Biomedical Signal Processing and Control最新文献_第3页

Automatic segmentation and measurement system of 3D point cloud images based on RGB-D camera for rat wounds

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-14 DOI: 10.1016/j.bspc.2025.107682

Tianci Hu , Chenghua Song , Jian Zhuang , Yi Lyu

Background and Objective

Accurate wound segmentation is an indispensable prerequisite for automated wound size measurement and healing process monitoring. Traditional assessment methods rely on time-consuming manual analysis, which are inefficient and suceptible to subjective judgment.

Methods

To simulate the changes of human wound healing, in this study, 3D point cloud data of 927 rat wounds at different healing stages were acquired, and a 3D point cloud segmentation model based on the improved PointNet++ was proposed to segment the wound area and get its 3D shape. Twenty-eight groups of simulated wounds were constructed, and reliable wound volumes were obtained by calculating the convex hulls of the simulated wound point clouds and regressing the convex hull volume with the actual wound volume.

Result

The improved model achieves 91.4 % in the intersection and mean intersection over union (mIoU) for wound segmentation, which is 1.57 % and 1.18 % higher than that of PointNet and the original PointNet++ model. Further, the volume of the convex hull was used to perform a regression analysis with the real volume of the simulated wound, and then the wound volume of the rats was calculated, in which the Pearson’s correlation coefficient was 0.996 and the R-square was 0.993, which indicated that there was a significant linear relationship between the two and proved that the wound volume measurements possessed a high degree of reliability.

Conclusion

This method acquires 3D wound morphology post-segmentation and provides accurate volume measurements, enhancing wound treatment monitoring and advancing 3D point cloud use in clinical settings.

{"title":"Automatic segmentation and measurement system of 3D point cloud images based on RGB-D camera for rat wounds","authors":"Tianci Hu , Chenghua Song , Jian Zhuang , Yi Lyu","doi":"10.1016/j.bspc.2025.107682","DOIUrl":"10.1016/j.bspc.2025.107682","url":null,"abstract":"<div><h3>Background and Objective</h3><div>Accurate wound segmentation is an indispensable prerequisite for automated wound size measurement and healing process monitoring. Traditional assessment methods rely on time-consuming manual analysis, which are inefficient and suceptible to subjective judgment.</div></div><div><h3>Methods</h3><div>To simulate the changes of human wound healing, in this study, 3D point cloud data of 927 rat wounds at different healing stages were acquired, and a 3D point cloud segmentation model based on the improved PointNet++ was proposed to segment the wound area and get its 3D shape. Twenty-eight groups of simulated wounds were constructed, and reliable wound volumes were obtained by calculating the convex hulls of the simulated wound point clouds and regressing the convex hull volume with the actual wound volume.</div></div><div><h3>Result</h3><div>The improved model achieves 91.4 % in the intersection and mean intersection over union (mIoU) for wound segmentation, which is 1.57 % and 1.18 % higher than that of PointNet and the original PointNet++ model. Further, the volume of the convex hull was used to perform a regression analysis with the real volume of the simulated wound, and then the wound volume of the rats was calculated, in which the Pearson’s correlation coefficient was 0.996 and the R-square was 0.993, which indicated that there was a significant linear relationship between the two and proved that the wound volume measurements possessed a high degree of reliability.</div></div><div><h3>Conclusion</h3><div>This method acquires 3D wound morphology post-segmentation and provides accurate volume measurements, enhancing wound treatment monitoring and advancing 3D point cloud use in clinical settings.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"104 ","pages":"Article 107682"},"PeriodicalIF":4.9,"publicationDate":"2025-02-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143402035","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}

引用次数: 0

PulseEmoNet: Pulse emotion network for speech emotion recognition

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-14 DOI: 10.1016/j.bspc.2025.107687

Huiyun Zhang , Gaigai Tang , Heming Huang , Zhu Yuan , Zongjin Li

In recent years, Speech Emotion Recognition (SER) has garnered significant attention due to its potential applications in human–computer interaction, healthcare, and affective computing. However, existing approaches often face challenges in handling the complex, multimodal nature of speech data and the variability in emotional expressions across different contexts. In this paper, we propose PulseEmoNet, a novel deep learning-based framework designed to enhance the robustness of SER systems by integrating pulse signal information with acoustic features. The key innovation of our approach lies in the development of a PulseEmoNet that effectively captures the temporal and physiological correlates of emotional states from speech signals. Experimental results on multiple benchmark datasets demonstrate the superiority of PulseEmoNet over existing models. On EMODB, SAVEE, and CASIA, PulseEmoNet achieved accuracies of 91.11 %, 78.75 %, and 93.08 %, respectively, outperforming previous methods like 3DRNN + Attention and GM-TCN. Additionally, it achieved 88.70 % on BodEMODB, 61.40 % on IEMOCAP, and 95.98 % on ESD. These results highlight the effectiveness of PulseEmoNet in diverse emotional recognition tasks, providing a promising solution for real-time, cross-domain SER applications.

{"title":"PulseEmoNet: Pulse emotion network for speech emotion recognition","authors":"Huiyun Zhang , Gaigai Tang , Heming Huang , Zhu Yuan , Zongjin Li","doi":"10.1016/j.bspc.2025.107687","DOIUrl":"10.1016/j.bspc.2025.107687","url":null,"abstract":"<div><div>In recent years, Speech Emotion Recognition (SER) has garnered significant attention due to its potential applications in human–computer interaction, healthcare, and affective computing. However, existing approaches often face challenges in handling the complex, multimodal nature of speech data and the variability in emotional expressions across different contexts. In this paper, we propose PulseEmoNet, a novel deep learning-based framework designed to enhance the robustness of SER systems by integrating pulse signal information with acoustic features. The key innovation of our approach lies in the development of a PulseEmoNet that effectively captures the temporal and physiological correlates of emotional states from speech signals. Experimental results on multiple benchmark datasets demonstrate the superiority of PulseEmoNet over existing models. On EMODB, SAVEE, and CASIA, PulseEmoNet achieved accuracies of 91.11 %, 78.75 %, and 93.08 %, respectively, outperforming previous methods like 3DRNN + Attention and GM-TCN. Additionally, it achieved 88.70 % on BodEMODB, 61.40 % on IEMOCAP, and 95.98 % on ESD. These results highlight the effectiveness of PulseEmoNet in diverse emotional recognition tasks, providing a promising solution for real-time, cross-domain SER applications.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107687"},"PeriodicalIF":4.9,"publicationDate":"2025-02-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143403003","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}

引用次数: 0

Skin whole slide image segmentation using lightweight-pruned transformer

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-14 DOI: 10.1016/j.bspc.2025.107624

Anum Abdul Salam , Muhammad Zeeshan Asaf , Muhammad Usman Akram , Ammad Ali , Mashaal Ibne Mashallah , Babar Rao , Samavia Khan , Bassem Rafiq , Bianca Sanabria , Muhammad Haroon Yousaf

The American Institute of Dermatology states that among every four individuals, one suffers from skin disease, adding a burden of up to 75 billion dollars in medical health care. Moreover, skin diseases contribute towards the psychological health of the patient making it even more difficult to cater for the disease. Traditional skin diagnosis pipeline initiates with sample extraction using biopsy, followed by chemical staining to enhance disease-associated structures. The stained sample is then further analyzed by the pathologist for diagnosis. To augment the detection of skin whole slide layers (epidermis, dermis, dermo-epidermal junction, keratin/stratum corneum, and slide background), we present a pruned SegFormer architecture (Derma-Pruned). Utilizing self-sufficient attention matrices, Gaussian positional embedding, and adaptive pruning has helped the model learn relevant features and has reduced redundant feature representations. An accuracy of 94.4% has been observed by the updated architecture when trained and tested on 32 whole slide images acquired, stained, and annotated for five layers by pathologists. We have also compared the performance of baseline models trained on unstained, virtually stained, and chemically stained whole slide images. Models trained on stained images performed significantly better than those trained on unstained images, moreover, a high cross-correlation score has been observed on images segmented from models trained using virtually stained and chemically stained images, emphasizing the accuracy of using virtually stained images in the skin disease diagnostics.

{"title":"Skin whole slide image segmentation using lightweight-pruned transformer","authors":"Anum Abdul Salam , Muhammad Zeeshan Asaf , Muhammad Usman Akram , Ammad Ali , Mashaal Ibne Mashallah , Babar Rao , Samavia Khan , Bassem Rafiq , Bianca Sanabria , Muhammad Haroon Yousaf","doi":"10.1016/j.bspc.2025.107624","DOIUrl":"10.1016/j.bspc.2025.107624","url":null,"abstract":"<div><div>The American Institute of Dermatology states that among every four individuals, one suffers from skin disease, adding a burden of up to 75 billion dollars in medical health care. Moreover, skin diseases contribute towards the psychological health of the patient making it even more difficult to cater for the disease. Traditional skin diagnosis pipeline initiates with sample extraction using biopsy, followed by chemical staining to enhance disease-associated structures. The stained sample is then further analyzed by the pathologist for diagnosis. To augment the detection of skin whole slide layers (epidermis, dermis, dermo-epidermal junction, keratin/stratum corneum, and slide background), we present a pruned SegFormer architecture (Derma-Pruned). Utilizing self-sufficient attention matrices, Gaussian positional embedding, and adaptive pruning has helped the model learn relevant features and has reduced redundant feature representations. An accuracy of 94.4% has been observed by the updated architecture when trained and tested on 32 whole slide images acquired, stained, and annotated for five layers by pathologists. We have also compared the performance of baseline models trained on unstained, virtually stained, and chemically stained whole slide images. Models trained on stained images performed significantly better than those trained on unstained images, moreover, a high cross-correlation score has been observed on images segmented from models trained using virtually stained and chemically stained images, emphasizing the accuracy of using virtually stained images in the skin disease diagnostics.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"106 ","pages":"Article 107624"},"PeriodicalIF":4.9,"publicationDate":"2025-02-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143403538","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}

引用次数: 0

MyoGPT: Augmenting and segmenting spatial muscle activation patterns in forearm using generative Pre-Trained Transformers

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-13 DOI: 10.1016/j.bspc.2025.107591

Wei Chen, Lihui Feng, Jihua Lu

Finger-muscle modeling in sEMG-based gesture interaction requires not only balancing the constraints of resolution, deployment, and cost on the hardware side, but also addressing the constraints of cross-population compatibility and complexity of calibration on the software side. We propose MyoGPT containing GPAT and AT based on generative pre-trained transformers. The GPAT is capable of spatially augmenting the signal on a sparse sEMG device that is less costly and easy to wear, i.e., enhancing from a 1-dimensional vector to a 2-dimensional sEMG pattern, which provides richer information for functional partitioning of muscles. With the spatially augmented data, the AT then segments the relevant muscle regions for driving finger movements with only two gestures to fulfill the calibration. Results show that the SSIM between the augmented sEMG pattern generated by GPAT and the ground truth in the public dataset reaches 76.28 %, and the SSIM for the segmentation of AT is 74.87 %. In addition, the model trained on the public dataset also achieved a SSIM of 68.16 % on our self-developed 16-channel sEMG armband (subjects not in the dataset), and the actual running time is 8.369 ms, which meets the real-time requirement.

{"title":"MyoGPT: Augmenting and segmenting spatial muscle activation patterns in forearm using generative Pre-Trained Transformers","authors":"Wei Chen, Lihui Feng, Jihua Lu","doi":"10.1016/j.bspc.2025.107591","DOIUrl":"10.1016/j.bspc.2025.107591","url":null,"abstract":"<div><div>Finger-muscle modeling in sEMG-based gesture interaction requires not only balancing the constraints of resolution, deployment, and cost on the hardware side, but also addressing the constraints of cross-population compatibility and complexity of calibration on the software side. We propose MyoGPT containing GPAT and AT based on generative pre-trained transformers. The GPAT is capable of spatially augmenting the signal on a sparse sEMG device that is less costly and easy to wear, i.e., enhancing from a 1-dimensional vector to a 2-dimensional sEMG pattern, which provides richer information for functional partitioning of muscles. With the spatially augmented data, the AT then segments the relevant muscle regions for driving finger movements with only two gestures to fulfill the calibration. Results show that the SSIM between the augmented sEMG pattern generated by GPAT and the ground truth in the public dataset reaches 76.28 %, and the SSIM for the segmentation of AT is 74.87 %. In addition, the model trained on the public dataset also achieved a SSIM of 68.16 % on our self-developed 16-channel sEMG armband (subjects not in the dataset), and the actual running time is 8.369 ms, which meets the real-time requirement.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107591"},"PeriodicalIF":4.9,"publicationDate":"2025-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143394732","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}

引用次数: 0

Study on tower crane drivers’ fatigue detection based on conditional empirical mode decomposition and multi-scale attention convolutional neural network

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-13 DOI: 10.1016/j.bspc.2025.107662

Daping Chen, Fuwang Wang

Tower crane drivers’ fatigue may cause safety hazards and serious work accidents. Therefore, the detection of fatigue is critically important. In the research field of driving fatigue of tower crane drivers, the detection method of electroencephalogram (EEG) signals based on drivers is one of the most commonly used methods. However, noise in the real building environment often disrupts this type of detection method, leading to low classification accuracy. To solve this problem, this study proposes a driving fatigue detection model based on conditional empirical mode decomposition and multi-scale attention convolutional neural network (CEMD-MACNN). Conditional empirical mode decomposition (CEMD) overcomes the problem that traditional empirical mode decomposition (EMD) ignores important information or does not sufficiently remove the noise component when analyzing the signal. A multi-scale attention convolutional neural network (MACNN) uses channel attention to adaptively select channels containing fatigue features when extracting features at different scales, thus improving the model’s noise immunity and suppressing the influence of noise. In this study, the driving fatigue detection experiment of tower crane drivers was carried out. The Emotiv device was used to collect the EEG signal of 10 subjects in 7 driving stages, and the EEG signals were divided into awake state and fatigue state using the Karolinska sleepiness scale (KSS). The results showed that the CEMD-MACNN methods achieved an average classification accuracy of 98.70% across 10 subjects. Compared with other traditional methods, CEMD-MACNN has better anti-noise performance and higher classification accuracy.

{"title":"Study on tower crane drivers’ fatigue detection based on conditional empirical mode decomposition and multi-scale attention convolutional neural network","authors":"Daping Chen, Fuwang Wang","doi":"10.1016/j.bspc.2025.107662","DOIUrl":"10.1016/j.bspc.2025.107662","url":null,"abstract":"<div><div>Tower crane drivers’ fatigue may cause safety hazards and serious work accidents. Therefore, the detection of fatigue is critically important. In the research field of driving fatigue of tower crane drivers, the detection method of electroencephalogram (EEG) signals based on drivers is one of the most commonly used methods. However, noise in the real building environment often disrupts this type of detection method, leading to low classification accuracy. To solve this problem, this study proposes a driving fatigue detection model based on conditional empirical mode decomposition and multi-scale attention convolutional neural network (CEMD-MACNN). Conditional empirical mode decomposition (CEMD) overcomes the problem that traditional empirical mode decomposition (EMD) ignores important information or does not sufficiently remove the noise component when analyzing the signal. A multi-scale attention convolutional neural network (MACNN) uses channel attention to adaptively select channels containing fatigue features when extracting features at different scales, thus improving the model’s noise immunity and suppressing the influence of noise. In this study, the driving fatigue detection experiment of tower crane drivers was carried out. The Emotiv device was used to collect the EEG signal of 10 subjects in 7 driving stages, and the EEG signals were divided into awake state and fatigue state using the Karolinska sleepiness scale (KSS). The results showed that the CEMD-MACNN methods achieved an average classification accuracy of 98.70% across 10 subjects. Compared with other traditional methods, CEMD-MACNN has better anti-noise performance and higher classification accuracy.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107662"},"PeriodicalIF":4.9,"publicationDate":"2025-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143394733","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}

引用次数: 0

Preoperative path planning of craniotomy surgical robot based on improved MDP-LQR-RRT* algorithm

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-13 DOI: 10.1016/j.bspc.2025.107647

Zhenzhong Liu, Mingyang Li, Runfeng Zhang, Guobin Zhang, Shilei Han, Kelong Chen

A craniotomy is significant in treating brain diseases and has gained immense attention. The study of preoperative path planning has always been a major research issue in studying craniotomy surgical robots (CSR). Reasonable path planning can effectively prevent various brain tissue injuries and subsequent damage to the brain. Thus, to promote the efficiency of preoperative path planning and improve the safety of surgery, this study proposes a novel preoperative path planning algorithm based on improved MDP-LQR-RRT*. First, the craniotomy approach was analyzed based on expert experience, and the architecture of CSR was introduced. Subsequently, a technique for dividing the bone window area was developed to determine the shape and location of the intracranial tumor before path planning. Then, we proposed the workflow of the MDP-LQR-RRT*, which introduces a workflow that uses the LQR controller to generate path points and utilizes the Markov decision process model to refine the path. Afterward, the effectiveness and accuracy of the proposed method were verified by comparing it with other benchmark methods under two-dimensional and three-dimensional scenes. Finally, experimental verification of the skull model was carried out. The results showed that the method could realize a balanced performance compared with other methods, which provides a foundation for the clinical application of surgery while significantly improving the safety of the procedure.

{"title":"Preoperative path planning of craniotomy surgical robot based on improved MDP-LQR-RRT* algorithm","authors":"Zhenzhong Liu, Mingyang Li, Runfeng Zhang, Guobin Zhang, Shilei Han, Kelong Chen","doi":"10.1016/j.bspc.2025.107647","DOIUrl":"10.1016/j.bspc.2025.107647","url":null,"abstract":"<div><div>A craniotomy is significant in treating brain diseases and has gained immense attention. The study of preoperative path planning has always been a major research issue in studying craniotomy surgical robots (CSR). Reasonable path planning can effectively prevent various brain tissue injuries and subsequent damage to the brain. Thus, to promote the efficiency of preoperative path planning and improve the safety of surgery, this study proposes a novel preoperative path planning algorithm based on improved MDP-LQR-RRT*. First, the craniotomy approach was analyzed based on expert experience, and the architecture of CSR was introduced. Subsequently, a technique for dividing the bone window area was developed to determine the shape and location of the intracranial tumor before path planning. Then, we proposed the workflow of the MDP-LQR-RRT*, which introduces a workflow that uses the LQR controller to generate path points and utilizes the Markov decision process model to refine the path. Afterward, the effectiveness and accuracy of the proposed method were verified by comparing it with other benchmark methods under two-dimensional and three-dimensional scenes. Finally, experimental verification of the skull model was carried out. The results showed that the method could realize a balanced performance compared with other methods, which provides a foundation for the clinical application of surgery while significantly improving the safety of the procedure.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107647"},"PeriodicalIF":4.9,"publicationDate":"2025-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143403002","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}

引用次数: 0

Spatiotemporal feature disentanglement for quality surveillance of left ventricular echocardiographic video using ST-R(2 + 1)D-ConvNeXt

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-13 DOI: 10.1016/j.bspc.2025.107671

Chin-Chieh Hsu , You-Wei Wang , Lung-Chun Lin , Ruey-Feng Chang

The left ventricle (LV), as the primary chamber responsible for systemic circulation, plays a crucial role in cardiac function assessment. Echocardiography which particularly focuses on LV, is vital for cardiac disease diagnosis. However, the diagnostic accuracy heavily depends on image quality, which requires systematic assessment. In this study, we propose a two-stage deep learning approach for echocardiographic quality surveillance using a dataset of 514 annotated videos. The first stage employs EchoNet, to extract LV volumes of interest. The second stage introduces ST-R(2 + 1)D-ConvNeXt, a novel ConvNeXt-based model designed to disentangle spatiotemporal features and leverage echocardiographic hallmarks within the apical-four-chamber (A4C) dynamic echocardiogram data. The proposed approach achieves an accuracy of 82.63 %, an Area Under the Curve (AUC) of 0.89, a sensitivity of 84.10 %, and a specificity of 81.08 % in classifying echocardiographic videos into high and low quality. Furthermore, through explainable AI techniques, our model identifies specific quality issues such as missing cardiac walls, distorted or poorly positioned chambers, and other anomalies, providing interpretable feedback for clinical applications.

{"title":"Spatiotemporal feature disentanglement for quality surveillance of left ventricular echocardiographic video using ST-R(2 + 1)D-ConvNeXt","authors":"Chin-Chieh Hsu , You-Wei Wang , Lung-Chun Lin , Ruey-Feng Chang","doi":"10.1016/j.bspc.2025.107671","DOIUrl":"10.1016/j.bspc.2025.107671","url":null,"abstract":"<div><div>The left ventricle (LV), as the primary chamber responsible for systemic circulation, plays a crucial role in cardiac function assessment. Echocardiography which particularly focuses on LV, is vital for cardiac disease diagnosis. However, the diagnostic accuracy heavily depends on image quality, which requires systematic assessment. In this study, we propose a two-stage deep learning approach for echocardiographic quality surveillance using a dataset of 514 annotated videos. The first stage employs EchoNet, to extract LV volumes of interest. The second stage introduces ST-R(2 + 1)D-ConvNeXt, a novel ConvNeXt-based model designed to disentangle spatiotemporal features and leverage echocardiographic hallmarks within the apical-four-chamber (A4C) dynamic echocardiogram data. The proposed approach achieves an accuracy of 82.63 %, an Area Under the Curve (AUC) of 0.89, a sensitivity of 84.10 %, and a specificity of 81.08 % in classifying echocardiographic videos into high and low quality. Furthermore, through explainable AI techniques, our model identifies specific quality issues such as missing cardiac walls, distorted or poorly positioned chambers, and other anomalies, providing interpretable feedback for clinical applications.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107671"},"PeriodicalIF":4.9,"publicationDate":"2025-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143394735","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}

引用次数: 0

Periodicity constrained and block accelerated thin plate spline approach for cardiac motion estimation

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-13 DOI: 10.1016/j.bspc.2025.107655

Yunfeng Yang , Lihui Zhu , Zekuan Yang , Yuqi Zhu , Qiyin Huang , Pengcheng Shi , Qiang Lin , Xiaohu Zhao , Zhenghui Hu

In this paper, we propose a periodicity constrained and block accelerated Thin Plate Spline (TPS) approach for cardiac motion estimation from periodic medical image sequences. The TPS transformation is confined to specific sub-blocks to cover the motion range of the matching points during the cardiac cycle, which captured sufficient motion information while preserving computational efficiency. A periodic constraint is introduced to ensure motional consistency throughout the entire cardiac motion. The feasibility of the proposed approach was validated using the Lenna test image, further validation was conducted using MRI datasets from the Cardiac Motion Analysis Challenge (CMAC), demonstrating accurate motion estimation capability with an endpoint error (

E E

) of less than 1 pixel and an angular error (

A E

) of less than 5 degrees. Finally, this approach was applied to real cardiac MRI data, and the motion estimation results were shown to be consistent with the assessment of medical experts. Experimental validation demonstrates that the proposed approach provides enhanced computational flexibility in motion estimation, while expert input ensures an optimal balance between computational efficiency and precision.

{"title":"Periodicity constrained and block accelerated thin plate spline approach for cardiac motion estimation","authors":"Yunfeng Yang , Lihui Zhu , Zekuan Yang , Yuqi Zhu , Qiyin Huang , Pengcheng Shi , Qiang Lin , Xiaohu Zhao , Zhenghui Hu","doi":"10.1016/j.bspc.2025.107655","DOIUrl":"10.1016/j.bspc.2025.107655","url":null,"abstract":"<div><div>In this paper, we propose a periodicity constrained and block accelerated Thin Plate Spline (TPS) approach for cardiac motion estimation from periodic medical image sequences. The TPS transformation is confined to specific sub-blocks to cover the motion range of the matching points during the cardiac cycle, which captured sufficient motion information while preserving computational efficiency. A periodic constraint is introduced to ensure motional consistency throughout the entire cardiac motion. The feasibility of the proposed approach was validated using the Lenna test image, further validation was conducted using MRI datasets from the Cardiac Motion Analysis Challenge (CMAC), demonstrating accurate motion estimation capability with an endpoint error (<span><math><mrow><mi>E</mi><mi>E</mi></mrow></math></span>) of less than 1 pixel and an angular error (<span><math><mrow><mi>A</mi><mi>E</mi></mrow></math></span>) of less than 5 degrees. Finally, this approach was applied to real cardiac MRI data, and the motion estimation results were shown to be consistent with the assessment of medical experts. Experimental validation demonstrates that the proposed approach provides enhanced computational flexibility in motion estimation, while expert input ensures an optimal balance between computational efficiency and precision.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"104 ","pages":"Article 107655"},"PeriodicalIF":4.9,"publicationDate":"2025-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143394632","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}

引用次数: 0

MECF-Net: A prostate cancer lymph node metastasis classification method based on 18F-PSMA-1007 and 18F-FDG dual-tracer PET/CT image feature optimization

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-13 DOI: 10.1016/j.bspc.2025.107651

Junchen Hao , Yan Cui , Huiyan Jiang , Guoyu Tong , Xuena Li

Prostate cancer is one of the most common cancers in men, and the presence of lymph node metastasis is critical for determining treatment options and assessing prognosis. ¹⁸F-FDG is a widely used tracer in PET imaging for tumor diagnosis, while ¹⁸F-PSMA-1007 typically exhibits high specificity in prostate cancer cells. Therefore, we propose a novel model for classifying prostate cancer lymph node metastasis, MECF-Net, which integrates ¹⁸F-PSMA-1007 and ¹⁸F-FDG PET/CT images. Specifically, to enhance feature perception at different channel levels and multi-scale spatial dimensions, we propose the Multi-Scale Feature Extraction (MSFE) branch, which combines Squeeze-and-Excitation Attention with a newly designed Multi-Scale Spatial Enhanced Attention (MSEA). The MSEA extracts spatial feature information of tumors at various scales by employing global average pooling and max pooling aggregation operations at multiple scales. Furthermore, we introduce a Local-Global Feature Complementary Fusion (LGFCF) branch, which constructs a series of complementary fusion blocks as basic units. These blocks consist of concatenated multi-scale grouped convolutions and point-wise convolutions, enabling the complementary extraction of intra-group local spatial features and inter-channel global features. Finally, at the end of MECF-Net, we design a Multi-Feature Adaptive Fusion (MF-AF) module, based on a dynamic weight allocation mechanism, to fuse the features extracted from different branch sub-networks. Our experimental results on a private dual-tracer ¹⁸F-PSMA-1007/¹⁸F-FDG PET/CT dataset and a public single-tracer ¹⁸F-FDG PET/CT dataset demonstrate the effectiveness of MECF-Net, which achieved 0.9269 and 0.885 in ACC, respectively, 0.9156 and 0.8838 in AUC, respectively, demonstrating superior performance compared to state-of-the-art networks, as well as generalizability on the single-tracer dataset.

{"title":"MECF-Net: A prostate cancer lymph node metastasis classification method based on 18F-PSMA-1007 and 18F-FDG dual-tracer PET/CT image feature optimization","authors":"Junchen Hao , Yan Cui , Huiyan Jiang , Guoyu Tong , Xuena Li","doi":"10.1016/j.bspc.2025.107651","DOIUrl":"10.1016/j.bspc.2025.107651","url":null,"abstract":"<div><div>Prostate cancer is one of the most common cancers in men, and the presence of lymph node metastasis is critical for determining treatment options and assessing prognosis. <sup>18</sup>F-FDG is a widely used tracer in PET imaging for tumor diagnosis, while <sup>18</sup>F-PSMA-1007 typically exhibits high specificity in prostate cancer cells. Therefore, we propose a novel model for classifying prostate cancer lymph node metastasis, MECF-Net, which integrates <sup>18</sup>F-PSMA-1007 and <sup>18</sup>F-FDG PET/CT images. Specifically, to enhance feature perception at different channel levels and multi-scale spatial dimensions, we propose the Multi-Scale Feature Extraction (MSFE) branch, which combines Squeeze-and-Excitation Attention with a newly designed Multi-Scale Spatial Enhanced Attention (MSEA). The MSEA extracts spatial feature information of tumors at various scales by employing global average pooling and max pooling aggregation operations at multiple scales. Furthermore, we introduce a Local-Global Feature Complementary Fusion (LGFCF) branch, which constructs a series of complementary fusion blocks as basic units. These blocks consist of concatenated multi-scale grouped convolutions and point-wise convolutions, enabling the complementary extraction of intra-group local spatial features and inter-channel global features. Finally, at the end of MECF-Net, we design a Multi-Feature Adaptive Fusion (MF-AF) module, based on a dynamic weight allocation mechanism, to fuse the features extracted from different branch sub-networks. Our experimental results on a private dual-tracer <sup>18</sup>F-PSMA-1007/<sup>18</sup>F-FDG PET/CT dataset and a public single-tracer <sup>18</sup>F-FDG PET/CT dataset demonstrate the effectiveness of MECF-Net, which achieved 0.9269 and 0.885 in ACC, respectively, 0.9156 and 0.8838 in AUC, respectively, demonstrating superior performance compared to state-of-the-art networks, as well as generalizability on the single-tracer dataset.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107651"},"PeriodicalIF":4.9,"publicationDate":"2025-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143394681","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}

引用次数: 0

Control of end-tidal carbon dioxide during phrenic nerve stimulation with mechanical ventilation

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-12 DOI: 10.1016/j.bspc.2025.107649

Arnhold Lohse , Felix Röhren , Philip von Platen , Carl-Friedrich Benner , Dmitrij Ziles , Marius Hühn , Matthias Manfred Deininger , Thomas Breuer , Steffen Leonhardt , Marian Walter

Mechanical ventilation maintains the gas exchange of patients in the intensive care unit which is life-saving, but prolonged ventilation results in diaphragm atrophy. Phrenic nerve stimulation can keep the diaphragm active so that atrophy might be avoided. To use phrenic nerve stimulation in a clinical setting, it is important to implement a closed-loop control system that automatically adjusts stimulation parameters to achieve the desired ventilation. This study presents the development of a robust cascaded control system for end-tidal carbon dioxide using phrenic nerve stimulation. The control system was validated in simulations with 100 virtual patients, in which the conditions of the phrenic nerve stimulation and the patient’s condition changed, as well as in animal trials using pigs. The control system proved to be robust to end-tidal carbon dioxide perturbations, such as changing stimulation efficiency, varying patient conditions, and disconnection, in both simulations and animal trials. Regarding reference tracking, the control system achieved a settling time of 5.5 min–14 min in simulations and of 7.3 min–38.8 min in animal trials. The proposed control system can be used for further development of feedback-controlled phrenic nerve stimulation in the intensive care unit.

{"title":"Control of end-tidal carbon dioxide during phrenic nerve stimulation with mechanical ventilation","authors":"Arnhold Lohse , Felix Röhren , Philip von Platen , Carl-Friedrich Benner , Dmitrij Ziles , Marius Hühn , Matthias Manfred Deininger , Thomas Breuer , Steffen Leonhardt , Marian Walter","doi":"10.1016/j.bspc.2025.107649","DOIUrl":"10.1016/j.bspc.2025.107649","url":null,"abstract":"<div><div>Mechanical ventilation maintains the gas exchange of patients in the intensive care unit which is life-saving, but prolonged ventilation results in diaphragm atrophy. Phrenic nerve stimulation can keep the diaphragm active so that atrophy might be avoided. To use phrenic nerve stimulation in a clinical setting, it is important to implement a closed-loop control system that automatically adjusts stimulation parameters to achieve the desired ventilation. This study presents the development of a robust cascaded control system for end-tidal carbon dioxide using phrenic nerve stimulation. The control system was validated in simulations with 100 virtual patients, in which the conditions of the phrenic nerve stimulation and the patient’s condition changed, as well as in animal trials using pigs. The control system proved to be robust to end-tidal carbon dioxide perturbations, such as changing stimulation efficiency, varying patient conditions, and disconnection, in both simulations and animal trials. Regarding reference tracking, the control system achieved a settling time of 5.5min–14min in simulations and of 7.3min–38.8min in animal trials. The proposed control system can be used for further development of feedback-controlled phrenic nerve stimulation in the intensive care unit.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107649"},"PeriodicalIF":4.9,"publicationDate":"2025-02-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143394679","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}

引用次数: 0