic Resonance (MR) parameters mapping in muscle Magnetic Resonance Imaging (mMRI) is predominantly performed using pattern recognition-based algorithms, which are characterised by high computational costs and scalability issues in the context of multi-parametric mapping.</p><p>Deep Learning (DL) has been demonstrated to be a robust and efficient method for rapid MR parameters mapping. However, its application in mMRI domain to investigate Neuromuscular Disorders (NMDs) has not yet been explored. In addition, data-driven DL models suffered in interpretation and explainability of the learning process. We developed a Physics Informed Neural Network called Myo-Regressor Deep Informed Neural NetwOrk (Myo-DINO) for efficient and explainable Fat Fraction (FF), water-T<sub>2</sub> (wT<sub>2</sub>) and B1 mapping from a cohort of NMDs.A total of 2165 slices (232 subjects) from Multi-Echo Spin Echo (MESE) images were selected as the input dataset for which FF, wT<sub>2</sub>,B1 ground truth maps were computed using the MyoQMRI toolbox. This toolbox exploits the Extended Phase Graph (EPG) theory with a two-component model (water and fat signal) and slice profile to simulate the signal evolution in the MESE framework. A customized U-Net architecture was implemented as the Myo-DINO architecture. The squared L<sub>2</sub> norm loss was complemented by two distinct physics models to define two ‘Physics-Informed’ loss functions: <em>Cycling Loss 1</em> embedded a mono-exponential model to describe the relaxation of water protons, while <em>Cycling Loss 2</em> incorporated the EPG theory with slice profile to model the magnetization dephasing under the effect of gradients and RF pulses. The Myo-DINO was trained with the hyperparameter value of the 'Physics-Informed' component held constant, i.e. λ<sub>model</sub> = 1, while different hyperparameter values (λ<sub>cnn</sub>) were applied to the squared L<sub>2</sub> norm component in both the cycling loss. In particular, hard (λ<sub>cnn</sub>=10), normal (λ<sub>cnn</sub>=1) and self-supervised (λ<sub>cnn</sub>=0) constraints were applied to gradually decrease the impact of the squared L<sub>2</sub> norm component on the ‘Physics Informed’ term during the Myo-DINO training process.</p><p>Myo-DINO achieved higher performance with <em>Cycling Loss 2</em> for FF, wT<sub>2</sub> and B1 prediction. In particular, high reconstruction similarity and quality (Structural Similarity Index > 0.92, Peak Signal to Noise ratio > 30.0 db) and small reconstruction error (Normalized Root Mean Squared Error < 0.038) to the reference maps were shown with self-supervised weighting of the <em>Cycling Loss 2</em>. In addition muscle-wise FF, wT<sub>2</sub> and B1 predicted values showed good agreement with the reference values. The Myo-DINO has been demonstrated to be a robust and efficient workflow for MR parameters mapping in the context of mMRI. This provides preliminary evidence that it can be an effective alter
{"title":"Myo-regressor Deep Informed Neural NetwOrk (Myo-DINO) for fast MR parameters mapping in neuromuscular disorders","authors":"","doi":"10.1016/j.cmpb.2024.108399","DOIUrl":"10.1016/j.cmpb.2024.108399","url":null,"abstract":"<div><p>Magnetic Resonance (MR) parameters mapping in muscle Magnetic Resonance Imaging (mMRI) is predominantly performed using pattern recognition-based algorithms, which are characterised by high computational costs and scalability issues in the context of multi-parametric mapping.</p><p>Deep Learning (DL) has been demonstrated to be a robust and efficient method for rapid MR parameters mapping. However, its application in mMRI domain to investigate Neuromuscular Disorders (NMDs) has not yet been explored. In addition, data-driven DL models suffered in interpretation and explainability of the learning process. We developed a Physics Informed Neural Network called Myo-Regressor Deep Informed Neural NetwOrk (Myo-DINO) for efficient and explainable Fat Fraction (FF), water-T<sub>2</sub> (wT<sub>2</sub>) and B1 mapping from a cohort of NMDs.A total of 2165 slices (232 subjects) from Multi-Echo Spin Echo (MESE) images were selected as the input dataset for which FF, wT<sub>2</sub>,B1 ground truth maps were computed using the MyoQMRI toolbox. This toolbox exploits the Extended Phase Graph (EPG) theory with a two-component model (water and fat signal) and slice profile to simulate the signal evolution in the MESE framework. A customized U-Net architecture was implemented as the Myo-DINO architecture. The squared L<sub>2</sub> norm loss was complemented by two distinct physics models to define two ‘Physics-Informed’ loss functions: <em>Cycling Loss 1</em> embedded a mono-exponential model to describe the relaxation of water protons, while <em>Cycling Loss 2</em> incorporated the EPG theory with slice profile to model the magnetization dephasing under the effect of gradients and RF pulses. The Myo-DINO was trained with the hyperparameter value of the 'Physics-Informed' component held constant, i.e. λ<sub>model</sub> = 1, while different hyperparameter values (λ<sub>cnn</sub>) were applied to the squared L<sub>2</sub> norm component in both the cycling loss. In particular, hard (λ<sub>cnn</sub>=10), normal (λ<sub>cnn</sub>=1) and self-supervised (λ<sub>cnn</sub>=0) constraints were applied to gradually decrease the impact of the squared L<sub>2</sub> norm component on the ‘Physics Informed’ term during the Myo-DINO training process.</p><p>Myo-DINO achieved higher performance with <em>Cycling Loss 2</em> for FF, wT<sub>2</sub> and B1 prediction. In particular, high reconstruction similarity and quality (Structural Similarity Index > 0.92, Peak Signal to Noise ratio > 30.0 db) and small reconstruction error (Normalized Root Mean Squared Error < 0.038) to the reference maps were shown with self-supervised weighting of the <em>Cycling Loss 2</em>. In addition muscle-wise FF, wT<sub>2</sub> and B1 predicted values showed good agreement with the reference values. The Myo-DINO has been demonstrated to be a robust and efficient workflow for MR parameters mapping in the context of mMRI. This provides preliminary evidence that it can be an effective alter","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":null,"pages":null},"PeriodicalIF":4.9,"publicationDate":"2024-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S0169260724003924/pdfft?md5=8ce4157dd71dac18b8abf08374f3fc22&pid=1-s2.0-S0169260724003924-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142137111","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-28DOI: 10.1016/j.cmpb.2024.108402
Background
This study aimed to predict early adolescent sleep problems using pregnancy and childbirth risk factors through machine learning algorithms, and to evaluate model performance internally and externally.
Methods
Data from the China Jintan Child Cohort study (CJCC; n=848) for model development and the US Healthy Brain and Behavior Study (HBBS; n=454) for external validation were employed. Maternal pregnancy histories, obstetric data, and adolescent sleep problems were collected. Several machine learning techniques were employed, including least absolute shrinkage and selection operator, logistic regression, random forest, naïve bayes, extreme gradient boosting, decision tree, and neural network. The area under the receiver operating characteristic curve, sensitivity, specificity, accuracy, and root mean square of residuals were used to evaluate model performance.
Results
Key predictors for CJCC adolescents’ sleep problems include gestational age, birthweight, duration of delivery, and maternal happiness during pregnancy. In HBBS adolescents, the duration of postnatal depressive emotions was the primary perinatal predictor. The prediction models developed in the CJCC had good-to-excellent internal validation performance but poor performance in predicting the sleep problems in HBBS adolescents.
Conclusion
The identification of specific perinatal risk factors associated with adolescent sleep problems can inform targeted interventions during and after pregnancy to mitigate these risks. Health providers should consider integrating these predictive factors into routine pre- and postnatal assessments to identify at-risk populations. The variability in model performance across different cohorts highlights the need for context-specific models and the cautious application of predictive analytics across diverse populations. Future research should focus on refining predictive models to account for such variations, potentially through the incorporation of additional socio-cultural factors and genetic markers. This study emphasizes the importance of personalized and culturally sensitive approaches in the prediction and management of adolescent sleep problems, leveraging advanced computational methods to enhance maternal and child health outcomes.
{"title":"Machine learning approach to investigate pregnancy and childbirth risk factors of sleep problems in early adolescence: Evidence from two cohort studies","authors":"","doi":"10.1016/j.cmpb.2024.108402","DOIUrl":"10.1016/j.cmpb.2024.108402","url":null,"abstract":"<div><h3>Background</h3><p>This study aimed to predict early adolescent sleep problems using pregnancy and childbirth risk factors through machine learning algorithms, and to evaluate model performance internally and externally.</p></div><div><h3>Methods</h3><p>Data from the China Jintan Child Cohort study (CJCC; n=848) for model development and the US Healthy Brain and Behavior Study (HBBS; n=454) for external validation were employed. Maternal pregnancy histories, obstetric data, and adolescent sleep problems were collected. Several machine learning techniques were employed, including least absolute shrinkage and selection operator, logistic regression, random forest, naïve bayes, extreme gradient boosting, decision tree, and neural network. The area under the receiver operating characteristic curve, sensitivity, specificity, accuracy, and root mean square of residuals were used to evaluate model performance.</p></div><div><h3>Results</h3><p>Key predictors for CJCC adolescents’ sleep problems include gestational age, birthweight, duration of delivery, and maternal happiness during pregnancy. In HBBS adolescents, the duration of postnatal depressive emotions was the primary perinatal predictor. The prediction models developed in the CJCC had good-to-excellent internal validation performance but poor performance in predicting the sleep problems in HBBS adolescents.</p></div><div><h3>Conclusion</h3><p>The identification of specific perinatal risk factors associated with adolescent sleep problems can inform targeted interventions during and after pregnancy to mitigate these risks. Health providers should consider integrating these predictive factors into routine pre- and postnatal assessments to identify at-risk populations. The variability in model performance across different cohorts highlights the need for context-specific models and the cautious application of predictive analytics across diverse populations. Future research should focus on refining predictive models to account for such variations, potentially through the incorporation of additional socio-cultural factors and genetic markers. This study emphasizes the importance of personalized and culturally sensitive approaches in the prediction and management of adolescent sleep problems, leveraging advanced computational methods to enhance maternal and child health outcomes.</p></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":null,"pages":null},"PeriodicalIF":4.9,"publicationDate":"2024-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142121788","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-28DOI: 10.1016/j.cmpb.2024.108398
Background and Objective:
Tendon segmentation is crucial for studying tendon-related pathologies like tendinopathy, tendinosis, etc. This step further enables detailed analysis of specific tendon regions using automated or semi-automated methods. This study specifically aims at the segmentation of Achilles tendon, the largest tendon in the human body.
Methods:
This study proposes a comprehensive end-to-end tendon segmentation module composed of a preliminary superpixel-based coarse segmentation preceding the final segmentation task. The final segmentation results are obtained through two distinct approaches. In the first approach, the coarsely generated superpixels are subjected to classification using Random Forest (RF) and Support Vector Machine (SVM) classifiers to classify whether each superpixel belongs to a tendon class or not (resulting in tendon segmentation). In the second approach, the arrangements of superpixels are converted to graphs instead of being treated as conventional image grids. This classification process uses a graph-based convolutional network (GCN) to determine whether each superpixel corresponds to a tendon class or not.
Results:
All experiments are conducted on a custom-made ankle MRI dataset. The dataset comprises 76 subjects and is divided into two sets: one for training (Dataset 1, trained and evaluated using leave-one-group-out cross-validation) and the other as unseen test data (Dataset 2). Using our first approach, the final test AUC (Area Under the ROC Curve) scores using RF and SVM classifiers on the test data (Dataset 2) are 0.992 and 0.987, respectively, with sensitivities of 0.904 and 0.966. On the other hand, using our second approach (GCN-based node classification), the AUC score for the test set is 0.933 with a sensitivity of 0.899.
Conclusions:
Our proposed pipeline demonstrates the efficacy of employing superpixel generation as a coarse segmentation technique for the final tendon segmentation. Whether utilizing RF, SVM-based superpixel classification, or GCN-based classification for tendon segmentation, our system consistently achieves commendable AUC scores, especially the non-graph-based approach. Given the limited dataset, our graph-based method did not perform as well as non-graph-based superpixel classifications; however, the results obtained provide valuable insights into how well the models can distinguish between tendons and non-tendons. This opens up opportunities for further exploration and improvement.
{"title":"Beyond pixel: Superpixel-based MRI segmentation through traditional machine learning and graph convolutional network","authors":"","doi":"10.1016/j.cmpb.2024.108398","DOIUrl":"10.1016/j.cmpb.2024.108398","url":null,"abstract":"<div><h3>Background and Objective:</h3><p>Tendon segmentation is crucial for studying tendon-related pathologies like tendinopathy, tendinosis, etc. This step further enables detailed analysis of specific tendon regions using automated or semi-automated methods. This study specifically aims at the segmentation of Achilles tendon, the largest tendon in the human body.</p></div><div><h3>Methods:</h3><p>This study proposes a comprehensive end-to-end tendon segmentation module composed of a preliminary superpixel-based coarse segmentation preceding the final segmentation task. The final segmentation results are obtained through two distinct approaches. In the first approach, the coarsely generated superpixels are subjected to classification using Random Forest (RF) and Support Vector Machine (SVM) classifiers to classify whether each superpixel belongs to a tendon class or not (resulting in tendon segmentation). In the second approach, the arrangements of superpixels are converted to graphs instead of being treated as conventional image grids. This classification process uses a graph-based convolutional network (GCN) to determine whether each superpixel corresponds to a tendon class or not.</p></div><div><h3>Results:</h3><p>All experiments are conducted on a custom-made ankle MRI dataset. The dataset comprises 76 subjects and is divided into two sets: one for training (Dataset 1, trained and evaluated using leave-one-group-out cross-validation) and the other as unseen test data (Dataset 2). Using our first approach, the final test AUC (Area Under the ROC Curve) scores using RF and SVM classifiers on the test data (Dataset 2) are 0.992 and 0.987, respectively, with sensitivities of 0.904 and 0.966. On the other hand, using our second approach (GCN-based node classification), the AUC score for the test set is 0.933 with a sensitivity of 0.899.</p></div><div><h3>Conclusions:</h3><p>Our proposed pipeline demonstrates the efficacy of employing superpixel generation as a coarse segmentation technique for the final tendon segmentation. Whether utilizing RF, SVM-based superpixel classification, or GCN-based classification for tendon segmentation, our system consistently achieves commendable AUC scores, especially the non-graph-based approach. Given the limited dataset, our graph-based method did not perform as well as non-graph-based superpixel classifications; however, the results obtained provide valuable insights into how well the models can distinguish between tendons and non-tendons. This opens up opportunities for further exploration and improvement.</p></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":null,"pages":null},"PeriodicalIF":4.9,"publicationDate":"2024-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S0169260724003912/pdfft?md5=ec546df96619be9e2501c6602cc6ef74&pid=1-s2.0-S0169260724003912-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142137112","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-26DOI: 10.1016/j.cmpb.2024.108395
Background and objective:
The use of deep learning to undertake shape analysis of the complexities of the human head holds great promise. However, there have traditionally been a number of barriers to accurate modelling, especially when operating on both a global and local level.
Methods:
In this work, we will discuss the application of the Swap Disentangled Variational Autoencoder (SD-VAE) with relevance to Crouzon, Apert and Muenke syndromes. The model is trained on a dataset of 3D meshes of healthy and syndromic patients which was increased in size with a novel data augmentation technique based on spectral interpolation. Thanks to its semantically meaningful and disentangled latent representation, SD-VAE is used to analyse and generate head shapes while considering the influence of different anatomical sub-units.
Results:
Although syndrome classification is performed on the entire mesh, it is also possible, for the first time, to analyse the influence of each region of the head on the syndromic phenotype. By manipulating specific parameters of the generative model, and producing procedure-specific new shapes, it is also possible to approximate the outcome of a range of craniofacial surgical procedures.
Conclusion:
This work opens new avenues to advance diagnosis, aids surgical planning and allows for the objective evaluation of surgical outcomes. Our code is available at github.com/simofoti/CraniofacialSD-VAE.
{"title":"Latent disentanglement in mesh variational autoencoders improves the diagnosis of craniofacial syndromes and aids surgical planning","authors":"","doi":"10.1016/j.cmpb.2024.108395","DOIUrl":"10.1016/j.cmpb.2024.108395","url":null,"abstract":"<div><h3>Background and objective:</h3><p>The use of deep learning to undertake shape analysis of the complexities of the human head holds great promise. However, there have traditionally been a number of barriers to accurate modelling, especially when operating on both a global and local level.</p></div><div><h3>Methods:</h3><p>In this work, we will discuss the application of the Swap Disentangled Variational Autoencoder (SD-VAE) with relevance to Crouzon, Apert and Muenke syndromes. The model is trained on a dataset of 3D meshes of healthy and syndromic patients which was increased in size with a novel data augmentation technique based on spectral interpolation. Thanks to its semantically meaningful and disentangled latent representation, SD-VAE is used to analyse and generate head shapes while considering the influence of different anatomical sub-units.</p></div><div><h3>Results:</h3><p>Although syndrome classification is performed on the entire mesh, it is also possible, for the first time, to analyse the influence of each region of the head on the syndromic phenotype. By manipulating specific parameters of the generative model, and producing procedure-specific new shapes, it is also possible to approximate the outcome of a range of craniofacial surgical procedures.</p></div><div><h3>Conclusion:</h3><p>This work opens new avenues to advance diagnosis, aids surgical planning and allows for the objective evaluation of surgical outcomes. Our code is available at <span><span>github.com/simofoti/CraniofacialSD-VAE</span><svg><path></path></svg></span>.</p></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":null,"pages":null},"PeriodicalIF":4.9,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S0169260724003882/pdfft?md5=7bb50e4fab3ca51c5867372f1437b450&pid=1-s2.0-S0169260724003882-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142097710","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}