Pub Date : 2024-11-18DOI: 10.1016/j.bspc.2024.107156
G. Dharani Devi , Neeraj Kumar , Manikandan J , V. Rekha
The process of detecting brain tumors entails capturing brain images, which are then scrutinized to identify any abnormalities. It is crucial to develop and validate medical image classification models in collaboration with healthcare professionals to ensure their safety and effectiveness in clinical settings. These models aid in categorizing the disease based on its type, facilitating appropriate treatment decisions. While traditional methods for brain tumor detection have been useful, they often face limitations in terms of accuracy, scalability, time sensitivity, and cost. To overcome these complexities, a novel Stacked Random Support Vector-based Hybrid Gazelle Coati (SRS-HGC) algorithm is developed to detect brain tumors. This method utilizes feature extraction to capture the shape and size of the tumor. Additionally, Support Vector Machine. Random Forest, and stacked ensemble techniques are employed to classify medical images, into categories such as tumor or non-tumor. In this research, the Hybrid Gazelle optimization and Coati Optimization algorithms are employed for tuning hyperparameter that enhances the efficiency. The analysis are carried out using the Brain Tumor Segmentation (BraTS2020), Br35H, Figshare Brain tumor and REMBRANDT datasets. The results are then compared by demonstrating the efficiency of the SRS-HGC technique in detecting brain tumor diseases.
检测脑肿瘤的过程需要捕捉脑部图像，然后对图像进行仔细检查，以确定是否存在异常。与医疗保健专业人员合作开发和验证医学图像分类模型以确保其在临床环境中的安全性和有效性至关重要。这些模型有助于根据疾病类型对疾病进行分类，从而做出适当的治疗决定。虽然传统的脑肿瘤检测方法很有用，但它们往往在准确性、可扩展性、时间敏感性和成本方面面临限制。为了克服这些复杂性，我们开发了一种新型的基于堆叠随机支持向量的混合瞪羚算法（SRS-HGC）来检测脑肿瘤。该方法利用特征提取来捕捉肿瘤的形状和大小。此外，支持向量机采用随机森林和堆叠集合技术将医学图像分为肿瘤或非肿瘤等类别。在这项研究中，采用了混合瞪羚优化和科蒂优化算法来调整超参数，以提高效率。分析使用了脑肿瘤分割（BraTS2020）、Br35H、Figshare Brain tumor 和 REMBRANDT 数据集。然后对结果进行比较，以证明 SRS-HGC 技术在检测脑肿瘤疾病方面的效率。
{"title":"Innovative brain tumor detection: Stacked random support vector-based hybrid gazelle coati algorithm","authors":"G. Dharani Devi , Neeraj Kumar , Manikandan J , V. Rekha","doi":"10.1016/j.bspc.2024.107156","DOIUrl":"10.1016/j.bspc.2024.107156","url":null,"abstract":"<div><div>The process of detecting brain tumors entails capturing brain images, which are then scrutinized to identify any abnormalities. It is crucial to develop and validate medical image classification models in collaboration with healthcare professionals to ensure their safety and effectiveness in clinical settings. These models aid in categorizing the disease based on its type, facilitating appropriate treatment decisions. While traditional methods for brain tumor detection have been useful, they often face limitations in terms of accuracy, scalability, time sensitivity, and cost. To overcome these complexities, a novel Stacked Random Support Vector-based Hybrid Gazelle Coati (SRS-HGC) algorithm is developed to detect brain tumors. This method utilizes feature extraction to capture the shape and size of the tumor. Additionally, Support Vector Machine. Random Forest, and stacked ensemble techniques are employed to classify medical images, into categories such as tumor or non-tumor. In this research, the Hybrid Gazelle optimization and Coati Optimization algorithms are employed for tuning hyperparameter that enhances the efficiency. The analysis are carried out using the Brain Tumor Segmentation (BraTS2020), Br35H, Figshare Brain tumor and REMBRANDT datasets. The results are then compared by demonstrating the efficiency of the SRS-HGC technique in detecting brain tumor diseases.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"101 ","pages":"Article 107156"},"PeriodicalIF":4.9,"publicationDate":"2024-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142652857","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 : 2024-11-17DOI: 10.1016/j.bspc.2024.107136
C. Binu Jeya Schafftar , A. Radhakrishnan , C. Emmy Prema
Distinguishing COVID-19 from other acute infectious pneumonia types remains a significant challenge, often complicated by overlapping symptoms and radiological features. This paper introduces a novel texture-based detection framework that enhances the classification accuracy of COVID-19 cases. The proposed method begins with a pre-processing task carried out by the Advanced Bilinear CLAHE (ABC) approach. This will enhance the image contrast and reduce noise revealing critical texture features than traditional methods. Here, one of the innovations is said to be the Texture Rectified Cross-Attention Transformer (TRCAT) for feature extraction. Unlike previous models, TRCAT integrates both handcrafted and deep features by texture enhancement block along with an attention-rectified convolutional block. This combination allows more detailed feature extraction enabling the model to differentiate variations in texture across different cases of pneumonia and COVID-19. Further, Mountain Gazelle Optimized Enhanced Support Vector Machine (MGO-ESVM) optimized by the Enhanced Chaotic Mountain Gazelle Optimization (EMGO) algorithm is introduced for classification. This optimization minimizes processing losses and enhances the accuracy of the classification, particularly in distinguishing COVID-19 from other forms of pneumonia. The method was evaluated on the COVID-19 Detection X-Ray Dataset, and the results are outstanding, with an accuracy rate of 99.34 %. Compared to other methods, namely ensemble CNN, this framework offers significantly improved performance in COVID-19 detection. For instance, in similar studies, accuracy rates have typically ranged between 95 % and 98.82 %. From the analysis, the TRCAT and MGO-ESVM framework consistently outperforms these methods by a margin of 1–2 %. These results underline the novelty and effectiveness of the proposed texture-based approach in enhancing diagnostic precision and generalizability.
{"title":"A novel optimized machine learning approach with texture rectified cross-attention based transformer for COVID-19 detection","authors":"C. Binu Jeya Schafftar , A. Radhakrishnan , C. Emmy Prema","doi":"10.1016/j.bspc.2024.107136","DOIUrl":"10.1016/j.bspc.2024.107136","url":null,"abstract":"<div><div>Distinguishing COVID-19 from other acute infectious pneumonia types remains a significant challenge, often complicated by overlapping symptoms and radiological features. This paper introduces a novel texture-based detection framework that enhances the classification accuracy of COVID-19 cases. The proposed method begins with a pre-processing task carried out by the Advanced Bilinear CLAHE (ABC) approach. This will enhance the image contrast and reduce noise revealing critical texture features than traditional methods. Here, one of the innovations is said to be the Texture Rectified Cross-Attention Transformer (TRCAT) for feature extraction. Unlike previous models, TRCAT integrates both handcrafted and deep features by texture enhancement block along with an attention-rectified convolutional block. This combination allows more detailed feature extraction enabling the model to differentiate variations in texture across different cases of pneumonia and COVID-19. Further, Mountain Gazelle Optimized Enhanced Support Vector Machine (MGO-ESVM) optimized by the Enhanced Chaotic Mountain Gazelle Optimization (EMGO) algorithm is introduced for classification. This optimization minimizes processing losses and enhances the accuracy of the classification, particularly in distinguishing COVID-19 from other forms of pneumonia. The method was evaluated on the COVID-19 Detection X-Ray Dataset, and the results are outstanding, with an accuracy rate of 99.34 %. Compared to other methods, namely ensemble CNN, this framework offers significantly improved performance in COVID-19 detection. For instance, in similar studies, accuracy rates have typically ranged between 95 % and 98.82 %. From the analysis, the TRCAT and MGO-ESVM framework consistently outperforms these methods by a margin of 1–2 %. These results underline the novelty and effectiveness of the proposed texture-based approach in enhancing diagnostic precision and generalizability.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"101 ","pages":"Article 107136"},"PeriodicalIF":4.9,"publicationDate":"2024-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142652853","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 : 2024-11-17DOI: 10.1016/j.bspc.2024.107146
Hongyi Pan , Jingpeng Miao , Jie Yu , Jingran Dong , Mingming Zhang , Xiaobing Wang , Jihong Feng
Retinal diseases such as age-related macular degeneration and diabetic macular edema will lead to irreversible blindness without timely diagnosis and treatment. Optical coherence tomography (OCT) has been widely utilized to detect retinal diseases because of its non-contact and non-invasive imaging peculiarities. Due to the lack of ophthalmic medical resources, automatic analyzing and diagnosing retinal OCT images is necessary with computer-aided diagnosis algorithms. In this study, we propose a lightweight retinal OCT image classification model integrating convolutional neural network (CNN) and Transformer to classify various retinal diseases with few parameters of the model. Local lesion features extracted by CNN can be encoded with the whole OCT image through the Transformer, which improves the classification ability. A convolutional block attention module is also integrated into our model to enhance the representational power. Compared with several classical models, our model achieves the best accuracy of 0.9800 and recall of 0.9799 with the least number of parameters and prediction time for an image on the OCT-C8 dataset. Moreover, on the OCT2017 dataset, our model outperforms the four state-of-the-art models except almost equal to another, achieving an average accuracy, precision, recall, specificity and F1-score of 0.9985, 0.9970, 0.9970, 0.9990, and 0.9970. Simultaneously, the number of parameters of our model has been reduced to just 1.28 M, and the average prediction time for an image is only 2.5 ms.
老年性黄斑变性和糖尿病性黄斑水肿等视网膜疾病如果得不到及时诊断和治疗，将导致不可逆转的失明。光学相干断层扫描（OCT）具有非接触、非侵入性成像的特点，已被广泛用于检测视网膜疾病。由于眼科医疗资源的匮乏，视网膜 OCT 图像的自动分析和诊断需要计算机辅助诊断算法。在这项研究中，我们提出了一种整合了卷积神经网络（CNN）和变换器的轻量级视网膜 OCT 图像分类模型，只需少量模型参数即可对各种视网膜疾病进行分类。CNN 提取的局部病变特征可通过 Transformer 与整个 OCT 图像进行编码，从而提高分类能力。我们的模型还集成了卷积块注意力模块，以增强表征能力。与几种经典模型相比，我们的模型在 OCT-C8 数据集上以最少的参数数量和最少的图像预测时间达到了最佳准确率 0.9800 和召回率 0.9799。此外，在 OCT2017 数据集上，我们的模型除几乎与另一个模型持平外，在准确度、精确度、召回率、特异性和 F1 分数上的平均值分别为 0.9985、0.9970、0.9970、0.9990 和 0.9970，优于四个最先进的模型。同时，我们的模型参数数也减少到了 1.28 M，图像的平均预测时间仅为 2.5 ms。
{"title":"A lightweight model for the retinal disease classification using optical coherence tomography","authors":"Hongyi Pan , Jingpeng Miao , Jie Yu , Jingran Dong , Mingming Zhang , Xiaobing Wang , Jihong Feng","doi":"10.1016/j.bspc.2024.107146","DOIUrl":"10.1016/j.bspc.2024.107146","url":null,"abstract":"<div><div>Retinal diseases such as age-related macular degeneration and diabetic macular edema will lead to irreversible blindness without timely diagnosis and treatment. Optical coherence tomography (OCT) has been widely utilized to detect retinal diseases because of its non-contact and non-invasive imaging peculiarities. Due to the lack of ophthalmic medical resources, automatic analyzing and diagnosing retinal OCT images is necessary with computer-aided diagnosis algorithms. In this study, we propose a lightweight retinal OCT image classification model integrating convolutional neural network (CNN) and Transformer to classify various retinal diseases with few parameters of the model. Local lesion features extracted by CNN can be encoded with the whole OCT image through the Transformer, which improves the classification ability. A convolutional block attention module is also integrated into our model to enhance the representational power. Compared with several classical models, our model achieves the best accuracy of 0.9800 and recall of 0.9799 with the least number of parameters and prediction time for an image on the OCT-C8 dataset. Moreover, on the OCT2017 dataset, our model outperforms the four state-of-the-art models except almost equal to another, achieving an average accuracy, precision, recall, specificity and F1-score of 0.9985, 0.9970, 0.9970, 0.9990, and 0.9970. Simultaneously, the number of parameters of our model has been reduced to just 1.28 M, and the average prediction time for an image is only 2.5 ms.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"101 ","pages":"Article 107146"},"PeriodicalIF":4.9,"publicationDate":"2024-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142652859","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 : 2024-11-16DOI: 10.1016/j.bspc.2024.107134
Siqi Zhao , Xvwen Gui , Jiacheng Zhang , Hao Feng , Bo Yang , Fanli Zhou , Hong Tang , Tao Liu
In recent years, electrocardiogram (ECG) monitoring has become the most effective method of monitoring cardiac rhythm in critically ill patients. It can detect a variety of arrhythmias, including atrial and ventricular premature beats, myocardial perfusion, etc. Nevertheless, the transmission and storage of large amounts of physiological data is a major challenge. To maintain signal integrity and increase transmission speed, data compression is necessary. Current research is increasingly focused on adaptive compression algorithms. These algorithms adapt coding strategies based on signal characteristics. ECG data compression technique combining empirical mode decomposition (EMD) and discrete wavelet transform (DWT) has been proposed. However, the intrinsic mode functions (IMFs) component generated from EMD decomposition suffers from a mode mixing problem. This paper proposes a scheme for decomposing ECG signals using ensemble empirical mode decomposition (EEMD) and recombining the components with DWT. The scheme compresses and quantizes the ECG signal using a uniform scalar dead-zone quantization method and further compresses the data using run-length coding. Evaluation parameters indicate that the proposed scheme has superior compression performance. Compressed signals can facilitate remote transmission and real-time monitoring, providing patients with more convenient medical services and promoting the development of healthcare.
{"title":"An improved ECG data compression scheme based on ensemble empirical mode decomposition","authors":"Siqi Zhao , Xvwen Gui , Jiacheng Zhang , Hao Feng , Bo Yang , Fanli Zhou , Hong Tang , Tao Liu","doi":"10.1016/j.bspc.2024.107134","DOIUrl":"10.1016/j.bspc.2024.107134","url":null,"abstract":"<div><div>In recent years, electrocardiogram (ECG) monitoring has become the most effective method of monitoring cardiac rhythm in critically ill patients. It can detect a variety of arrhythmias, including atrial and ventricular premature beats, myocardial perfusion, etc. Nevertheless, the transmission and storage of large amounts of physiological data is a major challenge. To maintain signal integrity and increase transmission speed, data compression is necessary. Current research is increasingly focused on adaptive compression algorithms. These algorithms adapt coding strategies based on signal characteristics. ECG data compression technique combining empirical mode decomposition (EMD) and discrete wavelet transform (DWT) has been proposed. However, the intrinsic mode functions (IMFs) component generated from EMD decomposition suffers from a mode mixing problem. This paper proposes a scheme for decomposing ECG signals using ensemble empirical mode decomposition (EEMD) and recombining the components with DWT. The scheme compresses and quantizes the ECG signal using a uniform scalar dead-zone quantization method and further compresses the data using run-length coding. Evaluation parameters indicate that the proposed scheme has superior compression performance. Compressed signals can facilitate remote transmission and real-time monitoring, providing patients with more convenient medical services and promoting the development of healthcare.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"101 ","pages":"Article 107134"},"PeriodicalIF":4.9,"publicationDate":"2024-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142652856","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 : 2024-11-16DOI: 10.1016/j.bspc.2024.107143
Mehran Taghipour-Gorjikolaie , Navid Ghavami , Lorenzo Papini , Mario Badia , Arianna Fracassini , Alessandra Bigotti , Gianmarco Palomba , Daniel Álvarez Sánchez-Bayuela , Cristina Romero Castellano , Riccardo Loretoni , Massimo Calabrese , Alberto Stefano Tagliafico , Mohammad Ghavami , Gianluigi Tiberi
Breast cancer is a global health concern, ranking as the second leading cause of death among women. Current screening methods, such as mammography, face limitations, particularly for women under 50 due to radiation concerns and frequency of examination restrictions. MammoWave, utilizing microwave signals (1 to 9 GHz), emerges as an innovative and safe technology for breast cancer detection. This paper focuses on the numerical data extracted from MammoWave, presenting a hierarchical approach to address challenges posed by a diverse dataset of over 1000 samples from two European hospitals. The proposed approach involves unsupervised clustering to classify data into two main groups, followed by binary classification within each group to distinguish healthy and non-healthy cases. Careful consideration is given to feature extraction methods and classifiers at each step. The unique influence of sub-bands within the 1 to 9 GHz range on the diagnosis model is observed, leading to the selection of suitable sub-bands, feature extraction methods, and classification models. An optimization algorithm and a defined cost function are employed to achieve high and balanced sensitivity, specificity, and accuracy values. Experimental results showcase a promising overall balanced performance of around 70 %, representing a significant milestone in breast cancer detection using microwave imaging. MammoWave, with its novel approach, provides a solution that overcomes age and frequency of examination related limitations associated with existing screening methods, contributing to enhanced breast health monitoring for a broader population.
{"title":"AI-based hierarchical approach for optimizing breast cancer detection using MammoWave device","authors":"Mehran Taghipour-Gorjikolaie , Navid Ghavami , Lorenzo Papini , Mario Badia , Arianna Fracassini , Alessandra Bigotti , Gianmarco Palomba , Daniel Álvarez Sánchez-Bayuela , Cristina Romero Castellano , Riccardo Loretoni , Massimo Calabrese , Alberto Stefano Tagliafico , Mohammad Ghavami , Gianluigi Tiberi","doi":"10.1016/j.bspc.2024.107143","DOIUrl":"10.1016/j.bspc.2024.107143","url":null,"abstract":"<div><div>Breast cancer is a global health concern, ranking as the second leading cause of death among women. Current screening methods, such as mammography, face limitations, particularly for women under 50 due to radiation concerns and frequency of examination restrictions. MammoWave, utilizing microwave signals (1 to 9 GHz), emerges as an innovative and safe technology for breast cancer detection. This paper focuses on the numerical data extracted from MammoWave, presenting a hierarchical approach to address challenges posed by a diverse dataset of over 1000 samples from two European hospitals. The proposed approach involves unsupervised clustering to classify data into two main groups, followed by binary classification within each group to distinguish healthy and non-healthy cases. Careful consideration is given to feature extraction methods and classifiers at each step. The unique influence of sub-bands within the 1 to 9 GHz range on the diagnosis model is observed, leading to the selection of suitable sub-bands, feature extraction methods, and classification models. An optimization algorithm and a defined cost function are employed to achieve high and balanced sensitivity, specificity, and accuracy values. Experimental results showcase a promising overall balanced performance of around 70 %, representing a significant milestone in breast cancer detection using microwave imaging. MammoWave, with its novel approach, provides a solution that overcomes age and frequency of examination related limitations associated with existing screening methods, contributing to enhanced breast health monitoring for a broader population.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"100 ","pages":"Article 107143"},"PeriodicalIF":4.9,"publicationDate":"2024-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142658273","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 : 2024-11-16DOI: 10.1016/j.bspc.2024.107138
Md. Sakhawat Hossain Rabbi , Md. Masbahul Bari , Tanoy Debnath , Anichur Rahman , Avik Kumar Das , Md. Parvez Hossain , Ghulam Muhammad
Heart disease is a global health concern with a high mortality rate, necessitating early, accurate, and reliable prediction methods for effective prevention and control. In this research, we combine principal component analysis and linear discriminant analysis to reduce dataset complexity and enhance the performance of heart disease classification models by selecting the most relevant features. We address the class imbalance by employing two balancing techniques: oversampling and the synthetic minority oversampling technique, which ensures a more representative dataset, leading to more accurate predictions. Our study develops a novel ensemble approach, utilizing a combination of random forest, support vector machine, K-nearest neighbors, logistic regression, decision tree, and Gaussian naive Bayes to significantly improve heart disease prediction accuracy. Furthermore, we implement advanced ensemble learning techniques, such as Stacking, Bagging, Voting, and Boosting, to achieve early and precise prediction of heart disease. The performance evaluation is conducted on three datasets: Cleveland Heart Disease, Framingham Heart Disease, and Indicators of Heart Disease Dataset (2020), ensuring a robust validation of our methods. The results demonstrate that the voting ensemble machine learning algorithm (VEMLA) achieved 92% accuracy on the Cleveland Heart Disease dataset, while the bagging ensemble machine learning algorithm (BEMLA) achieved 97% accuracy on both the Framingham Heart Disease and Indicators of Heart Disease (2020) datasets. Notably, the proposed BEMLA consistently outperformed other methods, showcasing its superiority in heart disease prediction. This study contributes a comprehensive and effective approach to heart disease diagnosis, outperforming individual classifiers and providing valuable insights for practical clinical applications.
{"title":"Performance evaluation of optimal ensemble learning approaches with PCA and LDA-based feature extraction for heart disease prediction","authors":"Md. Sakhawat Hossain Rabbi , Md. Masbahul Bari , Tanoy Debnath , Anichur Rahman , Avik Kumar Das , Md. Parvez Hossain , Ghulam Muhammad","doi":"10.1016/j.bspc.2024.107138","DOIUrl":"10.1016/j.bspc.2024.107138","url":null,"abstract":"<div><div>Heart disease is a global health concern with a high mortality rate, necessitating early, accurate, and reliable prediction methods for effective prevention and control. In this research, we combine principal component analysis and linear discriminant analysis to reduce dataset complexity and enhance the performance of heart disease classification models by selecting the most relevant features. We address the class imbalance by employing two balancing techniques: oversampling and the synthetic minority oversampling technique, which ensures a more representative dataset, leading to more accurate predictions. Our study develops a novel ensemble approach, utilizing a combination of random forest, support vector machine, K-nearest neighbors, logistic regression, decision tree, and Gaussian naive Bayes to significantly improve heart disease prediction accuracy. Furthermore, we implement advanced ensemble learning techniques, such as Stacking, Bagging, Voting, and Boosting, to achieve early and precise prediction of heart disease. The performance evaluation is conducted on three datasets: Cleveland Heart Disease, Framingham Heart Disease, and Indicators of Heart Disease Dataset (2020), ensuring a robust validation of our methods. The results demonstrate that the voting ensemble machine learning algorithm (VEMLA) achieved 92% accuracy on the Cleveland Heart Disease dataset, while the bagging ensemble machine learning algorithm (BEMLA) achieved 97% accuracy on both the Framingham Heart Disease and Indicators of Heart Disease (2020) datasets. Notably, the proposed BEMLA consistently outperformed other methods, showcasing its superiority in heart disease prediction. This study contributes a comprehensive and effective approach to heart disease diagnosis, outperforming individual classifiers and providing valuable insights for practical clinical applications.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"101 ","pages":"Article 107138"},"PeriodicalIF":4.9,"publicationDate":"2024-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142652858","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 : 2024-11-15DOI: 10.1016/j.bspc.2024.107188
Irina A. Mizeva , Natalia P. Podolyan , Oleg V. Mamontov , Anastasiia V. Sakovskaia , Alexei A. Kamshilin
Low-frequency oscillations in the human circulatory system is important for basic physiology and practical applications in clinical medicine. Our objective was to study which mechanism (central or local) is responsible for changes in blood flow fluctuations at around 0.1 Hz. We used the method of imaging photoplethysmography synchronized with electrocardiography to measure blood-flow response to local forearm heating of 18 healthy male volunteers. The dynamics of peripheral perfusion was revealed by a correlation processing of photoplethysmography data, and the central hemodynamics was assessed from the electrocardiogram. Wavelet analysis was used to estimate the dynamics of spectral components. Our results show that skin heating leads to multiple increase in local perfusion accompanied by drop in blood flow oscillations at 0.1 Hz, whereas no changes in heart rate variability was observed. After switching off the heating, perfusion remains at the high level, regardless decrease in skin temperature. The 0.1 Hz oscillations are smoothly recovered to the base level. In conclusion, we confirm the local nature of fluctuations in peripheral blood flow in the frequency band of about 0.1 Hz. A significant, but time-delayed, recovery of fluctuation energy in this frequency range after cessation of the skin warming was discovered. This study reveals a novel factor involved in the regulation microcirculatory vascular tone. A comprehensive study of hemodynamics using the new technique of imaging photoplethysmography synchronized with electrocardiography is a prerequisite for development of a valuable diagnostic tool.
{"title":"Study of 0.1-Hz vasomotion in microcirculation under local heating by means of imaging photoplethysmography","authors":"Irina A. Mizeva , Natalia P. Podolyan , Oleg V. Mamontov , Anastasiia V. Sakovskaia , Alexei A. Kamshilin","doi":"10.1016/j.bspc.2024.107188","DOIUrl":"10.1016/j.bspc.2024.107188","url":null,"abstract":"<div><div>Low-frequency oscillations in the human circulatory system is important for basic physiology and practical applications in clinical medicine. Our objective was to study which mechanism (central or local) is responsible for changes in blood flow fluctuations at around 0.1 Hz. We used the method of imaging photoplethysmography synchronized with electrocardiography to measure blood-flow response to local forearm heating of 18 healthy male volunteers. The dynamics of peripheral perfusion was revealed by a correlation processing of photoplethysmography data, and the central hemodynamics was assessed from the electrocardiogram. Wavelet analysis was used to estimate the dynamics of spectral components. Our results show that skin heating leads to multiple increase in local perfusion accompanied by drop in blood flow oscillations at 0.1 Hz, whereas no changes in heart rate variability was observed. After switching off the heating, perfusion remains at the high level, regardless decrease in skin temperature. The 0.1 Hz oscillations are smoothly recovered to the base level. In conclusion, we confirm the local nature of fluctuations in peripheral blood flow in the frequency band of about 0.1 Hz. A significant, but time-delayed, recovery of fluctuation energy in this frequency range after cessation of the skin warming was discovered. This study reveals a novel factor involved in the regulation microcirculatory vascular tone. A comprehensive study of hemodynamics using the new technique of imaging photoplethysmography synchronized with electrocardiography is a prerequisite for development of a valuable diagnostic tool.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"100 ","pages":"Article 107188"},"PeriodicalIF":4.9,"publicationDate":"2024-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142658270","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 : 2024-11-15DOI: 10.1016/j.bspc.2024.107148
Jinjie Guo , Tao Feng , Penghu Wei , Jinguo Huang , Yanfeng Yang , Yiping Wang , Gongpeng Cao , Yuda Huang , Guixia Kang , Guoguang Zhao
Accurate localization of seizure onset zones (SOZs) in patients with drug-resistant epilepsy is essential for improving prognostic outcomes. This process can be significantly enhanced through effective network representation and analysis of functional dependencies among brain regions. However, traditional network construction methods often lack generalizability due to individual variability. Furthermore, the independent design of the network construction and analysis modules restricts the overall optimization of localization frameworks. In this study, we propose a novel deep learning framework that integrates graph building and analysis modules for seizure localization. The graph building block adaptively generates customized network representations from Stereo-Electroencephalography (SEEG) data of individual patients by extracting feature vectors of each channel and calculating functional connectivity weights among channels with these vectors. While the GCN-and-LSTM-based graph analysis block identifies abnormal nodes corresponding to SOZs by aggregating spatial and temporal information in the network representations. The graph analysis block is trained alongside the graph building block via the seizure prediction task. Attention weights assigned to each channel are utilized to characterize epileptogenicity, facilitating precise localization of the SOZ. Our method demonstrates superior performance, surpassing baseline and state-of-the-art approaches in 9 of 13 patients from a public dataset and 11 of 14 patients from a clinical dataset. Visualization of the identified brain regions aligns well with labeled SOZs. Furthermore, the adaptive functional brain network reveals that the connectivity density among SOZ channels is greater than that of other brain regions, corroborating existing clinical findings and further confirming the model’s reliability and interpretability.
{"title":"Adaptive graph learning with SEEG data for improved seizure localization: Considerations of generalization and simplicity","authors":"Jinjie Guo , Tao Feng , Penghu Wei , Jinguo Huang , Yanfeng Yang , Yiping Wang , Gongpeng Cao , Yuda Huang , Guixia Kang , Guoguang Zhao","doi":"10.1016/j.bspc.2024.107148","DOIUrl":"10.1016/j.bspc.2024.107148","url":null,"abstract":"<div><div>Accurate localization of seizure onset zones (SOZs) in patients with drug-resistant epilepsy is essential for improving prognostic outcomes. This process can be significantly enhanced through effective network representation and analysis of functional dependencies among brain regions. However, traditional network construction methods often lack generalizability due to individual variability. Furthermore, the independent design of the network construction and analysis modules restricts the overall optimization of localization frameworks. In this study, we propose a novel deep learning framework that integrates graph building and analysis modules for seizure localization. The graph building block adaptively generates customized network representations from Stereo-Electroencephalography (SEEG) data of individual patients by extracting feature vectors of each channel and calculating functional connectivity weights among channels with these vectors. While the GCN-and-LSTM-based graph analysis block identifies abnormal nodes corresponding to SOZs by aggregating spatial and temporal information in the network representations. The graph analysis block is trained alongside the graph building block via the seizure prediction task. Attention weights assigned to each channel are utilized to characterize epileptogenicity, facilitating precise localization of the SOZ. Our method demonstrates superior performance, surpassing baseline and state-of-the-art approaches in 9 of 13 patients from a public dataset and 11 of 14 patients from a clinical dataset. Visualization of the identified brain regions aligns well with labeled SOZs. Furthermore, the adaptive functional brain network reveals that the connectivity density among SOZ channels is greater than that of other brain regions, corroborating existing clinical findings and further confirming the model’s reliability and interpretability.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"101 ","pages":"Article 107148"},"PeriodicalIF":4.9,"publicationDate":"2024-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142652855","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 : 2024-11-15DOI: 10.1016/j.bspc.2024.107169
Jing Wu , Zhenyi Ge , Helin Huang , Hairui Wang , Nan Li , Chunqiang Hu , Cuizhen Pan , Xiaomei Wu
For automated analysis of mitral valve regurgitation in non-Doppler-based 2D echocardiography, there is limited work on combining quantitative tasks for cardiac targets, such as semantic segmentation and motion tracking, with qualitative detection and etiological classification of mitral regurgitation, thus overlooking the common features that exist among these tasks. Therefore, we proposed a multi-task learning model called Attention-guided ResUNet-MTL (abbreviated as ARUNet-MTL), to address the task from a holistic view. Specifically, the model was built on the U-shaped architecture and emphasized the importance of inherent correlation among tasks by allowing all three tasks to share the encoder structure. Meanwhile, an attention mechanism called MSA was incorporated to improve the temporal continuity of segmentation image sequences by leveraging the bidirectional deformation field information achieved in motion tracking tasks. Besides, during the training phase of the model, the loss function was designed to assimilate two key aspects: the fidelity of cardiac anatomical structures in segmentation and motion tracking results, and the enforcement smooth transitions and coherence between consecutive frames in the sequence. Through 5-fold cross validation, the accuracy for mitral regurgitation etiological classification was 0.8946 and 0.9179 at the video level and subject level, respectively, while the macro-F1 score was 0.8976 and 0.9176, respectively. The segmentation results for the left atrium, left ventricle, and mitral valve yielded Dice coefficients of 0.9438, 0.9157, and 0.7951. Additionally, validation experiments were performed on two public datasets to verify the robustness of the model’s segmentation and motion tracking branches.
{"title":"Attention-guided model for mitral regurgitation analysis based on multi-task learning","authors":"Jing Wu , Zhenyi Ge , Helin Huang , Hairui Wang , Nan Li , Chunqiang Hu , Cuizhen Pan , Xiaomei Wu","doi":"10.1016/j.bspc.2024.107169","DOIUrl":"10.1016/j.bspc.2024.107169","url":null,"abstract":"<div><div>For automated analysis of mitral valve regurgitation in non-Doppler-based 2D echocardiography, there is limited work on combining quantitative tasks for cardiac targets, such as semantic segmentation and motion tracking, with qualitative detection and etiological classification of mitral regurgitation, thus overlooking the common features that exist among these tasks. Therefore, we proposed a multi-task learning model called Attention-guided ResUNet-MTL (abbreviated as ARUNet-MTL), to address the task from a holistic view. Specifically, the model was built on the U-shaped architecture and emphasized the importance of inherent correlation among tasks by allowing all three tasks to share the encoder structure. Meanwhile, an attention mechanism called MSA was incorporated to improve the temporal continuity of segmentation image sequences by leveraging the bidirectional deformation field information achieved in motion tracking tasks. Besides, during the training phase of the model, the loss function was designed to assimilate two key aspects: the fidelity of cardiac anatomical structures in segmentation and motion tracking results, and the enforcement smooth transitions and coherence between consecutive frames in the sequence. Through 5-fold cross validation, the accuracy for mitral regurgitation etiological classification was 0.8946 and 0.9179 at the video level and subject level, respectively, while the macro-F1 score was 0.8976 and 0.9176, respectively. The segmentation results for the left atrium, left ventricle, and mitral valve yielded Dice coefficients of 0.9438, 0.9157, and 0.7951. Additionally, validation experiments were performed on two public datasets to verify the robustness of the model’s segmentation and motion tracking branches.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"101 ","pages":"Article 107169"},"PeriodicalIF":4.9,"publicationDate":"2024-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142652852","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}
Thoracic imaging is vital for diagnosing lung diseases, as it provides a detailed visualization of the lungs. Despite significant advancements in medical imaging techniques, These methods pose critical challenges, such as high costs and the use of radiation in certain devices, which can raise serious concerns, limit accessibility, and increase potential health risks. Therefore main aim of this study addressing these issues by utilizing electrical impedance tomography (EIT), which is a non-invasive imaging technique that mitigates the risks of radiation exposure, reduces costs, and simplifies the interpretation of complex lung disease related patterns seen in traditional imaging methods. As EIT emerges as a promising imaging technique, this study investigates and develops a deep learning-based framework for classifying lung diseases using reconstructed EIT images. The proposed framework includes three feature extraction methods: Initial-Pretrained Weights Models (ResNet-50 and DenseNet-201), fine-tuned convolutional 3D networks, and fine-tuned convolutional 3D accompanied by dense layer networks. Various machine models fed by extracted features were employed for lung sound disease classification both as individual learners and ensemble classifiers. The framework was evaluated on three classification tasks: binary classification (healthy vs. non-healthy) achieving 89.55% accuracy, 3-class classification (obstructive-related, restrictive-related, and healthy) achieving 55.29% accuracy, and 5-class classification (asthma, chronic obstructive pulmonary disease, interstitial lung disease, pulmonary infection, and healthy) achieving 44.54% accuracy. The proposed methods outperform state-of-the-art results and introduce novel approaches to EIT imaging classification.
胸部成像可提供肺部的详细图像，对诊断肺部疾病至关重要。尽管医学成像技术取得了长足的进步，但这些方法也带来了严峻的挑战，如高昂的成本和某些设备中辐射的使用，这些都会引起严重的担忧，限制了可及性，并增加了潜在的健康风险。因此，本研究的主要目的是利用电阻抗断层扫描（EIT）来解决这些问题，EIT 是一种非侵入性成像技术，可降低辐射风险、降低成本，并简化对传统成像方法中出现的复杂肺部疾病相关模式的解释。随着 EIT 成为一种前景广阔的成像技术，本研究调查并开发了一种基于深度学习的框架，用于利用重建的 EIT 图像对肺部疾病进行分类。所提出的框架包括三种特征提取方法：初始预训练加权模型（ResNet-50 和 DenseNet-201）、微调卷积三维网络和微调卷积三维伴密集层网络。在肺部疾病分类中，采用了以提取的特征为反馈的各种机器模型，既可以作为单个学习器，也可以作为集合分类器。该框架在三个分类任务中进行了评估：二元分类（健康与非健康）准确率达到 89.55%，三类分类（阻塞性相关、限制性相关和健康）准确率达到 55.29%，五类分类（哮喘、慢性阻塞性肺病、间质性肺病、肺部感染和健康）准确率达到 44.54%。所提出的方法优于最先进的结果，并为 EIT 成像分类引入了新方法。
{"title":"Deep learning-driven feature engineering for lung disease classification through electrical impedance tomography imaging","authors":"Berke Cansiz, Coskuvar Utkan Kilinc, Gorkem Serbes","doi":"10.1016/j.bspc.2024.107124","DOIUrl":"10.1016/j.bspc.2024.107124","url":null,"abstract":"<div><div>Thoracic imaging is vital for diagnosing lung diseases, as it provides a detailed visualization of the lungs. Despite significant advancements in medical imaging techniques, These methods pose critical challenges, such as high costs and the use of radiation in certain devices, which can raise serious concerns, limit accessibility, and increase potential health risks. Therefore main aim of this study addressing these issues by utilizing electrical impedance tomography (EIT), which is a non-invasive imaging technique that mitigates the risks of radiation exposure, reduces costs, and simplifies the interpretation of complex lung disease related patterns seen in traditional imaging methods. As EIT emerges as a promising imaging technique, this study investigates and develops a deep learning-based framework for classifying lung diseases using reconstructed EIT images. The proposed framework includes three feature extraction methods: Initial-Pretrained Weights Models (ResNet-50 and DenseNet-201), fine-tuned convolutional 3D networks, and fine-tuned convolutional 3D accompanied by dense layer networks. Various machine models fed by extracted features were employed for lung sound disease classification both as individual learners and ensemble classifiers. The framework was evaluated on three classification tasks: binary classification (healthy vs. non-healthy) achieving 89.55% accuracy, 3-class classification (obstructive-related, restrictive-related, and healthy) achieving 55.29% accuracy, and 5-class classification (asthma, chronic obstructive pulmonary disease, interstitial lung disease, pulmonary infection, and healthy) achieving 44.54% accuracy. The proposed methods outperform state-of-the-art results and introduce novel approaches to EIT imaging classification.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"100 ","pages":"Article 107124"},"PeriodicalIF":4.9,"publicationDate":"2024-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142658274","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}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号