Biomedical Signal Processing and Control最新文献_第7页

Evaluating the effectiveness of nighttime natural virtual scene on relaxation and sleepiness

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-08 DOI: 10.1016/j.bspc.2025.107596

En-Chen Chen , Tsai-Yen Li

The natural environment significantly affects the release of human stress and sleep quality. Variations in lighting and color brought about by the sun play a crucial role in the regulation of our physiological functions and circadian rhythm. This study aims to examine the effects of virtual beach environments at different times of the day (including day and nighttime) on relaxation and the promotion of sleepiness, compared to a control group using self-relaxation methods. The research variables included self-relaxation, daytime, and nighttime scenarios, with the nighttime scene featuring a starry sky and a campfire. We analyze participants’ physiological and psychological responses to uncover the impact of virtual scenes and self-regulation on transitioning from stress to relaxation. Psychological questionnaires gauged the levels of relaxation, calmness, and sleepiness. Physiologically, brain waves, heart rate variability, and heartbeat changes were evaluated. The results of our study indicate that all three groups could relax from stress, with participants immersed in the starry sky scene exhibiting the most significant reduction in heart rate and an increased sense of sleepiness. The self-relaxation group showed higher theta waves than other relaxation groups, possibly tied to deep contemplation. Through this study, we gained a better understanding of the effects of natural virtual reality environments on stress relief and the promotion of drowsiness. We plan to use the starry sky with a campfire virtual scene in experiments involving individuals who struggle with sleep difficulties in the future.

{"title":"Evaluating the effectiveness of nighttime natural virtual scene on relaxation and sleepiness","authors":"En-Chen Chen , Tsai-Yen Li","doi":"10.1016/j.bspc.2025.107596","DOIUrl":"10.1016/j.bspc.2025.107596","url":null,"abstract":"<div><div>The natural environment significantly affects the release of human stress and sleep quality. Variations in lighting and color brought about by the sun play a crucial role in the regulation of our physiological functions and circadian rhythm. This study aims to examine the effects of virtual beach environments at different times of the day (including day and nighttime) on relaxation and the promotion of sleepiness, compared to a control group using self-relaxation methods. The research variables included self-relaxation, daytime, and nighttime scenarios, with the nighttime scene featuring a starry sky and a campfire. We analyze participants’ physiological and psychological responses to uncover the impact of virtual scenes and self-regulation on transitioning from stress to relaxation. Psychological questionnaires gauged the levels of relaxation, calmness, and sleepiness. Physiologically, brain waves, heart rate variability, and heartbeat changes were evaluated. The results of our study indicate that all three groups could relax from stress, with participants immersed in the starry sky scene exhibiting the most significant reduction in heart rate and an increased sense of sleepiness. The self-relaxation group showed higher theta waves than other relaxation groups, possibly tied to deep contemplation. Through this study, we gained a better understanding of the effects of natural virtual reality environments on stress relief and the promotion of drowsiness. We plan to use the starry sky with a campfire virtual scene in experiments involving individuals who struggle with sleep difficulties in the future.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107596"},"PeriodicalIF":4.9,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143350991","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

Anomaly detection applied to the classification of cytology images

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-08 DOI: 10.1016/j.bspc.2025.107625

Carlo Bruno Marta , Manuel Doblaré , Jónathan Heras , Gadea Mata , Teresa Ramírez

Cytology is a branch of pathology that diagnoses diseases and identifies tumours by looking at single cells, or small clusters of cells, using images observed under microscopes. Traditionally, pathologists manually analyse cytology images, a time-consuming and subjective task that could be significantly accelerated and improved through the application of computer vision and deep learning techniques. However, existing deep learning methods for cytology images need annotation at the cell level, a laborious and cumbersome process that requires the ability of pathologists with high experience. In this paper, we tackle this issue by using an anomaly detection approach that requires only annotation at the level of cytology images. Our approach splits cytology images into patches, and then uses an anomaly detection model to highlight anomalous cells. For the anomaly detection model, several reconstruction-based and embedding-based methods have been studied, the latter showing a better performance than the former. In particular, the best reconstruction-based method, based on a GAN model, achieved a perfect recall, a precision of 73.61%, and an AUROC of 69.8%; whereas, the best embedding-based method, being the PatchCore algorithm with a ResNet 50 backbone, obtained a perfect recall, a precision of 98.39%, and an AUROC of 99.98%. Finally, in order to facilitate the usage of our approach by pathologists, an ImageJ macro has been implemented. Thanks to this work, the analysis of cytology images and the diagnosis of associated diseases will be faster and more reliable.

{"title":"Anomaly detection applied to the classification of cytology images","authors":"Carlo Bruno Marta , Manuel Doblaré , Jónathan Heras , Gadea Mata , Teresa Ramírez","doi":"10.1016/j.bspc.2025.107625","DOIUrl":"10.1016/j.bspc.2025.107625","url":null,"abstract":"<div><div>Cytology is a branch of pathology that diagnoses diseases and identifies tumours by looking at single cells, or small clusters of cells, using images observed under microscopes. Traditionally, pathologists manually analyse cytology images, a time-consuming and subjective task that could be significantly accelerated and improved through the application of computer vision and deep learning techniques. However, existing deep learning methods for cytology images need annotation at the cell level, a laborious and cumbersome process that requires the ability of pathologists with high experience. In this paper, we tackle this issue by using an anomaly detection approach that requires only annotation at the level of cytology images. Our approach splits cytology images into patches, and then uses an anomaly detection model to highlight anomalous cells. For the anomaly detection model, several reconstruction-based and embedding-based methods have been studied, the latter showing a better performance than the former. In particular, the best reconstruction-based method, based on a GAN model, achieved a perfect recall, a precision of 73.61%, and an AUROC of 69.8%; whereas, the best embedding-based method, being the PatchCore algorithm with a ResNet 50 backbone, obtained a perfect recall, a precision of 98.39%, and an AUROC of 99.98%. Finally, in order to facilitate the usage of our approach by pathologists, an ImageJ macro has been implemented. Thanks to this work, the analysis of cytology images and the diagnosis of associated diseases will be faster and more reliable.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107625"},"PeriodicalIF":4.9,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143350988","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

scCAT: Single-cell Combined graph Attentional clustering for scRNA-seq analysis

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-08 DOI: 10.1016/j.bspc.2025.107502

Yufei Gao , Wenbo Zhang , Yameng Zhang , Yujie Shi , Lei Shi , Hailing Wang , Guozhen Cheng

The advent of single-cell RNA sequencing (scRNA-seq) has revolutionized gene expression research at the single-cell level, enabling the study of cellular heterogeneity and identification of rare cell populations. Deep clustering is crucial for analyzing scRNA-seq datasets by assigning cells into subpopulations. However, inherent sparsity and variability in gene expression pose challenges to clustering accuracy. To address these issues, a novel unsupervised deep clustering approach named single-cell Combined graph Attentional clustering (scCAT) is introduced. The method designs a dual-branch joint dimensionality reduction (JDR) module to learn gene expression. This strategy preserves key variance while capturing complex nonlinear relationships, effectively addressing the high-dimensionality challenges of single-cell data. Additionally, a Zero-inflated negative binomial (ZINB) distribution is integrated within the JDR to tackle significant noise, sparsity, and zero-inflation challenges. A graph attention autoencoder (GATE) then processes the graph-structured data, enhancing the integration of integrating cellular topological relationships and gene expressions. Finally, a k-means-based self-optimization technique refines clustering while synchronizing with representation learning. Experimental evaluations on eight scRNA-seq datasets demonstrate that scCAT significantly improves clustering performance, mitigates the impact of inherent data defects. The embeddings learned from both linear and non-linear perspectives eliminate inter-component interactions, uncovering complex underlying structures and enhancing the recognition of cellular topological relationships when combined with graph neural networks.

{"title":"scCAT: Single-cell Combined graph Attentional clustering for scRNA-seq analysis","authors":"Yufei Gao , Wenbo Zhang , Yameng Zhang , Yujie Shi , Lei Shi , Hailing Wang , Guozhen Cheng","doi":"10.1016/j.bspc.2025.107502","DOIUrl":"10.1016/j.bspc.2025.107502","url":null,"abstract":"<div><div>The advent of single-cell RNA sequencing (scRNA-seq) has revolutionized gene expression research at the single-cell level, enabling the study of cellular heterogeneity and identification of rare cell populations. Deep clustering is crucial for analyzing scRNA-seq datasets by assigning cells into subpopulations. However, inherent sparsity and variability in gene expression pose challenges to clustering accuracy. To address these issues, a novel unsupervised deep clustering approach named single-cell Combined graph Attentional clustering (scCAT) is introduced. The method designs a dual-branch joint dimensionality reduction (JDR) module to learn gene expression. This strategy preserves key variance while capturing complex nonlinear relationships, effectively addressing the high-dimensionality challenges of single-cell data. Additionally, a Zero-inflated negative binomial (ZINB) distribution is integrated within the JDR to tackle significant noise, sparsity, and zero-inflation challenges. A graph attention autoencoder (GATE) then processes the graph-structured data, enhancing the integration of integrating cellular topological relationships and gene expressions. Finally, a k-means-based self-optimization technique refines clustering while synchronizing with representation learning. Experimental evaluations on eight scRNA-seq datasets demonstrate that scCAT significantly improves clustering performance, mitigates the impact of inherent data defects. The embeddings learned from both linear and non-linear perspectives eliminate inter-component interactions, uncovering complex underlying structures and enhancing the recognition of cellular topological relationships when combined with graph neural networks.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107502"},"PeriodicalIF":4.9,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143372899","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

GCN-LSTM: A hybrid graph convolutional network model for schizophrenia classification

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-08 DOI: 10.1016/j.bspc.2025.107657

Bethany Gosala , Avnish Ramvinay Singh , Himanshu Tiwari , Manjari Gupta

Schizophrenia is a complex mental disorder that influences one’s perceptions, thought processes, social behavior, emotional responses, etc. Electroencephalography is a non-invasive brain imaging technique that measures the brain’s electrical activity. The EEG signals are used to study and analyze the human brain. Graphs have always been one of the best ways to represent information. With the inspiration from graphs, in this paper, we developed a novel GCN-LSTM model, a graph-based hybrid deep learning model for classifying schizophrenia from Healthy Control. We used the Institute of Psychiatry and Neurology in Warsaw, Poland dataset to experiment with the developed models; Raw EEG signals were pre-processed and divided into segments of 5-sec and 8-sec. We extracted 14 different features from these epochs, 7 each from the time and frequency domains. After feature extraction, we constructed the graphs out of epochs of 5-sec and 8-sec, where EEG electrodes are considered as nodes and how signal flows between EEG channels as edges. These graphs were fed to the developed GCN-LSTM model for the classification. We also used different seeds and 5-fold cross-validation to avoid overfitting. We conducted several experiments and achieved average accuracy across all seeds as 99.25 ± 0.24 for the GCN-LSTM model with 8-sec epoch data, also Precision of 99.28 ± 0.22 %, F1 score of 99.24 ± 0.24 %, Specificity of 98.73 ± 0.64; Sensitivity of 99.67 ± 0.28 and AUC of 99.20 ± 0.27. We used t-test and one-way ANOVA to study the statistical significance of the extracted features. We found zero crossing rate, mobility (Hjorth parameter), peak frequency, and gamma band.

{"title":"GCN-LSTM: A hybrid graph convolutional network model for schizophrenia classification","authors":"Bethany Gosala , Avnish Ramvinay Singh , Himanshu Tiwari , Manjari Gupta","doi":"10.1016/j.bspc.2025.107657","DOIUrl":"10.1016/j.bspc.2025.107657","url":null,"abstract":"<div><div>Schizophrenia is a complex mental disorder that influences one’s perceptions, thought processes, social behavior, emotional responses, etc. Electroencephalography is a non-invasive brain imaging technique that measures the brain’s electrical activity. The EEG signals are used to study and analyze the human brain. Graphs have always been one of the best ways to represent information. With the inspiration from graphs, in this paper, we developed a novel GCN-LSTM model, a graph-based hybrid deep learning model for classifying schizophrenia from Healthy Control. We used the Institute of Psychiatry and Neurology in Warsaw, Poland dataset to experiment with the developed models; Raw EEG signals were pre-processed and divided into segments of 5-sec and 8-sec. We extracted 14 different features from these epochs, 7 each from the time and frequency domains. After feature extraction, we constructed the graphs out of epochs of 5-sec and 8-sec, where EEG electrodes are considered as nodes and how signal flows between EEG channels as edges. These graphs were fed to the developed GCN-LSTM model for the classification. We also used different seeds and 5-fold cross-validation to avoid overfitting. We conducted several experiments and achieved average accuracy across all seeds as 99.25 ± 0.24 for the GCN-LSTM model with 8-sec epoch data, also Precision of 99.28 ± 0.22 %, F1 score of 99.24 ± 0.24 %, Specificity of 98.73 ± 0.64; Sensitivity of 99.67 ± 0.28 and AUC of 99.20 ± 0.27. We used <em>t</em>-test and one-way ANOVA to study the statistical significance of the extracted features. We found zero crossing rate, mobility (Hjorth parameter), peak frequency, and gamma band.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107657"},"PeriodicalIF":4.9,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143350017","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

Vascular reactivity characterized by PPG-derived pulse wave velocity

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-08 DOI: 10.1016/j.bspc.2025.107641

Pablo Armañac-Julián , Spyridon Kontaxis , Jesús Lázaro , Andrius Rapalis , Marius Brazaitis , Vaidotas Marozas , Pablo Laguna , Raquel Bailón , Eduardo Gil

Vascular reactivity is the capacity of the blood vessels to adapt under physiological and environmental stimuli. Heat stress causes changes at vascular level affecting pulse wave velocity (PWV), which can be non-invasively obtained using pulse photoplethysmography (PPG). The study aim is to characterize non-invasive and reliable PPG-derived PWV surrogates that are able to assess vascular reactivity, using data from fifteen healthy male volunteers under heat stress conditions. Pulse arrival time (PAT) is a recognized PWV surrogate measure, but our study explores further by including pulse transit time difference (PTTD) and pulse wave decomposition analysis (PDA). Our results indicate a significant linear decrease in PAT and PDA under heat stress, with an approximate 15% reduction compared to the relax phase, closely correlating with heart rate (HR) alterations. This correlation is likely influenced by factors such as the pre-ejection period or stroke volume changes. In contrast, PTTD demonstrates a distinct pattern: it exhibits significant and rapid changes during the initial exposure to heat stress, with an approximate 30% reduction, yet shows minimal intra-stage variations (around 0 ms/min compared to 2.5 ms/min in PAT). This suggests that PTTD, in measuring acute sympathetic activation responses, effectively minimizes the impact of HR-related phenomena that significantly influence PAT and PDA measurements. Our study highlights PTTD as an underexplored yet promising measure for accurately assessing vasoconstriction and vascular reactivity.

{"title":"Vascular reactivity characterized by PPG-derived pulse wave velocity","authors":"Pablo Armañac-Julián , Spyridon Kontaxis , Jesús Lázaro , Andrius Rapalis , Marius Brazaitis , Vaidotas Marozas , Pablo Laguna , Raquel Bailón , Eduardo Gil","doi":"10.1016/j.bspc.2025.107641","DOIUrl":"10.1016/j.bspc.2025.107641","url":null,"abstract":"<div><div>Vascular reactivity is the capacity of the blood vessels to adapt under physiological and environmental stimuli. Heat stress causes changes at vascular level affecting pulse wave velocity (PWV), which can be non-invasively obtained using pulse photoplethysmography (PPG). The study aim is to characterize non-invasive and reliable PPG-derived PWV surrogates that are able to assess vascular reactivity, using data from fifteen healthy male volunteers under heat stress conditions. Pulse arrival time (PAT) is a recognized PWV surrogate measure, but our study explores further by including pulse transit time difference (PTTD) and pulse wave decomposition analysis (PDA). Our results indicate a significant linear decrease in PAT and PDA under heat stress, with an approximate 15% reduction compared to the relax phase, closely correlating with heart rate (HR) alterations. This correlation is likely influenced by factors such as the pre-ejection period or stroke volume changes. In contrast, PTTD demonstrates a distinct pattern: it exhibits significant and rapid changes during the initial exposure to heat stress, with an approximate 30% reduction, yet shows minimal intra-stage variations (around 0 ms/min compared to 2.5 ms/min in PAT). This suggests that PTTD, in measuring acute sympathetic activation responses, effectively minimizes the impact of HR-related phenomena that significantly influence PAT and PDA measurements. Our study highlights PTTD as an underexplored yet promising measure for accurately assessing vasoconstriction and vascular reactivity.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107641"},"PeriodicalIF":4.9,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143350986","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

ProsGradNet: An effective and structured CNN approach for prostate cancer grading from histopathology images

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-08 DOI: 10.1016/j.bspc.2025.107626

Akshaya Prabhu , Sravya Nedungatt , Shyam Lal , Jyoti Kini

Prostate cancer (PCa) is one of the most prevalent and potentially fatal malignancies affecting men globally. The incidence of prostate cancer is expected to double by 2040, posing significant health challenges. This anticipated increase underscores the urgent need for early and precise diagnosis to facilitate effective treatment and management. Histopathological analysis using Gleason grading system plays a pivotal role in clinical decision making by classifying cancer subtypes based on their cellular characteristics. This paper proposes a novel deep CNN model named as Prostate Grading Network (ProsGradNet), for the automatic grading of PCa from histopathological images. Central to the approach is the novel Context Guided Shared Channel Residual (CGSCR) block, that introduces structured methods for channel splitting and clustering, by varying group sizes. By grouping channels into 2, 4, and 8, it prioritizes deeper layer features, enhancing local semantic content and abstract feature representation. This methodological advancement significantly boosts classification accuracy, achieving an impressive 92.88% on Prostate Gleason dataset, outperforming other CNN models. To demonstrate the generalizability of ProsGradNet over different datasets, experiments are performed on Kasturba Medical College (KMC) Kidney dataset as well. The results further confirm the superiority of the proposed ProsGradNet model, with a classification accuracy of 92.68% on the KMC Kidney dataset. This demonstrates the model’s potential to be applied effectively across various histopathological datasets, making it a valuable tool to fight against cancer.

{"title":"ProsGradNet: An effective and structured CNN approach for prostate cancer grading from histopathology images","authors":"Akshaya Prabhu , Sravya Nedungatt , Shyam Lal , Jyoti Kini","doi":"10.1016/j.bspc.2025.107626","DOIUrl":"10.1016/j.bspc.2025.107626","url":null,"abstract":"<div><div>Prostate cancer (PCa) is one of the most prevalent and potentially fatal malignancies affecting men globally. The incidence of prostate cancer is expected to double by 2040, posing significant health challenges. This anticipated increase underscores the urgent need for early and precise diagnosis to facilitate effective treatment and management. Histopathological analysis using Gleason grading system plays a pivotal role in clinical decision making by classifying cancer subtypes based on their cellular characteristics. This paper proposes a novel deep CNN model named as Prostate Grading Network (ProsGradNet), for the automatic grading of PCa from histopathological images. Central to the approach is the novel Context Guided Shared Channel Residual (CGSCR) block, that introduces structured methods for channel splitting and clustering, by varying group sizes. By grouping channels into 2, 4, and 8, it prioritizes deeper layer features, enhancing local semantic content and abstract feature representation. This methodological advancement significantly boosts classification accuracy, achieving an impressive 92.88% on Prostate Gleason dataset, outperforming other CNN models. To demonstrate the generalizability of ProsGradNet over different datasets, experiments are performed on Kasturba Medical College (KMC) Kidney dataset as well. The results further confirm the superiority of the proposed ProsGradNet model, with a classification accuracy of 92.68% on the KMC Kidney dataset. This demonstrates the model’s potential to be applied effectively across various histopathological datasets, making it a valuable tool to fight against cancer.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107626"},"PeriodicalIF":4.9,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143350987","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

Robust semi-supervised extraction of information using functional near-infrared spectroscopy for diagnosing depression

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-08 DOI: 10.1016/j.bspc.2025.107571

Shi Qiao , Jitao Zhong , Lu Zhang , Hele Liu , Jiangang Li , Hong Peng , Bin Hu

Depression has become one of the major psychological disorders faced by contemporary human beings, and the current depression diagnosis model, which is based on the doctor’s questioning as the main diagnostic basis, can no longer meet the requirements of early detection and treatment of depression. To this end, this paper proposes a novel feature extraction algorithm, Robust Semi-Supervised Information Extraction (RSSIE), which is a joint optimization process of the

l_{2, 1}

-norm, the graph Laplace operator, and some data labels, different from the traditional Non-negative Matrix Factorization (NMF), or Conceptual Factorization (CF), which decomposes the original high-dimensional matrix into two low-dimensional matrices only, in contrast, our proposed algorithm takes into account the robustness of the features and the flow structure of the features, makes full use of the existing labeling information, enhances the ability of the base matrix to contribute to depression diagnosis, and significantly improves the classification accuracy compared to other relevant methods. In addition, we developed an audio stimulation paradigm for functional near-infrared spectroscopy (fNIRS) measurements in task-state experiments. Finally, our algorithm shows the best classification results for negative audio stimuli, i.e., accuracy (92.5%), specificity (93.3%), sensitivity (91.5%), and AUC (91.0%), which is superior to traditional machine learning algorithms and can be used as an effective feature extraction method for depression diagnosis.

{"title":"Robust semi-supervised extraction of information using functional near-infrared spectroscopy for diagnosing depression","authors":"Shi Qiao , Jitao Zhong , Lu Zhang , Hele Liu , Jiangang Li , Hong Peng , Bin Hu","doi":"10.1016/j.bspc.2025.107571","DOIUrl":"10.1016/j.bspc.2025.107571","url":null,"abstract":"<div><div>Depression has become one of the major psychological disorders faced by contemporary human beings, and the current depression diagnosis model, which is based on the doctor’s questioning as the main diagnostic basis, can no longer meet the requirements of early detection and treatment of depression. To this end, this paper proposes a novel feature extraction algorithm, Robust Semi-Supervised Information Extraction (RSSIE), which is a joint optimization process of the <span><math><msub><mrow><mi>l</mi></mrow><mrow><mn>2</mn><mo>,</mo><mn>1</mn></mrow></msub></math></span>-norm, the graph Laplace operator, and some data labels, different from the traditional Non-negative Matrix Factorization (NMF), or Conceptual Factorization (CF), which decomposes the original high-dimensional matrix into two low-dimensional matrices only, in contrast, our proposed algorithm takes into account the robustness of the features and the flow structure of the features, makes full use of the existing labeling information, enhances the ability of the base matrix to contribute to depression diagnosis, and significantly improves the classification accuracy compared to other relevant methods. In addition, we developed an audio stimulation paradigm for functional near-infrared spectroscopy (fNIRS) measurements in task-state experiments. Finally, our algorithm shows the best classification results for negative audio stimuli, i.e., accuracy (92.5%), specificity (93.3%), sensitivity (91.5%), and AUC (91.0%), which is superior to traditional machine learning algorithms and can be used as an effective feature extraction method for depression diagnosis.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107571"},"PeriodicalIF":4.9,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143350993","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

Surgical tool detection in open surgery based on improved-YOLOv8

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-08 DOI: 10.1016/j.bspc.2025.107548

Zhaokun Xu, Feng Luo, Feng Chen, Hang Wu, Ming Yu

Efficient detection of surgical tools provides positional and categorized information in visual method for intelligent analysis of surgical videos. However, detection methods mainly focus on Minimally Invasive Surgery (MIS). Few studies have been conducted on open surgery, mainly limited by the lack of public database. Complicated operation and incomplete surgical tools truncated by hands lead to more challenges during detection in open surgery. To overcome these difficulties, we introduce AEDCSSAD dataset with 3 open surgeries and complete annotations for surgical tool detection and employ RET-YOLOv8 as the tool detection framework. The redesigned C2f structure and the Efficient Channel Attention (ECA) module is adopted to improve feature extraction ability. We propose the two-stage bounding box regression loss function to develop the efficiency of training. In addition, we explore the effect of the size distribution of surgical tools among images on detection performance. The results prove that the balanced size proportion in train, validation and test set has significant importance to improve the detection performance. Our method is evaluated on both open surgery and MIS, and achieves 92.7 % mean Average Precision (mAP) on AEDCSSAD. It is worth noting that the RET-YOLOv8 method reaches 3.1 % improvement of Average Precision (AP) in surgical consumable detection. Our algorithm achieves 2.2 % improvement of mAP with decreased FLOPS by 3.7 % and parameters by 6.0 % on m2cai16-tool-locations dataset. The experiments show that our detection algorithm performs effectively with potential in detecting surgical tools with flexible volume.

{"title":"Surgical tool detection in open surgery based on improved-YOLOv8","authors":"Zhaokun Xu, Feng Luo, Feng Chen, Hang Wu, Ming Yu","doi":"10.1016/j.bspc.2025.107548","DOIUrl":"10.1016/j.bspc.2025.107548","url":null,"abstract":"<div><div>Efficient detection of surgical tools provides positional and categorized information in visual method for intelligent analysis of surgical videos. However, detection methods mainly focus on Minimally Invasive Surgery (MIS). Few studies have been conducted on open surgery, mainly limited by the lack of public database. Complicated operation and incomplete surgical tools truncated by hands lead to more challenges during detection in open surgery. To overcome these difficulties, we introduce AEDCSSAD dataset with 3 open surgeries and complete annotations for surgical tool detection and employ RET-YOLOv8 as the tool detection framework. The redesigned C2f structure and the Efficient Channel Attention (ECA) module is adopted to improve feature extraction ability. We propose the two-stage bounding box regression loss function to develop the efficiency of training. In addition, we explore the effect of the size distribution of surgical tools among images on detection performance. The results prove that the balanced size proportion in train, validation and test set has significant importance to improve the detection performance. Our method is evaluated on both open surgery and MIS, and achieves 92.7 % mean Average Precision (mAP) on AEDCSSAD. It is worth noting that the RET-YOLOv8 method reaches 3.1 % improvement of Average Precision (AP) in surgical consumable detection. Our algorithm achieves 2.2 % improvement of mAP with decreased FLOPS by 3.7 % and parameters by 6.0 % on m2cai16-tool-locations dataset. The experiments show that our detection algorithm performs effectively with potential in detecting surgical tools with flexible volume.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107548"},"PeriodicalIF":4.9,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143372898","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

Attention enabled precise haemoglobin prediction: A GNN based non-invasive approach

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-08 DOI: 10.1016/j.bspc.2025.107604

Sunanda Das, Abhishek Kesarwani, Dakshina Ranjan Kisku, Mamata Dalui

Anaemia, a widespread condition affecting billions globally, is characterized by low blood haemoglobin levels. Traditional invasive methods for detecting anaemia are expensive and challenging to implement on a global scale, highlighting the need for a non-invasive, cost-effective, and user-friendly solution. In response to this need, this study aims to develop a non-invasive anaemia detection system by developing an innovative, fully automated, portable, low-cost embedded device. The current approach integrates advanced computational techniques with the traditional practice of estimating blood haemoglobin levels by observing pallor in the fingernail of a human subject. The proposed system functions by inducing changes in nail pallor through controlled pressure application and release facilitated by the indigenous device, and then measuring the rate of colour changes to correlate with blood haemoglobin concentration. Our method involves extracting features from the nail pallor video using AneumoNet model followed by Graph Neural Network (GNN) module for feature selection. These selected features are then fed into seven state-of-the-art regression models followed by a multi-head self-attention regression model that combines the strengths of the multiple state-of-the-art regression models and captures intricate relationships between inputs, thereby enhancing prediction accuracy and robustness. The proposed approach effectively predicts haemoglobin levels, achieving an RMSE of 0.43, prediction accuracy of 94%, classification accuracy of 97.04%, sensitivity and specificity of 0.98 and 0.94 respectively, over 644 samples, demonstrating its efficacy as compared to other state-of-the-art solutions. In conclusion, this method sets a new benchmark in non-invasive diagnostic techniques, enhancing both clinical and remote healthcare applications.

{"title":"Attention enabled precise haemoglobin prediction: A GNN based non-invasive approach","authors":"Sunanda Das, Abhishek Kesarwani, Dakshina Ranjan Kisku, Mamata Dalui","doi":"10.1016/j.bspc.2025.107604","DOIUrl":"10.1016/j.bspc.2025.107604","url":null,"abstract":"<div><div>Anaemia, a widespread condition affecting billions globally, is characterized by low blood haemoglobin levels. Traditional invasive methods for detecting anaemia are expensive and challenging to implement on a global scale, highlighting the need for a non-invasive, cost-effective, and user-friendly solution. In response to this need, this study aims to develop a non-invasive anaemia detection system by developing an innovative, fully automated, portable, low-cost embedded device. The current approach integrates advanced computational techniques with the traditional practice of estimating blood haemoglobin levels by observing pallor in the fingernail of a human subject. The proposed system functions by inducing changes in nail pallor through controlled pressure application and release facilitated by the indigenous device, and then measuring the rate of colour changes to correlate with blood haemoglobin concentration. Our method involves extracting features from the nail pallor video using <em>AneumoNet</em> model followed by Graph Neural Network (GNN) module for feature selection. These selected features are then fed into seven state-of-the-art regression models followed by a multi-head self-attention regression model that combines the strengths of the multiple state-of-the-art regression models and captures intricate relationships between inputs, thereby enhancing prediction accuracy and robustness. The proposed approach effectively predicts haemoglobin levels, achieving an RMSE of 0.43, prediction accuracy of 94%, classification accuracy of 97.04%, sensitivity and specificity of 0.98 and 0.94 respectively, over 644 samples, demonstrating its efficacy as compared to other state-of-the-art solutions. In conclusion, this method sets a new benchmark in non-invasive diagnostic techniques, enhancing both clinical and remote healthcare applications.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107604"},"PeriodicalIF":4.9,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143350990","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

A novel twin vision transformer framework for crop disease classification with deformable attention

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2025-02-07 DOI: 10.1016/j.bspc.2025.107551

Smitha Padshetty , Ambika

In recent years, crop disease has grown substantially due to the lack of immunity in crops and climate changes. This results in crop destruction and poor crop cultivation. Quick and accurate classification of crop disease is crucial for enhancing agricultural yield and preventing crop damage. This research proposes a novel Twin Vision Transformer framework for accurate and quick classification of crop disease. The proposed model utilized a Depthwise Separable Visual Geometry Group (16) technique to extract the crop image features that incorporated Depthwise Separable Convolution and channel normalization. The Depthwise Separable Convolution is deployed for computing the channel and space separation and mitigating the computational complexity. The channel normalization is employed to normalize each activation using local responses. Further, the Twin Vision Transformer model integrates a transformer encoder, Edge Aware Enhancement Module, and classifier head for accurate disease classification. The transformer encoder layer is applied to capture the complex relationship that encompasses deformable attention for identifying the image patch’s relevant part and enhancing the flexibility of the model, and spatial-channel attention for enhancing the crucial feature channels and spatial locations. Moreover, the residual Multilayer Perceptron with Swish-Gated Linear Units is employed to enhance the generalization capability of the model by mitigating overfitting. The Edge Aware Enhancement Module is developed to improve the features around the identified edges. Additionally, the Squeeze and Excitation block in the classifier head is applied to emphasize the features of crop disease and to enhance the model’s classification accuracy. The proposed Twin Vision Transformer model provided a higher classification accuracy of 98.88 %, and the experimental analysis confirms that the proposed Twin Vision Transformer model is more accurate and reliable in classifying crop disease than existing crop disease classification approaches.

{"title":"A novel twin vision transformer framework for crop disease classification with deformable attention","authors":"Smitha Padshetty , Ambika","doi":"10.1016/j.bspc.2025.107551","DOIUrl":"10.1016/j.bspc.2025.107551","url":null,"abstract":"<div><div>In recent years, crop disease has grown substantially due to the lack of immunity in crops and climate changes. This results in crop destruction and poor crop cultivation. Quick and accurate classification of crop disease is crucial for enhancing agricultural yield and preventing crop damage. This research proposes a novel Twin Vision Transformer framework for accurate and quick classification of crop disease. The proposed model utilized a Depthwise Separable Visual Geometry Group (16) technique to extract the crop image features that incorporated Depthwise Separable Convolution and channel normalization. The Depthwise Separable Convolution is deployed for computing the channel and space separation and mitigating the computational complexity. The channel normalization is employed to normalize each activation using local responses. Further, the Twin Vision Transformer model integrates a transformer encoder, Edge Aware Enhancement Module, and classifier head for accurate disease classification. The transformer encoder layer is applied to capture the complex relationship that encompasses deformable attention for identifying the image patch’s relevant part and enhancing the flexibility of the model, and spatial-channel attention for enhancing the crucial feature channels and spatial locations. Moreover, the residual Multilayer Perceptron with Swish-Gated Linear Units is employed to enhance the generalization capability of the model by mitigating overfitting. The Edge Aware Enhancement Module is developed to improve the features around the identified edges. Additionally, the Squeeze and Excitation block in the classifier head is applied to emphasize the features of crop disease and to enhance the model’s classification accuracy. The proposed Twin Vision Transformer model provided a higher classification accuracy of 98.88 %, and the experimental analysis confirms that the proposed Twin Vision Transformer model is more accurate and reliable in classifying crop disease than existing crop disease classification approaches.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"105 ","pages":"Article 107551"},"PeriodicalIF":4.9,"publicationDate":"2025-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143240358","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