Background: Deep learning models for patch classification in whole-slide images (WSIs) have shown promise in assisting follicular lymphoma grading. However, these models often require pathologists to identify centroblasts and manually provide refined labels for model optimization. Objective: To address this limitation, we propose PseudoCell , an object detection framework for automated centroblast detection in WSI, eliminating the need for extensive pathologist's refined labels. Methods: PseudoCell leverages a combination of pathologist-provided centroblast labels and pseudo-negative labels generated from undersampled false-positive predictions based on cell morphology features. This approach reduces the reliance on time-consuming manual annotations. Results: Our framework significantly reduces the workload for pathologists by accurately identifying and narrowing down areas of interest containing centroblasts. Depending on the confidence threshold, PseudoCell can eliminate 58.18-99.35% of irrelevant tissue areas on WSI, streamlining the diagnostic process. Conclusion: This study presents PseudoCell as a practical and efficient prescreening method for centroblast detection, eliminating the need for refined labels from pathologists. The discussion section provides detailed guidance for implementing PseudoCell in clinical practice.
背景：用于全切片图像（WSI）斑块分类的深度学习模型在辅助滤泡性淋巴瘤分级方面大有可为。然而，这些模型通常需要病理学家识别中心母细胞，并手动提供用于模型优化的精细标签。目的：为了解决这一局限性，我们提出了一个对象检测框架--PseudoCell，用于自动检测 WSI 中的成中心细胞，无需病理学家提供大量的精细标签。方法PseudoCell 综合利用了病理学家提供的中心母细胞标签和根据细胞形态特征从采样不足的假阳性预测中生成的伪阴性标签。这种方法减少了对耗时的人工注释的依赖。结果我们的框架能准确识别并缩小含有中心母细胞的感兴趣区，从而大大减轻了病理学家的工作量。根据置信度阈值的不同，PseudoCell 可以消除 WSI 上 58.18-99.35% 的无关组织区域，从而简化诊断过程。结论本研究提出的伪细胞是一种实用、高效的成中心细胞检测预筛选方法，无需病理学家进行精细标记。讨论部分为在临床实践中使用伪细胞提供了详细指导。
{"title":"PseudoCell: Hard Negative Mining as Pseudo Labeling for Deep Learning-Based Centroblast Cell Detection","authors":"Narongrid Seesawad;Piyalitt Ittichaiwong;Thapanun Sudhawiyangkul;Phattarapong Sawangjai;Peti Thuwajit;Paisarn Boonsakan;Supasan Sripodok;Kanyakorn Veerakanjana;Komgrid Charngkaew;Ananya Pongpaibul;Napat Angkathunyakul;Narit Hnoohom;Sumeth Yuenyong;Chanitra Thuwajit;Theerawit Wilaiprasitporn","doi":"10.1109/OJEMB.2024.3407351","DOIUrl":"10.1109/OJEMB.2024.3407351","url":null,"abstract":"<italic>Background:</i>\u0000 Deep learning models for patch classification in whole-slide images (WSIs) have shown promise in assisting follicular lymphoma grading. However, these models often require pathologists to identify centroblasts and manually provide refined labels for model optimization. \u0000<italic>Objective:</i>\u0000 To address this limitation, we propose \u0000<italic>PseudoCell</i>\u0000, an object detection framework for automated centroblast detection in WSI, eliminating the need for extensive pathologist's refined labels. \u0000<italic>Methods:</i>\u0000 \u0000<italic>PseudoCell</i>\u0000 leverages a combination of pathologist-provided centroblast labels and pseudo-negative labels generated from undersampled false-positive predictions based on cell morphology features. This approach reduces the reliance on time-consuming manual annotations. \u0000<italic>Results:</i>\u0000 Our framework significantly reduces the workload for pathologists by accurately identifying and narrowing down areas of interest containing centroblasts. Depending on the confidence threshold, \u0000<italic>PseudoCell</i>\u0000 can eliminate 58.18-99.35% of irrelevant tissue areas on WSI, streamlining the diagnostic process. \u0000<italic>Conclusion:</i>\u0000 This study presents \u0000<italic>PseudoCell</i>\u0000 as a practical and efficient prescreening method for centroblast detection, eliminating the need for refined labels from pathologists. The discussion section provides detailed guidance for implementing \u0000<italic>PseudoCell</i>\u0000 in clinical practice.","PeriodicalId":33825,"journal":{"name":"IEEE Open Journal of Engineering in Medicine and Biology","volume":"5 ","pages":"514-523"},"PeriodicalIF":2.7,"publicationDate":"2024-03-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10542389","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141191635","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Goal: FetSAM represents a cutting-edge deep learning model aimed at revolutionizing fetal head ultrasound segmentation, thereby elevating prenatal diagnostic precision. Methods: Utilizing a comprehensive dataset–the largest to date for fetal head metrics–FetSAM incorporates prompt-based learning. It distinguishes itself with a dual loss mechanism, combining Weighted DiceLoss and Weighted Lovasz Loss, optimized through AdamW and underscored by class weight adjustments for better segmentation balance. Performance benchmarks against prominent models such as U-Net, DeepLabV3, and Segformer highlight its efficacy. Results: FetSAM delivers unparalleled segmentation accuracy, demonstrated by a DSC of 0.90117, HD of 1.86484, and ASD of 0.46645. Conclusion: FetSAM sets a new benchmark in AI-enhanced prenatal ultrasound analysis, providing a robust, precise tool for clinical applications and pushing the envelope of prenatal care with its groundbreaking dataset and segmentation capabilities.
{"title":"FetSAM: Advanced Segmentation Techniques for Fetal Head Biometrics in Ultrasound Imagery","authors":"Mahmood Alzubaidi;Uzair Shah;Marco Agus;Mowafa Househ","doi":"10.1109/OJEMB.2024.3382487","DOIUrl":"10.1109/OJEMB.2024.3382487","url":null,"abstract":"<italic>Goal:</i>\u0000 FetSAM represents a cutting-edge deep learning model aimed at revolutionizing fetal head ultrasound segmentation, thereby elevating prenatal diagnostic precision. \u0000<italic>Methods:</i>\u0000 Utilizing a comprehensive dataset–the largest to date for fetal head metrics–FetSAM incorporates prompt-based learning. It distinguishes itself with a dual loss mechanism, combining Weighted DiceLoss and Weighted Lovasz Loss, optimized through AdamW and underscored by class weight adjustments for better segmentation balance. Performance benchmarks against prominent models such as U-Net, DeepLabV3, and Segformer highlight its efficacy. \u0000<italic>Results:</i>\u0000 FetSAM delivers unparalleled segmentation accuracy, demonstrated by a DSC of 0.90117, HD of 1.86484, and ASD of 0.46645. \u0000<italic>Conclusion:</i>\u0000 FetSAM sets a new benchmark in AI-enhanced prenatal ultrasound analysis, providing a robust, precise tool for clinical applications and pushing the envelope of prenatal care with its groundbreaking dataset and segmentation capabilities.","PeriodicalId":33825,"journal":{"name":"IEEE Open Journal of Engineering in Medicine and Biology","volume":"5 ","pages":"281-295"},"PeriodicalIF":5.8,"publicationDate":"2024-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10480532","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140312995","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-27DOI: 10.1109/OJEMB.2024.3405666
Meng Jiao;Changyue Song;Xiaochen Xian;Shihao Yang;Feng Liu
Sleep Apnea (SA) is a prevalent sleep disorder with multifaceted etiologies that can have severe consequences for patients. Diagnosing SA traditionally relies on the in-laboratory polysomnogram (PSG), which records various human physiological activities overnight. SA diagnosis involves manual scoring by qualified physicians. Traditional machine learning methods for SA detection depend on hand-crafted features, making feature selection pivotal for downstream classification tasks. In recent years, deep learning has gained popularity in SA detection due to its capability for automatic feature extraction and superior classification accuracy. This study introduces a Deep Attention Network with Multi-Temporal Information Fusion (DAN-MTIF) for SA detection using single-lead electrocardiogram (ECG) signals. This framework utilizes three 1D convolutional neural network (CNN) blocks to extract features from R-R intervals and R-peak amplitudes using segments of varying lengths. Recognizing that features derived from different temporal scales vary in their contribution to classification, we integrate a multi-head attention module with a self-attention mechanism to learn the weights for each feature vector. Comprehensive experiments and comparisons between two paradigms of classical machine learning approaches and deep learning approaches are conducted. Our experiment results demonstrate that (1) compared with benchmark methods, the proposed DAN-MTIF exhibits excellent performance with 0.9106 accuracy, 0.9396 precision, 0.8470 sensitivity, 0.9588 specificity, and 0.8909 $F_{1}$ score at per-segment level; (2) DAN-MTIF can effectively extract features with a higher degree of discrimination from ECG segments of multiple timescales than those with a single time scale, ensuring a better SA detection performance; (3) the overall performance of deep learning methods is better than the classical machine learning algorithms, highlighting the superior performance of deep learning approaches for SA detection.
睡眠呼吸暂停（SA）是一种普遍存在的睡眠障碍，其病因是多方面的，可对患者造成严重后果。睡眠呼吸暂停的诊断传统上依赖于实验室多导睡眠图（PSG），它记录了人体在一夜之间的各种生理活动。SA 诊断需要由合格的医生进行人工评分。用于 SA 检测的传统机器学习方法依赖于手工创建的特征，因此特征选择对于下游分类任务至关重要。近年来，深度学习因其自动提取特征的能力和出色的分类准确性，在 SA 检测中越来越受欢迎。本研究介绍了利用单导联心电图（ECG）信号进行 SA 检测的多时空信息融合深度注意力网络（DAN-MTIF）。该框架利用三个一维卷积神经网络（CNN）块，使用不同长度的片段从 R-R 间期和 R 峰振幅中提取特征。由于从不同时间尺度提取的特征对分类的贡献各不相同，我们将多头注意模块与自我注意机制相结合，以学习每个特征向量的权重。我们在经典机器学习方法和深度学习方法的两种范例之间进行了全面的实验和比较。实验结果表明：(1) 与基准方法相比，DAN-MTIF 的准确度为 0.9106、精确度为 0.9396、灵敏度为 0.8470、特异度为 0.9588 和 0.8909 $F_{1}$ 的得分；（2）DAN-MTIF 能有效地从多个时间尺度的心电图片段中提取比单一时间尺度的心电图片段具有更高辨别度的特征，保证了更好的 SA 检测性能；（3）深度学习方法的整体性能优于经典的机器学习算法，凸显了深度学习方法在 SA 检测中的优越性能。
{"title":"Deep Attention Networks With Multi-Temporal Information Fusion for Sleep Apnea Detection","authors":"Meng Jiao;Changyue Song;Xiaochen Xian;Shihao Yang;Feng Liu","doi":"10.1109/OJEMB.2024.3405666","DOIUrl":"10.1109/OJEMB.2024.3405666","url":null,"abstract":"Sleep Apnea (SA) is a prevalent sleep disorder with multifaceted etiologies that can have severe consequences for patients. Diagnosing SA traditionally relies on the in-laboratory polysomnogram (PSG), which records various human physiological activities overnight. SA diagnosis involves manual scoring by qualified physicians. Traditional machine learning methods for SA detection depend on hand-crafted features, making feature selection pivotal for downstream classification tasks. In recent years, deep learning has gained popularity in SA detection due to its capability for automatic feature extraction and superior classification accuracy. This study introduces a Deep Attention Network with Multi-Temporal Information Fusion (DAN-MTIF) for SA detection using single-lead electrocardiogram (ECG) signals. This framework utilizes three 1D convolutional neural network (CNN) blocks to extract features from R-R intervals and R-peak amplitudes using segments of varying lengths. Recognizing that features derived from different temporal scales vary in their contribution to classification, we integrate a multi-head attention module with a self-attention mechanism to learn the weights for each feature vector. Comprehensive experiments and comparisons between two paradigms of classical machine learning approaches and deep learning approaches are conducted. Our experiment results demonstrate that (1) compared with benchmark methods, the proposed DAN-MTIF exhibits excellent performance with 0.9106 accuracy, 0.9396 precision, 0.8470 sensitivity, 0.9588 specificity, and 0.8909 \u0000<inline-formula><tex-math>$F_{1}$</tex-math></inline-formula>\u0000 score at per-segment level; (2) DAN-MTIF can effectively extract features with a higher degree of discrimination from ECG segments of multiple timescales than those with a single time scale, ensuring a better SA detection performance; (3) the overall performance of deep learning methods is better than the classical machine learning algorithms, highlighting the superior performance of deep learning approaches for SA detection.","PeriodicalId":33825,"journal":{"name":"IEEE Open Journal of Engineering in Medicine and Biology","volume":"5 ","pages":"792-802"},"PeriodicalIF":2.7,"publicationDate":"2024-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10539178","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141170429","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Goal: In patients with coronary artery disease, the prediction of future cardiac events such as myocardial infarction (MI) remains a major challenge. In this work, we propose a novel anatomy-informed multimodal deep learning framework to predict future MI from clinical data and Invasive Coronary Angiography (ICA) images. Methods: The images are analyzed by Convolutional Neural Networks (CNNs) guided by anatomical information, and the clinical data by an Artificial Neural Network (ANN). Embeddings from both sources are then merged to provide a patient-level prediction. Results: The results of our framework on a clinical study of 445 patients admitted with acute coronary syndromes confirms that multimodal learning increases the predictive power and achieves good performance (AUC: $0.67pm 0.04$ & F1-Score: $0.36pm 0.12$ ), which outperforms the prediction obtained by each modality independently as well as that of interventional cardiologists (AUC: 0.54 & F1-Score: 0.18). Conclusions: To the best of our knowledge, this is the first attempt towards combining multimodal data through a deep learning framework for future MI prediction. Although it demonstrates the superiority of multi-modal approaches over single modality, the results do not yet meet the necessary criteria for practical application.
{"title":"Anatomy-Informed Multimodal Learning for Myocardial Infarction Prediction","authors":"Ivan-Daniel Sievering;Ortal Senouf;Thabo Mahendiran;David Nanchen;Stephane Fournier;Olivier Muller;Pascal Frossard;Emmanuel Abbé;Dorina Thanou","doi":"10.1109/OJEMB.2024.3403948","DOIUrl":"10.1109/OJEMB.2024.3403948","url":null,"abstract":"<italic>Goal:</i>\u0000 In patients with coronary artery disease, the prediction of future cardiac events such as myocardial infarction (MI) remains a major challenge. In this work, we propose a novel anatomy-informed multimodal deep learning framework to predict future MI from clinical data and Invasive Coronary Angiography (ICA) images. \u0000<italic>Methods:</i>\u0000 The images are analyzed by Convolutional Neural Networks (CNNs) guided by anatomical information, and the clinical data by an Artificial Neural Network (ANN). Embeddings from both sources are then merged to provide a patient-level prediction. \u0000<italic>Results:</i>\u0000 The results of our framework on a clinical study of 445 patients admitted with acute coronary syndromes confirms that multimodal learning increases the predictive power and achieves good performance (AUC: \u0000<inline-formula><tex-math>$0.67pm 0.04$</tex-math></inline-formula>\u0000 & F1-Score: \u0000<inline-formula><tex-math>$0.36pm 0.12$</tex-math></inline-formula>\u0000), which outperforms the prediction obtained by each modality independently as well as that of interventional cardiologists (AUC: 0.54 & F1-Score: 0.18). \u0000<italic>Conclusions:</i>\u0000 To the best of our knowledge, this is the first attempt towards combining multimodal data through a deep learning framework for future MI prediction. Although it demonstrates the superiority of multi-modal approaches over single modality, the results do not yet meet the necessary criteria for practical application.","PeriodicalId":33825,"journal":{"name":"IEEE Open Journal of Engineering in Medicine and Biology","volume":"5 ","pages":"837-845"},"PeriodicalIF":2.7,"publicationDate":"2024-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10540036","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141170475","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-25DOI: 10.1109/OJEMB.2024.3381475
E. Losanno;M. Badi;E. Roussinova;A. Bogaard;M. Delacombaz;S. Shokur;S. Micera
Objective: Brain-body interfaces (BBIs) have emerged as a very promising solution for restoring voluntary hand control in people with upper-limb paralysis. The BBI module decoding motor commands from brain signals should provide the user with intuitive, accurate, and stable control. Here, we present a preliminary investigation in a monkey of a brain decoding strategy based on the direct coupling between the activity of intrinsic neural ensembles and output variables, aiming at achieving ease of learning and long-term robustness. Results: We identified an intrinsic low-dimensional space (called manifold) capturing the co-variation patterns of the monkey's neural activity associated to reach-to-grasp movements. We then tested the animal's ability to directly control a computer cursor using cortical activation along the manifold axes. By daily recalibrating only scaling factors, we achieved rapid learning and stable high performance in simple, incremental 2D tasks over more than 12 weeks of experiments. Finally, we showed that this brain decoding strategy can be effectively coupled to peripheral nerve stimulation to trigger voluntary hand movements. Conclusions: These results represent a proof of concept of manifold-based direct control for BBI applications.
{"title":"An Investigation of Manifold-Based Direct Control for a Brain-to-Body Neural Bypass","authors":"E. Losanno;M. Badi;E. Roussinova;A. Bogaard;M. Delacombaz;S. Shokur;S. Micera","doi":"10.1109/OJEMB.2024.3381475","DOIUrl":"10.1109/OJEMB.2024.3381475","url":null,"abstract":"<italic>Objective:</i>\u0000 Brain-body interfaces (BBIs) have emerged as a very promising solution for restoring voluntary hand control in people with upper-limb paralysis. The BBI module decoding motor commands from brain signals should provide the user with intuitive, accurate, and stable control. Here, we present a preliminary investigation in a monkey of a brain decoding strategy based on the direct coupling between the activity of intrinsic neural ensembles and output variables, aiming at achieving ease of learning and long-term robustness. \u0000<italic>Results:</i>\u0000 We identified an intrinsic low-dimensional space (called manifold) capturing the co-variation patterns of the monkey's neural activity associated to reach-to-grasp movements. We then tested the animal's ability to directly control a computer cursor using cortical activation along the manifold axes. By daily recalibrating only scaling factors, we achieved rapid learning and stable high performance in simple, incremental 2D tasks over more than 12 weeks of experiments. Finally, we showed that this brain decoding strategy can be effectively coupled to peripheral nerve stimulation to trigger voluntary hand movements. \u0000<italic>Conclusions:</i>\u0000 These results represent a proof of concept of manifold-based direct control for BBI applications.","PeriodicalId":33825,"journal":{"name":"IEEE Open Journal of Engineering in Medicine and Biology","volume":"5 ","pages":"271-280"},"PeriodicalIF":5.8,"publicationDate":"2024-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10478790","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140302370","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-25DOI: 10.1109/OJEMB.2024.3380352
Vladimir Atanasoski;Jovana Petrović;Lana Popović Maneski;Marjan Miletić;Miloš Babić;Aleksandra Nikolić;Dorin Panescu;Marija D. Ivanović
Goal: Clinical interpretation of an electrocardiogram (ECG) can be detrimentally affected by noise. Removal of the electromyographic (EMG) noise is particularly challenging due to its spectral overlap with the QRS complex. The existing EMG-denoising algorithms often distort signal morphology, thus obscuring diagnostically relevant information. Methods: Here, a new iterative regeneration method (IRM) for efficient EMG-noise suppression is proposed. The main hypothesis is that the temporary removal of the dominant ECG components enables extraction of the noise with the minimum alteration to the signal. The method is validated on SimEMG database of simultaneously recorded reference and noisy signals, MIT-BIH arrhythmia database and synthesized ECG signals, both with the noise from MIT Noise Stress Test Database. Results: IRM denoising and morphology-preserving performance is superior to the wavelet- and FIR-based benchmark methods. Conclusions : IRM is reliable, computationally non-intensive, fast and applicable to any number of ECG channels recorded by mobile or standard ECG devices.
目标：心电图（ECG）的临床解读可能会受到噪声的不利影响。由于肌电图（EMG）噪声的频谱与 QRS 波群重叠，因此去除肌电图噪声尤其具有挑战性。现有的肌电图去噪算法往往会扭曲信号形态，从而掩盖诊断相关信息。方法：本文提出了一种新的迭代再生方法（IRM），用于有效抑制肌电图噪声。其主要假设是，暂时去除主要的心电图成分可在提取噪声的同时将对信号的改变降到最低。该方法在同时记录参考信号和噪声信号的 SimEMG 数据库、MIT-BIH 心律失常数据库和合成心电信号上进行了验证。结果IRM 去噪和形态保持性能优于基于小波和 FIR 的基准方法。结论：IRMIRM 可靠、计算不密集、速度快，适用于移动或标准心电图设备记录的任何数量的心电图通道。
{"title":"A Morphology-Preserving Algorithm for Denoising of EMG-Contaminated ECG Signals","authors":"Vladimir Atanasoski;Jovana Petrović;Lana Popović Maneski;Marjan Miletić;Miloš Babić;Aleksandra Nikolić;Dorin Panescu;Marija D. Ivanović","doi":"10.1109/OJEMB.2024.3380352","DOIUrl":"10.1109/OJEMB.2024.3380352","url":null,"abstract":"<italic>Goal:</i>\u0000 Clinical interpretation of an electrocardiogram (ECG) can be detrimentally affected by noise. Removal of the electromyographic (EMG) noise is particularly challenging due to its spectral overlap with the QRS complex. The existing EMG-denoising algorithms often distort signal morphology, thus obscuring diagnostically relevant information. \u0000<italic>Methods:</i>\u0000 Here, a new iterative regeneration method (IRM) for efficient EMG-noise suppression is proposed. The main hypothesis is that the temporary removal of the dominant ECG components enables extraction of the noise with the minimum alteration to the signal. The method is validated on SimEMG database of simultaneously recorded reference and noisy signals, MIT-BIH arrhythmia database and synthesized ECG signals, both with the noise from MIT Noise Stress Test Database. \u0000<italic>Results:</i>\u0000 IRM denoising and morphology-preserving performance is superior to the wavelet- and FIR-based benchmark methods. \u0000<italic>Conclusions</i>\u0000: IRM is reliable, computationally non-intensive, fast and applicable to any number of ECG channels recorded by mobile or standard ECG devices.","PeriodicalId":33825,"journal":{"name":"IEEE Open Journal of Engineering in Medicine and Biology","volume":"5 ","pages":"296-305"},"PeriodicalIF":5.8,"publicationDate":"2024-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10479179","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140302635","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-23DOI: 10.1109/OJEMB.2024.3404653
Yu-Dong Zhang
Researchers in biomedical engineering are increasingly turning to weakly-supervised deep learning (WSDL) techniques [1] to tackle challenges in biomedical data analysis, which often involves noisy, limited, or imprecise expert annotations [2]. WSDL methods have emerged as a solution to alleviate the manual annotation burden for structured biomedical data like signals, images, and videos [3] while enabling deep neural network models to learn from larger-scale datasets at a reduced annotation cost. With the proliferation of advanced deep learning techniques such as generative adversarial networks (GANs), graph neural networks (GNNs) [4], vision transformers (ViTs) [5], and deep reinforcement learning (DRL) models [6], research endeavors are focused on solving WSDL problems and applying these techniques to various biomedical analysis tasks.
{"title":"Guest Editorial Introduction to the Special Section on Weakly-Supervised Deep Learning and Its Applications","authors":"Yu-Dong Zhang","doi":"10.1109/OJEMB.2024.3404653","DOIUrl":"10.1109/OJEMB.2024.3404653","url":null,"abstract":"Researchers in biomedical engineering are increasingly turning to weakly-supervised deep learning (WSDL) techniques [1] to tackle challenges in biomedical data analysis, which often involves noisy, limited, or imprecise expert annotations [2]. WSDL methods have emerged as a solution to alleviate the manual annotation burden for structured biomedical data like signals, images, and videos [3] while enabling deep neural network models to learn from larger-scale datasets at a reduced annotation cost. With the proliferation of advanced deep learning techniques such as generative adversarial networks (GANs), graph neural networks (GNNs) [4], vision transformers (ViTs) [5], and deep reinforcement learning (DRL) models [6], research endeavors are focused on solving WSDL problems and applying these techniques to various biomedical analysis tasks.","PeriodicalId":33825,"journal":{"name":"IEEE Open Journal of Engineering in Medicine and Biology","volume":"5 ","pages":"393-395"},"PeriodicalIF":5.8,"publicationDate":"2024-03-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10537991","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141150870","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-20DOI: 10.1109/OJEMB.2024.3402151
Chiara Romano;Emanuele Maiorana;Annunziata Nusca;Simone Circhetta;Sergio Silvestri;Schena Emiliano;Gian Paolo Ussia;Carlo Massaroni
Goal: To evaluate the suitability of seismocardiogram (SCG) and gyrocardiogram (GCG) recorded at the skin level to classify aortic stenosis (AS) patients from healthy volunteers, and to determine the optimal sensor position for the classification. Methods: SCG and GCG were recorded along three axes at five chest locations of fifteen healthy subjects and AS patients. Signal frames underwent feature extraction in frequency and time-frequency domains. Then, binary classification was performed through three machine learning and three deep learning methods, considering SCG, GCG, and their combination. Results: The highest classification accuracies were achieved using Support Vector Machine (SVM) classifier, with the best sensor locations being at the mitral valve for SCG signals (92.3% accuracy) and at the pulmonary valve for GCG (92.1%). Combining SCG and GCG data allows for further improvement in the achievable accuracy (93.5%). Jointly exploiting SCG and GCG signals and both SVM- and ResNet18-based classifiers, 40 s of monitoring allows for reaching 97.2% accuracy with a single sensor on the pulmonary valve. Conclusions: Combining SCG and GCG with adequate machine learning and deep learning classifiers allows reliable classification of AS patients.
{"title":"Classification of Aortic Stenosis Patients via ECG-Independent Multi-Site Measurements of Cardiac-Induced Accelerations and Angular Velocities at the Skin Level","authors":"Chiara Romano;Emanuele Maiorana;Annunziata Nusca;Simone Circhetta;Sergio Silvestri;Schena Emiliano;Gian Paolo Ussia;Carlo Massaroni","doi":"10.1109/OJEMB.2024.3402151","DOIUrl":"10.1109/OJEMB.2024.3402151","url":null,"abstract":"<italic>Goal:</i>\u0000 To evaluate the suitability of seismocardiogram (SCG) and gyrocardiogram (GCG) recorded at the skin level to classify aortic stenosis (AS) patients from healthy volunteers, and to determine the optimal sensor position for the classification. \u0000<italic>Methods:</i>\u0000 SCG and GCG were recorded along three axes at five chest locations of fifteen healthy subjects and AS patients. Signal frames underwent feature extraction in frequency and time-frequency domains. Then, binary classification was performed through three machine learning and three deep learning methods, considering SCG, GCG, and their combination. \u0000<italic>Results:</i>\u0000 The highest classification accuracies were achieved using Support Vector Machine (SVM) classifier, with the best sensor locations being at the mitral valve for SCG signals (92.3% accuracy) and at the pulmonary valve for GCG (92.1%). Combining SCG and GCG data allows for further improvement in the achievable accuracy (93.5%). Jointly exploiting SCG and GCG signals and both SVM- and ResNet18-based classifiers, 40 s of monitoring allows for reaching 97.2% accuracy with a single sensor on the pulmonary valve. \u0000<italic>Conclusions:</i>\u0000 Combining SCG and GCG with adequate machine learning and deep learning classifiers allows reliable classification of AS patients.","PeriodicalId":33825,"journal":{"name":"IEEE Open Journal of Engineering in Medicine and Biology","volume":"5 ","pages":"867-876"},"PeriodicalIF":2.7,"publicationDate":"2024-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10534834","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141150869","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Goal: Recently, large datasets of biosignals acquired during surgery became available. As they offer multiple physiological signals measured in parallel, multimodal analysis – which involves their joint analysis – can be conducted and could provide deeper insights than unimodal analysis based on a single signal. However, it is unclear what percentage of intraoperatively acquired data is suitable for multimodal analysis. Due to the large amount of data, manual inspection and labelling into suitable and unsuitable segments are not feasible. Nevertheless, multimodal analysis is performed successfully in sleep studies since many years as their signals have proven suitable. Hence, this study evaluates the suitability to perform multimodal analysis on a surgery dataset (VitalDB) using a multi-center sleep dataset (SIESTA) as reference. Methods: We applied widely known algorithms entitled “signal quality indicators” to the common biosignals in both datasets, namely electrocardiography, electroencephalography, and respiratory signals split in segments of 10 s duration. As there are no multimodal methods available, we used only unimodal signal quality indicators. In case, all three signals were determined as being adequate by the indicators, we assumed that the whole signal segment was suitable for multimodal analysis. Results: 82% of SIESTA and 72% of VitalDB are suitable for multimodal analysis. Unsuitable signal segments exhibit constant or physiologically unreasonable values. Histogram examination indicated similar signal quality distributions between the datasets, albeit with potential statistical biases due to different measurement setups. Conclusions: The majority of data within VitalDB is suitable for multimodal analysis.
{"title":"Quantifying the Suitability of Biosignals Acquired During Surgery for Multimodal Analysis","authors":"Ennio Idrobo-Ávila;Gergő Bognár;Dagmar Krefting;Thomas Penzel;Péter Kovács;Nicolai Spicher","doi":"10.1109/OJEMB.2024.3379733","DOIUrl":"10.1109/OJEMB.2024.3379733","url":null,"abstract":"<italic>Goal:</i>\u0000 Recently, large datasets of biosignals acquired during surgery became available. As they offer multiple physiological signals measured in parallel, multimodal analysis – which involves their joint analysis – can be conducted and could provide deeper insights than unimodal analysis based on a single signal. However, it is unclear what percentage of intraoperatively acquired data is suitable for multimodal analysis. Due to the large amount of data, manual inspection and labelling into suitable and unsuitable segments are not feasible. Nevertheless, multimodal analysis is performed successfully in sleep studies since many years as their signals have proven suitable. Hence, this study evaluates the suitability to perform multimodal analysis on a surgery dataset (VitalDB) using a multi-center sleep dataset (SIESTA) as reference. \u0000<italic>Methods:</i>\u0000 We applied widely known algorithms entitled “signal quality indicators” to the common biosignals in both datasets, namely electrocardiography, electroencephalography, and respiratory signals split in segments of 10 s duration. As there are no multimodal methods available, we used only unimodal signal quality indicators. In case, all three signals were determined as being adequate by the indicators, we assumed that the whole signal segment was suitable for multimodal analysis. \u0000<italic>Results:</i>\u0000 82% of SIESTA and 72% of VitalDB are suitable for multimodal analysis. Unsuitable signal segments exhibit constant or physiologically unreasonable values. Histogram examination indicated similar signal quality distributions between the datasets, albeit with potential statistical biases due to different measurement setups. \u0000<italic>Conclusions:</i>\u0000 The majority of data within VitalDB is suitable for multimodal analysis.","PeriodicalId":33825,"journal":{"name":"IEEE Open Journal of Engineering in Medicine and Biology","volume":"5 ","pages":"250-260"},"PeriodicalIF":5.8,"publicationDate":"2024-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10476670","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140202163","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-20DOI: 10.1109/OJEMB.2024.3402998
Sabrina Zumbo;Stefano Mandija;Ettore F. Meliadò;Peter Stijnman;Thierry G. Meerbothe;Cornelis A.T. van den Berg;Tommaso Isernia;Martina T. Bevacqua
Magnetic Resonance imaging based Electrical Properties Tomography (MR-EPT) is a non-invasive technique that measures the electrical properties (EPs) of biological tissues. In this work, we present and numerically investigate the performance of an unrolled, physics-assisted method for 2D MR-EPT reconstructions, where a cascade of Convolutional Neural Networks is used to compute the contrast update. Each network takes in input the EPs and the gradient descent direction (encoding the physics underlying the adopted scattering model) and returns as output the updated contrast function. The network is trained and tested in silico using 2D slices of realistic brain models at 128 MHz. Results show the capability of the proposed procedure to reconstruct EPs maps with quality comparable to that of the popular Contrast Source Inversion-EPT, while significantly reducing the computational time.
{"title":"Unrolled Optimization via Physics-Assisted Convolutional Neural Network for MR-Based Electrical Properties Tomography: A Numerical Investigation","authors":"Sabrina Zumbo;Stefano Mandija;Ettore F. Meliadò;Peter Stijnman;Thierry G. Meerbothe;Cornelis A.T. van den Berg;Tommaso Isernia;Martina T. Bevacqua","doi":"10.1109/OJEMB.2024.3402998","DOIUrl":"10.1109/OJEMB.2024.3402998","url":null,"abstract":"Magnetic Resonance imaging based Electrical Properties Tomography (MR-EPT) is a non-invasive technique that measures the electrical properties (EPs) of biological tissues. In this work, we present and numerically investigate the performance of an unrolled, physics-assisted method for 2D MR-EPT reconstructions, where a cascade of Convolutional Neural Networks is used to compute the contrast update. Each network takes in input the EPs and the gradient descent direction (encoding the physics underlying the adopted scattering model) and returns as output the updated contrast function. The network is trained and tested in silico using 2D slices of realistic brain models at 128 MHz. Results show the capability of the proposed procedure to reconstruct EPs maps with quality comparable to that of the popular Contrast Source Inversion-EPT, while significantly reducing the computational time.","PeriodicalId":33825,"journal":{"name":"IEEE Open Journal of Engineering in Medicine and Biology","volume":"5 ","pages":"505-513"},"PeriodicalIF":2.7,"publicationDate":"2024-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10534835","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141150877","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：