Biomedical Signal Processing and Control最新文献

Design of an optimized rotation-invariant coordinate convolutional neural network driven medical IoT recommendation system integrating sentiment analysis for improved patient preference prediction 设计一种优化的旋转不变坐标卷积神经网络驱动的医疗物联网推荐系统，集成情感分析，改进患者偏好预测

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2026-06-15 Epub Date: 2026-02-08 DOI: 10.1016/j.bspc.2026.109742

Rethina Kumar B , P. Sudhakaran , M. Baritha Begum , S. Rajeswari

Chronic and lifestyle-related diseases are rising globally, creating significant societal and economic burdens. To support effective long-term patient monitoring, an Optimized Rotation-Invariant Coordinate Convolutional Neural Network-driven Medical IoT Recommendation System integrating Sentiment Analysis for Improved Patient Preference Prediction (RICNN-IoT-SA-IPP) is proposed. The system collects multimodal data, including physiological and behavioural signals from IoT-based healthcare sensors and combines it with patient feedback sourced from electronic health records and medical consultation platforms. A Fast Guided Median Filter (FGMF) is employed to denoise and normalize the input, followed by spatial feature extraction utilizing Synchro-Transient-Extracting Transform (STET). These features are analyzed through a Multimodal Contrastive Domain Sharing Generative Adversarial Network (MCDSGAN) to infer patient sentiment. A Rotation-Invariant Coordinate Convolutional Neural Network (RICNN) then performs preference prediction. To enhance prediction accuracy, the Levy Pelican Optimization Algorithm (LPOA) is used for optimizing feature weights and model parameters. The system performance is evaluated using Accuracy, Precision, Recall, F1-Score, Mean Absolute Error (MAE), Mean Squared Error (MSE) and Computational Time. The proposed RICNN-IoT-SA-IPP model achieved 99.32% accuracy and 98.34% precision, while maintaining low error rates with MAE = 0.0855 and MSE = 0.0864, respectively. When compared with existing models, these outcomes represent an improvement of approximately 3–5% in classification metrics and a significant reduction in prediction error. This demonstrates that the proposed framework provides highly accurate, reliable, and computationally efficient patient preference predictions.

慢性病和与生活方式有关的疾病正在全球上升，造成重大的社会和经济负担。为了支持有效的长期患者监测，提出了一种优化的旋转不变坐标卷积神经网络驱动的医疗物联网推荐系统，该系统集成了改进患者偏好预测的情感分析（RICNN-IoT-SA-IPP）。该系统收集多模式数据，包括来自基于物联网的医疗传感器的生理和行为信号，并将其与来自电子健康记录和医疗咨询平台的患者反馈相结合。采用快速引导中值滤波（FGMF）对输入进行去噪和归一化处理，然后利用同步瞬态提取变换（STET）进行空间特征提取。这些特征通过多模态对比域共享生成对抗网络（MCDSGAN）进行分析，以推断患者的情绪。然后使用旋转不变坐标卷积神经网络（RICNN）进行偏好预测。为了提高预测精度，采用Levy Pelican Optimization Algorithm （LPOA）对特征权值和模型参数进行优化。系统性能评估使用准确性，精密度，召回率，F1-Score，平均绝对误差（MAE），均方误差（MSE）和计算时间。所提出的RICNN-IoT-SA-IPP模型准确率为99.32%，精度为98.34%，同时保持较低的错误率，MAE = 0.0855, MSE = 0.0864。与现有模型相比，这些结果在分类指标上提高了约3-5%，并显著降低了预测误差。这表明所提出的框架提供了高度准确、可靠和计算效率高的患者偏好预测。

{"title":"Design of an optimized rotation-invariant coordinate convolutional neural network driven medical IoT recommendation system integrating sentiment analysis for improved patient preference prediction","authors":"Rethina Kumar B , P. Sudhakaran , M. Baritha Begum , S. Rajeswari","doi":"10.1016/j.bspc.2026.109742","DOIUrl":"10.1016/j.bspc.2026.109742","url":null,"abstract":"<div><div>Chronic and lifestyle-related diseases are rising globally, creating significant societal and economic burdens. To support effective long-term patient monitoring, an Optimized Rotation-Invariant Coordinate Convolutional Neural Network-driven Medical IoT Recommendation System integrating Sentiment Analysis for Improved Patient Preference Prediction (RICNN-IoT-SA-IPP) is proposed. The system collects multimodal data, including physiological and behavioural signals from IoT-based healthcare sensors and combines it with patient feedback sourced from electronic health records and medical consultation platforms. A Fast Guided Median Filter (FGMF) is employed to denoise and normalize the input, followed by spatial feature extraction utilizing Synchro-Transient-Extracting Transform (STET). These features are analyzed through a Multimodal Contrastive Domain Sharing Generative Adversarial Network (MCDSGAN) to infer patient sentiment. A Rotation-Invariant Coordinate Convolutional Neural Network (RICNN) then performs preference prediction. To enhance prediction accuracy, the Levy Pelican Optimization Algorithm (LPOA) is used for optimizing feature weights and model parameters. The system performance is evaluated using Accuracy, Precision, Recall, F1-Score, Mean Absolute Error (MAE), Mean Squared Error (MSE) and Computational Time. The proposed RICNN-IoT-SA-IPP model achieved 99.32% accuracy and 98.34% precision, while maintaining low error rates with MAE = 0.0855 and MSE = 0.0864, respectively. When compared with existing models, these outcomes represent an improvement of approximately 3–5% in classification metrics and a significant reduction in prediction error. This demonstrates that the proposed framework provides highly accurate, reliable, and computationally efficient patient preference predictions.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"119 ","pages":"Article 109742"},"PeriodicalIF":4.9,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191952","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

Spike sequences classification for dengue and Zika infections in mosquito neurons using deep pre-trained models 利用深度预训练模型对蚊子神经元中登革热和寨卡感染的刺突序列进行分类

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2026-06-15 Epub Date: 2026-02-09 DOI: 10.1016/j.bspc.2026.109748

Danial Sharifrazi , Nouman Javed , Roohallah Alizadehsani , Prasad N. Paradkar , U.Rajendra Acharya , Asim Bhatti

Mosquito-borne diseases are severe hazards to the health of both animals and humans. Aedes aegypti mosquitos are the primary vectors of several medically significant diseases, including dengue and Zika. Therefore, a thorough understanding of the neurons of mosquitos transmitting these diseases can be extremely beneficial in disease prevention. We hope to better comprehend the unique pattern found in considerable values of signal retrieved from mosquito neurons, known as spikes. There is currently no open-source neural spike sequence classification technique for mosquitos. To obtain outstanding outcomes, we demonstrate how to extract and classify spikes from mosquito neuron inputs using transfer learning approaches. Consequently, we highlight the role of deep pre-trained models that were trained using ImageNet weights.

The proposed methodology uses electrical spiking activity data from mosquito neurons collected with microelectrode array technology. To assess the method’s performance, data from 0, 1, 2, 3, and 7 days post-infection, reaching more than 15 million samples, were used. In this study, we also look at the influence of days post-infection on recognizing spikes in mosquito neurons.

Overall, we attempted for the first time to analyze the distinctive pattern in the spike sequence of mosquito neurons using Artificial Intelligence (AI) approaches and to determine the impact of these spikes over time.

蚊媒疾病是严重危害动物和人类健康的疾病。埃及伊蚊是几种医学上重要疾病的主要媒介，包括登革热和寨卡病毒。因此，彻底了解传播这些疾病的蚊子的神经元对预防疾病非常有益。我们希望更好地理解从蚊子神经元中获取的大量信号中发现的独特模式，即所谓的尖峰。目前还没有开源的蚊子神经脉冲序列分类技术。为了获得突出的结果，我们演示了如何使用迁移学习方法从蚊子神经元输入中提取和分类spike。因此，我们强调了使用ImageNet权重训练的深度预训练模型的作用。所提出的方法使用微电极阵列技术收集的蚊子神经元的电尖峰活动数据。为了评估该方法的性能，使用了感染后0、1、2、3和7天的数据，涉及超过1500万个样本。在这项研究中，我们还研究了感染后几天对蚊子神经元识别尖峰的影响。总体而言，我们首次尝试使用人工智能（AI）方法分析蚊子神经元尖峰序列的独特模式，并确定这些尖峰随时间的影响。

{"title":"Spike sequences classification for dengue and Zika infections in mosquito neurons using deep pre-trained models","authors":"Danial Sharifrazi , Nouman Javed , Roohallah Alizadehsani , Prasad N. Paradkar , U.Rajendra Acharya , Asim Bhatti","doi":"10.1016/j.bspc.2026.109748","DOIUrl":"10.1016/j.bspc.2026.109748","url":null,"abstract":"<div><div>Mosquito-borne diseases are severe hazards to the health of both animals and humans. <em>Aedes aegypti</em> mosquitos are the primary vectors of several medically significant diseases, including dengue and Zika. Therefore, a thorough understanding of the neurons of mosquitos transmitting these diseases can be extremely beneficial in disease prevention. We hope to better comprehend the unique pattern found in considerable values of signal retrieved from mosquito neurons, known as spikes. There is currently no open-source neural spike sequence classification technique for mosquitos. To obtain outstanding outcomes, we demonstrate how to extract and classify spikes from mosquito neuron inputs using transfer learning approaches. Consequently, we highlight the role of deep pre-trained models that were trained using ImageNet weights.</div><div>The proposed methodology uses electrical spiking activity data from mosquito neurons collected with microelectrode array technology. To assess the method’s performance, data from 0, 1, 2, 3, and 7 days post-infection, reaching more than 15 million samples, were used. In this study, we also look at the influence of days post-infection on recognizing spikes in mosquito neurons.</div><div>Overall, we attempted for the first time to analyze the distinctive pattern in the spike sequence of mosquito neurons using Artificial Intelligence (AI) approaches and to determine the impact of these spikes over time.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"119 ","pages":"Article 109748"},"PeriodicalIF":4.9,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191947","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

CNN–LSTM based deep learning approach for remote photoplethysmography and cardiac activity monitoring leveraging minimal data 基于CNN-LSTM的深度学习方法，利用最小数据进行远程光容积脉搏波和心脏活动监测

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2026-06-15 Epub Date: 2026-02-07 DOI: 10.1016/j.bspc.2026.109787

Saran Zeb , Xiaocong Lian , Wajid Mumtaz , Kegang Wang

Precise heart rate measurement is crucial for assessing an individual’s health, offering valuable insights into their heart condition and cardiovascular activity. Remote photoplethysmography (rPPG) provides a contactless technique for measuring heart rate and monitoring cardiac activity without direct skin contact. This method is especially advantageous in scenarios where direct contact is not feasible or desirable, such as during pandemics to avoid infection risk. Despite significant progress, rPPG still faces challenges, including variations in illumination, motion artifacts, sensor noise, and skin tone variations, which complicate accurate waveform capture. This study proposes a deep neural network model comprising convolutional and LSTM layers to accurately detect the blood volume pulse (BVP) and monitor cardiac activity from RGB and NIR video frames. Notably, the model was trained on single-subject, single-channel data, effectively predicting the PPG waveform and estimating heart rate. Furthermore, the model was trained and tested on videos from four different devices—a webcam, smartphone, RealSense color camera, and NIR camera, ensuring robustness across various sensor types. The model was evaluated on four publicly available datasets—VIPL-HR, MR-NIRP, UBFC-rPPG, and MPSC-rPPG—achieving a mean absolute error (MAE) of 2.64 beats per minute for the same subject and 4.84 beats per minute for different subjects. Achieving robust and accurate heart rate estimation from a single subject across various sensors and challenging scenarios underscores the potential of this contactless model for cardiovascular assessment.

精确的心率测量对于评估个人健康状况至关重要，为了解他们的心脏状况和心血管活动提供了有价值的见解。远程光电容积脉搏波描记（rPPG）提供了一种非接触式技术来测量心率和监测心脏活动，而无需直接接触皮肤。这种方法在不可行或不需要直接接触的情况下特别有利，例如在大流行期间，以避免感染风险。尽管取得了重大进展，但rPPG仍然面临挑战，包括光照变化、运动伪影、传感器噪声和肤色变化，这些都会使准确的波形捕获复杂化。本研究提出了一种包含卷积层和LSTM层的深度神经网络模型，用于准确检测RGB和NIR视频帧的血容量脉冲（BVP）并监测心脏活动。值得注意的是，该模型是在单受试者、单通道数据上训练的，有效地预测了PPG波形并估计了心率。此外，该模型在四种不同设备（网络摄像头、智能手机、RealSense彩色相机和近红外相机）的视频上进行了训练和测试，以确保在各种传感器类型上的稳健性。该模型在四个公开可用的数据集（vipl - hr、MR-NIRP、UBFC-rPPG和mpsc - rppg）上进行了评估，同一受试者的平均绝对误差（MAE）为每分钟2.64次，不同受试者的平均绝对误差为每分钟4.84次。通过各种传感器和具有挑战性的场景，从单个受试者获得稳健和准确的心率估计，强调了这种非接触式心血管评估模型的潜力。

{"title":"CNN–LSTM based deep learning approach for remote photoplethysmography and cardiac activity monitoring leveraging minimal data","authors":"Saran Zeb , Xiaocong Lian , Wajid Mumtaz , Kegang Wang","doi":"10.1016/j.bspc.2026.109787","DOIUrl":"10.1016/j.bspc.2026.109787","url":null,"abstract":"<div><div>Precise heart rate measurement is crucial for assessing an individual’s health, offering valuable insights into their heart condition and cardiovascular activity. Remote photoplethysmography (rPPG) provides a contactless technique for measuring heart rate and monitoring cardiac activity without direct skin contact. This method is especially advantageous in scenarios where direct contact is not feasible or desirable, such as during pandemics to avoid infection risk. Despite significant progress, rPPG still faces challenges, including variations in illumination, motion artifacts, sensor noise, and skin tone variations, which complicate accurate waveform capture. This study proposes a deep neural network model comprising convolutional and LSTM layers to accurately detect the blood volume pulse (BVP) and monitor cardiac activity from RGB and NIR video frames. Notably, the model was trained on single-subject, single-channel data, effectively predicting the PPG waveform and estimating heart rate. Furthermore, the model was trained and tested on videos from four different devices—a webcam, smartphone, RealSense color camera, and NIR camera, ensuring robustness across various sensor types. The model was evaluated on four publicly available datasets—VIPL-HR, MR-NIRP, UBFC-rPPG, and MPSC-rPPG—achieving a mean absolute error (MAE) of 2.64 beats per minute for the same subject and 4.84 beats per minute for different subjects. Achieving robust and accurate heart rate estimation from a single subject across various sensors and challenging scenarios underscores the potential of this contactless model for cardiovascular assessment.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"119 ","pages":"Article 109787"},"PeriodicalIF":4.9,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192798","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

Mathematical modelling of atherogenesis: temperamental endothelial permeability 动脉粥样硬化的数学模型：内皮渗透性的变化

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2026-06-15 Epub Date: 2026-02-10 DOI: 10.1016/j.bspc.2026.109718

Shankar Narayan S , Aishwarya R , Nidhi S Vaishnaw

A significant contributor to the development of atherosclerosis is endothelial dysfunction, which is typified by elevated permeability. In order to understand the intricate interactions among low-density lipoprotein (LDL), cytokines (A), inflammatory immune cells (M), endothelial permeability (E), and vascular remodeling (R), we construct an evolving mathematical model in the present research. Using PID control theory, we introduce a novel approach to modulate endothelial permeability, demonstrating how proportional

(k_{p} = 0.01)

, integral

(k_{i} = 0.001)

, and derivatives

(k_{d} = 0.01)

control terms can stabilize the system and restore endothelial function. Our simulations reveal nonlinear relationships and critical thresholds, where a

10 %

increase in LDL leads to a

25 %

rise in endothelial permeability, highlighting the sensitivity of the endothelium to small changes in LDL levels. Heatmap and other plot analyses further elucidate the system’s dynamics, showing that low levels of LDL (below

2 \times 10^{- 3} g . c m^{- 3}

) and cytokines (below

10^{- 7} g . c m^{- 3}

) are sufficient to induce significant endothelial dysfunction. At higher concentrations, permeability stabilizes near

E \approx 12 \times 10^{- 3} c m^{3} / (g . d a y)

. These findings underscore the importance of early intervention and multi-targeted therapies to mitigate endothelial damage and slow atherosclerosis progression. This study advances our understanding of the molecular mechanisms driving endothelial permeability and provides a computational framework for designing personalised therapeutic strategies.

动脉粥样硬化的一个重要因素是内皮功能障碍，其典型特征是通透性升高。为了了解低密度脂蛋白（LDL）、细胞因子(A)、炎症免疫细胞(M)、内皮通透性(E)和血管重塑(R)之间复杂的相互作用，我们在本研究中构建了一个不断发展的数学模型。利用PID控制理论，我们引入了一种调节内皮通透性的新方法，展示了比例（kp=0.01）、积分（ki=0.001）和导数（kd=0.01）控制项如何稳定系统并恢复内皮功能。我们的模拟揭示了非线性关系和临界阈值，其中LDL增加10%导致内皮通透性增加25%，突出了内皮对LDL水平微小变化的敏感性。热图和其他图分析进一步阐明了系统动力学，显示低水平的LDL（低于2×10-3g.cm-3）和细胞因子（低于10-7g.cm-3）足以诱导显着的内皮功能障碍。在较高浓度下，渗透率稳定在E≈12×10-3cm3/(g.day)附近。这些发现强调了早期干预和多靶向治疗对于减轻内皮损伤和减缓动脉粥样硬化进展的重要性。这项研究促进了我们对驱动内皮细胞渗透性的分子机制的理解，并为设计个性化治疗策略提供了一个计算框架。

{"title":"Mathematical modelling of atherogenesis: temperamental endothelial permeability","authors":"Shankar Narayan S , Aishwarya R , Nidhi S Vaishnaw","doi":"10.1016/j.bspc.2026.109718","DOIUrl":"10.1016/j.bspc.2026.109718","url":null,"abstract":"<div><div>A significant contributor to the development of atherosclerosis is endothelial dysfunction, which is typified by elevated permeability. In order to understand the intricate interactions among low-density lipoprotein (LDL), cytokines (A), inflammatory immune cells (M), endothelial permeability (E), and vascular remodeling (R), we construct an evolving mathematical model in the present research. Using PID control theory, we introduce a novel approach to modulate endothelial permeability, demonstrating how proportional <span><math><mrow><mo>(</mo><msub><mi>k</mi><mi>p</mi></msub><mo>=</mo><mn>0.01</mn><mo>)</mo></mrow></math></span>, integral <span><math><mrow><mo>(</mo><msub><mi>k</mi><mi>i</mi></msub><mo>=</mo><mn>0.001</mn><mo>)</mo></mrow></math></span>, and derivatives <span><math><mrow><mo>(</mo><msub><mi>k</mi><mi>d</mi></msub><mo>=</mo><mn>0.01</mn><mo>)</mo></mrow></math></span> control terms can stabilize the system and restore endothelial function. Our simulations reveal nonlinear relationships and critical thresholds, where a <span><math><mrow><mn>10</mn><mo>%</mo></mrow></math></span> increase in LDL leads to a <span><math><mrow><mn>25</mn><mo>%</mo></mrow></math></span> rise in endothelial permeability, highlighting the sensitivity of the endothelium to small changes in LDL levels. Heatmap and other plot analyses further elucidate the system’s dynamics, showing that low levels of LDL (below <span><math><mrow><mn>2</mn><mo>×</mo><msup><mrow><mn>10</mn></mrow><mrow><mo>-</mo><mn>3</mn></mrow></msup><mi>g</mi><mo>.</mo><mi>c</mi><msup><mrow><mi>m</mi></mrow><mrow><mo>-</mo><mn>3</mn></mrow></msup></mrow></math></span>) and cytokines (below <span><math><mrow><msup><mrow><mn>10</mn></mrow><mrow><mo>-</mo><mn>7</mn></mrow></msup><mi>g</mi><mo>.</mo><mi>c</mi><msup><mrow><mi>m</mi></mrow><mrow><mo>-</mo><mn>3</mn></mrow></msup></mrow></math></span>) are sufficient to induce significant endothelial dysfunction. At higher concentrations, permeability stabilizes near <span><math><mrow><mi>E</mi><mo>≈</mo><mn>12</mn><mo>×</mo><msup><mrow><mn>10</mn></mrow><mrow><mo>-</mo><mn>3</mn></mrow></msup><mi>c</mi><msup><mrow><mi>m</mi></mrow><mn>3</mn></msup><mo>/</mo><mrow><mo>(</mo><mi>g</mi><mo>.</mo><mi>d</mi><mi>a</mi><mi>y</mi><mo>)</mo></mrow></mrow></math></span>. These findings underscore the importance of early intervention and multi-targeted therapies to mitigate endothelial damage and slow atherosclerosis progression. This study advances our understanding of the molecular mechanisms driving endothelial permeability and provides a computational framework for designing personalised therapeutic strategies.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"119 ","pages":"Article 109718"},"PeriodicalIF":4.9,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192856","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

DAAP-NET: Automatic identification and quantitative analysis of gastric wall structure for cancer screening using gastric ultrasound images DAAP-NET：利用胃超声图像对胃壁结构进行自动识别和定量分析，用于肿瘤筛查

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2026-06-15 Epub Date: 2026-02-09 DOI: 10.1016/j.bspc.2026.109648

Mustafain Rehman , Zhijun Liu , Miao Fan , Ahsan Humayun , Mingze Ding , Bin Liu

Objectives

Gastric cancer remains a significant public health challenge worldwide, ranked fifth in incidence and fourth in mortality among all malignant tumors. Early gastric cancer (EGC) detection is critical to improving survival rates.

Methods

We propose a new segmentation model, Dual Attention Atrous Pixel Network (DAAP-Net), designed to predict the gastric wall and stomach cavity in gastric ultrasound images. Post-segmentation, the predicted gastric wall mask extracts a region of interest (ROI) focused on the five-layer wall structure. The ROI undergoes a spatially diffused iterative enhancement (SDIE) technique to suppress intra-layer noise while preserving inter-layer transitions. We apply edge detection to the SDIE-refined ROI and compute layer thickness as pixel distances between successive edges, and normalize them into the proportion vector

x

. A scalar deviation

d

from the normal baseline quantifies abnormality.

Results

DAAP-Net outperforms state-of-the-art segmentation methods, achieving Intersection over Union scores of 0.7720 ± 0.0618 for normal gastric wall, 0.9007 ± 0.0495 for normal stomach cavity, 0.7607 ± 0.0780 for cancer gastric wall, and 0.8843 ± 0.0561 for cancer stomach cavity. Quantitative analysis shows gastric wall layer parameters differ markedly; the edge-derived deviation metric

d

separates cohorts, with normal mean 0.128 and cancer mean 0.508.

Conclusions

Our research highlights structural differences between normal and cancerous gastric walls, providing a reliable and non-invasive method for EGC detection. Current limitations include manual ROI selection and occasional errors in low-contrast regions. Future work includes automated ROI selection, adding a benign-labeled cohort, a multi-center dataset, and improving model accuracy for real-time clinical applications.

目的胃癌仍然是全球重大的公共卫生挑战，在所有恶性肿瘤中发病率排名第五，死亡率排名第四。早期胃癌（EGC）检测是提高生存率的关键。方法提出了一种新的分割模型——双注意像素网络（DAAP-Net），用于预测胃超声图像中的胃壁和胃腔。分割后，预测的胃壁掩膜提取一个关注五层胃壁结构的感兴趣区域（ROI）。ROI采用空间扩散迭代增强（SDIE）技术来抑制层内噪声，同时保持层间过渡。我们将边缘检测应用于sdie细化的ROI，并将层厚度作为连续边缘之间的像素距离计算，并将其归一化为比例向量x。从正常基线的标量偏差d量化异常。ResultsDAAP-Net优于最先进的分割方法,实现十字路口在联盟得分为0.7720±0.0618正常胃壁,0.9007±0.0495正常胃腔,为癌症胃壁0.7607±0.0780,0.8843±0.0561对癌症的胃腔。定量分析显示胃壁各层参数差异明显；边缘衍生偏差度量d分隔队列，正常平均值为0.128，癌症平均值为0.508。结论sour研究突出了正常和癌变胃壁的结构差异，为EGC的检测提供了可靠、无创的方法。目前的限制包括手动ROI选择和在低对比度区域偶尔出现错误。未来的工作包括自动化ROI选择，添加良性标记队列，多中心数据集，以及提高实时临床应用的模型准确性。

{"title":"DAAP-NET: Automatic identification and quantitative analysis of gastric wall structure for cancer screening using gastric ultrasound images","authors":"Mustafain Rehman , Zhijun Liu , Miao Fan , Ahsan Humayun , Mingze Ding , Bin Liu","doi":"10.1016/j.bspc.2026.109648","DOIUrl":"10.1016/j.bspc.2026.109648","url":null,"abstract":"<div><h3>Objectives</h3><div>Gastric cancer remains a significant public health challenge worldwide, ranked fifth in incidence and fourth in mortality among all malignant tumors. Early gastric cancer (EGC) detection is critical to improving survival rates.</div></div><div><h3>Methods</h3><div>We propose a new segmentation model, Dual Attention Atrous Pixel Network (DAAP-Net), designed to predict the gastric wall and stomach cavity in gastric ultrasound images. Post-segmentation, the predicted gastric wall mask extracts a region of interest (ROI) focused on the five-layer wall structure. The ROI undergoes a spatially diffused iterative enhancement (SDIE) technique to suppress intra-layer noise while preserving inter-layer transitions. We apply edge detection to the SDIE-refined ROI and compute layer thickness as pixel distances between successive edges, and normalize them into the proportion vector <span><math><mrow><mtext>x</mtext></mrow></math></span>. A scalar deviation <span><math><mrow><mtext>d</mtext></mrow></math></span> from the normal baseline quantifies abnormality.</div></div><div><h3>Results</h3><div>DAAP-Net outperforms state-of-the-art segmentation methods, achieving Intersection over Union scores of 0.7720 ± 0.0618 for normal gastric wall, 0.9007 ± 0.0495 for normal stomach cavity, 0.7607 ± 0.0780 for cancer gastric wall, and 0.8843 ± 0.0561 for cancer stomach cavity. Quantitative analysis shows gastric wall layer parameters differ markedly; the edge-derived deviation metric <span><math><mrow><mtext>d</mtext></mrow></math></span> separates cohorts, with normal mean 0.128 and cancer mean 0.508.</div></div><div><h3>Conclusions</h3><div>Our research highlights structural differences between normal and cancerous gastric walls, providing a reliable and non-invasive method for EGC detection. Current limitations include manual ROI selection and occasional errors in low-contrast regions. Future work includes automated ROI selection, adding a benign-labeled cohort, a multi-center dataset, and improving model accuracy for real-time clinical applications.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"119 ","pages":"Article 109648"},"PeriodicalIF":4.9,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191948","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

IoT based air quality monitoring and asthma alerts driven by non-crossing quantile regression neural networks 基于物联网的空气质量监测和哮喘警报，由非交叉分位数回归神经网络驱动

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2026-06-15 Epub Date: 2026-02-09 DOI: 10.1016/j.bspc.2026.109565

Abhijit Das , B.M. Chandrakala , N Shobha , J. Reshma , Vikranth Bhoothpur , Rakesh Kumar Godi

Asthma is a chronic respiratory disease that remains difficult to manage due to variable symptoms and diverse environmental triggers. Conventional monitoring approaches often rely on costly equipment and subjective self-reports, limiting timely interventions. Moreover, existing deep learning models suffer from issues like limited data quality, poor handling of outliers and lack of accurate risk assessment. To overcome these complications, IoT Based Air Quality Monitoring and Asthma Alerts Driven by Non-Crossing Quantile Regression Neural Networks (AM-IoT-NCQRNN) is proposed. Initially, the data is collected from Air Quality and Health Impact Dataset. Then the input data is preprocessed under Robust Maximum Correntropy Kalman Filter (RMCKF) to handle missing elements, noise and outliers. RMCKF is for its correntropy-based similarity, offering superior outlier suppression compared to median filters, low-rank imputation and standard Kalman filtering. Afterwards, the preprocessed data is given to the Non-Crossing Quantile Regression Neural Network (NCQRNN) which predicts and classifies health impact scores of asthma as very high, high, moderate, very low and low. NCQRNN applies a non-crossing quantile constraint, ensuring stable and interpretable risk estimation compared to Regression Neural Networks (RNNs) that yield inconsistent boundaries under fluctuating inputs. The proposed approach is implemented as a smartphone application, with real-time data collected through an IoT-based system using a Raspberry Pi and estimated using metrics such as accuracy, precision, recall, f1-score, specificity, ROC and computational time. Finally, the performance of proposed AM-IoT-NCQRNN method attains 19.76%, 24.00% and 19.07% higher accuracy and 29.56%, 24.22% and 28.57% higher precision when compared with existing methods.

哮喘是一种慢性呼吸道疾病，由于各种症状和各种环境触发因素，仍然难以控制。传统的监测方法往往依赖于昂贵的设备和主观的自我报告，限制了及时的干预。此外，现有的深度学习模型还存在数据质量有限、异常值处理差、缺乏准确的风险评估等问题。为了克服这些复杂性，提出了基于物联网的非交叉分位数回归神经网络（AM-IoT-NCQRNN）驱动的空气质量监测和哮喘警报。最初，数据是从空气质量和健康影响数据集收集的。然后在鲁棒最大相关卡尔曼滤波（RMCKF）下对输入数据进行预处理，处理缺失元素、噪声和异常值。RMCKF是基于相关系数的相似性，与中值滤波器、低秩插值和标准卡尔曼滤波相比，它提供了更好的离群值抑制。然后，将预处理后的数据输入到非交叉分位数回归神经网络（NCQRNN）中，对哮喘的健康影响评分进行预测和分类，分为非常高、高、中等、非常低和低。NCQRNN采用非交叉分位数约束，与在波动输入下产生不一致边界的回归神经网络（rnn）相比，确保了稳定和可解释的风险估计。所提出的方法是作为智能手机应用程序实现的，通过使用树莓派的基于物联网的系统收集实时数据，并使用准确度、精密度、召回率、f1评分、特异性、ROC和计算时间等指标进行估计。最后，与现有方法相比，本文提出的AM-IoT-NCQRNN方法的准确率分别提高了19.76%、24.00%和19.07%，精度分别提高了29.56%、24.22%和28.57%。

{"title":"IoT based air quality monitoring and asthma alerts driven by non-crossing quantile regression neural networks","authors":"Abhijit Das , B.M. Chandrakala , N Shobha , J. Reshma , Vikranth Bhoothpur , Rakesh Kumar Godi","doi":"10.1016/j.bspc.2026.109565","DOIUrl":"10.1016/j.bspc.2026.109565","url":null,"abstract":"<div><div>Asthma is a chronic respiratory disease that remains difficult to manage due to variable symptoms and diverse environmental triggers. Conventional monitoring approaches often rely on costly equipment and subjective self-reports, limiting timely interventions. Moreover, existing deep learning models suffer from issues like limited data quality, poor handling of outliers and lack of accurate risk assessment. To overcome these complications, IoT Based Air Quality Monitoring and Asthma Alerts Driven by Non-Crossing Quantile Regression Neural Networks (AM-IoT-NCQRNN) is proposed. Initially, the data is collected from Air Quality and Health Impact Dataset. Then the input data is preprocessed under Robust Maximum Correntropy Kalman Filter (RMCKF) to handle missing elements, noise and outliers. RMCKF is for its correntropy-based similarity, offering superior outlier suppression compared to median filters, low-rank imputation and standard Kalman filtering. Afterwards, the preprocessed data is given to the Non-Crossing Quantile Regression Neural Network (NCQRNN) which predicts and classifies health impact scores of asthma as very high, high, moderate, very low and low. NCQRNN applies a non-crossing quantile constraint, ensuring stable and interpretable risk estimation compared to Regression Neural Networks (RNNs) that yield inconsistent boundaries under fluctuating inputs. The proposed approach is implemented as a smartphone application, with real-time data collected through an IoT-based system using a Raspberry Pi and estimated using metrics such as accuracy, precision, recall, f1-score, specificity, ROC and computational time. Finally, the performance of proposed AM-IoT-NCQRNN method attains 19.76%, 24.00% and 19.07% higher accuracy and 29.56%, 24.22% and 28.57% higher precision when compared with existing methods.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"119 ","pages":"Article 109565"},"PeriodicalIF":4.9,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191955","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

Cortical network dynamics and neural decoding of fine motor complexity via fNIRS and attention-based deep learning 基于fNIRS和基于注意的深度学习的皮层网络动力学和精细运动复杂性的神经解码

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2026-06-15 Epub Date: 2026-02-09 DOI: 10.1016/j.bspc.2026.109758

Xing Ji , Zhong Yin , Yifei Bi , Kaiwei Yu , Yize Li , Jiafa Chen , Dawei Zhang

Fine motor decline serves as a critical early biomarker for neurodegenerative diseases like Parkinson’s disease, making its accurate assessment essential for early detection and intervention. While functional near-infrared spectroscopy (fNIRS) offers a portable, non-invasive neuroimaging solution, the precise cortical dynamics underlying varying levels of motor complexity remain underexplored. This study aims to investigate how fine motor task complexity modulates cortical activation and functional network topology. A secondary objective is to develop and validate a high-performance deep learning model to classify motor complexity levels from fNIRS signals. fNIRS data were recorded from healthy participants performing five fine-motor tasks of increasing complexity, and activation analyses were combined with graph-theoretical metrics to characterize neurophysiological responses. To classify the complexity of fine motor tasks from fNIRS signals, this study developed a bidirectional long short-term memory (Bi-LSTM) model. Performance evaluation used leave-one-out cross-validation, supplemented by multi-seed training to improve robustness. The model achieved an average classification accuracy of 90.67% ± 7.07% (95% CI: ± 2.68%) and an AUC of 0.9720 ± 0.0431, outperforming traditional support vector machine (by 21.3%) and Bi-LSTM (by 10.97%). These results demonstrate the model’s strong generalization across subjects and its ability to capture temporal-spatial patterns of cortical activation associated with increasing task complexity, providing a promising foundation for fine motor decoding and adaptive neurorehabilitation.

精细运动衰退是帕金森病等神经退行性疾病的重要早期生物标志物，对其进行准确评估对于早期发现和干预至关重要。虽然功能性近红外光谱（fNIRS）提供了一种便携式、非侵入性的神经成像解决方案，但不同运动复杂性水平下的精确皮层动力学仍未得到充分探索。本研究旨在探讨精细运动任务复杂性如何调节皮层激活和功能网络拓扑结构。第二个目标是开发和验证高性能深度学习模型，以从近红外光谱信号中分类运动复杂性水平。研究人员记录了健康参与者执行五项复杂精细运动任务时的fNIRS数据，并将激活分析与图形理论指标相结合，以表征神经生理反应。为了从近红外信号中对精细运动任务的复杂性进行分类，本研究建立了双向长短期记忆（Bi-LSTM）模型。性能评估采用留一交叉验证，辅以多种子训练来提高鲁棒性。该模型的平均分类准确率为90.67%±7.07% (95% CI：±2.68%)，AUC为0.9720±0.0431，优于传统的支持向量机（21.3%）和Bi-LSTM（10.97%）。这些结果表明，该模型具有很强的通用性，并且能够捕捉到与任务复杂性增加相关的皮层激活的时空模式，为精细运动解码和适应性神经康复提供了有希望的基础。

{"title":"Cortical network dynamics and neural decoding of fine motor complexity via fNIRS and attention-based deep learning","authors":"Xing Ji , Zhong Yin , Yifei Bi , Kaiwei Yu , Yize Li , Jiafa Chen , Dawei Zhang","doi":"10.1016/j.bspc.2026.109758","DOIUrl":"10.1016/j.bspc.2026.109758","url":null,"abstract":"<div><div>Fine motor decline serves as a critical early biomarker for neurodegenerative diseases like Parkinson’s disease, making its accurate assessment essential for early detection and intervention. While functional near-infrared spectroscopy (fNIRS) offers a portable, non-invasive neuroimaging solution, the precise cortical dynamics underlying varying levels of motor complexity remain underexplored. This study aims to investigate how fine motor task complexity modulates cortical activation and functional network topology. A secondary objective is to develop and validate a high-performance deep learning model to classify motor complexity levels from fNIRS signals. fNIRS data were recorded from healthy participants performing five fine-motor tasks of increasing complexity, and activation analyses were combined with graph-theoretical metrics to characterize neurophysiological responses. To classify the complexity of fine motor tasks from fNIRS signals, this study developed a bidirectional long short-term memory (Bi-LSTM) model. Performance evaluation used leave-one-out cross-validation, supplemented by multi-seed training to improve robustness. The model achieved an average classification accuracy of 90.67% ± 7.07% (95% CI: ± 2.68%) and an AUC of 0.9720 ± 0.0431, outperforming traditional support vector machine (by 21.3%) and Bi-LSTM (by 10.97%). These results demonstrate the model’s strong generalization across subjects and its ability to capture temporal-spatial patterns of cortical activation associated with increasing task complexity, providing a promising foundation for fine motor decoding and adaptive neurorehabilitation.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"119 ","pages":"Article 109758"},"PeriodicalIF":4.9,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191953","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

3D CNN-based method for automatic reorientation of 11C-acetate cardiac PET images using anchor point detection 基于3D cnn的锚点检测11C-acetate心脏PET图像自动重定向方法

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2026-06-15 Epub Date: 2026-02-14 DOI: 10.1016/j.bspc.2026.109814

Shuai Liu , Tan Gong , Ximin Shi , Xue Lin , Ligang Fang , Xiaoying Tang , Fei Shang , Li Huo

Background

Interpreting and diagnosing cardiac PET images in the transaxial plane could complicate image assessment and hinder the detection of perfusion defects. Therefore, reorienting cardiac PET images from the transaxial plane to the short-axis plane is essential.

Purpose

A convolutional neural network (CNN)-based method for anchor point detection was proposed to enable the automatic reorientation of ¹¹C-acetate cardiac PET images.

Methods

A total of 57 subjects who underwent ¹¹C-acetate PET/CT imaging were enrolled in this study. Forty subjects were assigned to the training set, and 17 subjects to the testing set. Three anchor points (the apex of the left ventricle, the center of the left ventricle base and the center of the right ventricle) were manually annotated and used as the gold standard. A 3D CNN incorporating residual modules and fully connected layers was developed to predict the coordinates of three anchor points. A composite loss function was designed to guide the model training.

Results

The predicted coordinates demonstrated a significant correlation with the gold standard (ICCs > 0.75, p < 0.05). Across 17 segments, the average normalized root mean square error (NRMSE) was below 0.082, and the average relative difference was less than 8.69%. No significant differences in pharmacokinetic parameters were observed between manual annotation and the proposed method (all p > 0.05). An NRMSE of 0.053 was achieved on the simulated pseudo image.

Conclusions

The 3D CNN-based method for anchor point detection demonstrated performance comparable to the manual approach, providing a novel and effective solution for the reorientation of ¹¹C-acetate cardiac PET image.

背景：对心脏经轴位PET图像的解读和诊断会使图像评估复杂化，阻碍灌注缺陷的检测。因此，将心脏PET图像从跨轴平面重新定位到短轴平面是必要的。目的提出一种基于卷积神经网络（CNN）的锚点检测方法，实现11C-acetate心脏PET图像的自动重定向。方法对57例经11c -乙酸酯PET/CT扫描的患者进行研究。40名受试者被分配到训练集，17名受试者被分配到测试集。人工标注三个锚点（左心室顶点、左心室基底中心和右心室中心）作为金标准。基于残差模块和全连通层的三维CNN预测了三个锚点的坐标。设计了一个复合损失函数来指导模型的训练。结果预测坐标与金标准具有显著相关性（ICCs >； 0.75,p <； 0.05）。17个区间的平均归一化均方根误差（NRMSE）均小于0.082，平均相对差小于8.69%。手工标注与本方法在药代动力学参数上无显著差异（p均为 >； 0.05）。在模拟的伪图像上实现了0.053的NRMSE。结论基于3D cnn的锚点检测方法具有与人工方法相当的性能，为11C-acetate心脏PET图像的重新定位提供了一种新颖有效的解决方案。

{"title":"3D CNN-based method for automatic reorientation of 11C-acetate cardiac PET images using anchor point detection","authors":"Shuai Liu , Tan Gong , Ximin Shi , Xue Lin , Ligang Fang , Xiaoying Tang , Fei Shang , Li Huo","doi":"10.1016/j.bspc.2026.109814","DOIUrl":"10.1016/j.bspc.2026.109814","url":null,"abstract":"<div><h3>Background</h3><div>Interpreting and diagnosing cardiac PET images in the transaxial plane could complicate image assessment and hinder the detection of perfusion defects. Therefore, reorienting cardiac PET images from the transaxial plane to the short-axis plane is essential.</div></div><div><h3>Purpose</h3><div>A convolutional neural network (CNN)-based method for anchor point detection was proposed to enable the automatic reorientation of <sup>11</sup>C-acetate cardiac PET images.</div></div><div><h3>Methods</h3><div>A total of 57 subjects who underwent <sup>11</sup>C-acetate PET/CT imaging were enrolled in this study. Forty subjects were assigned to the training set, and 17 subjects to the testing set. Three anchor points (the apex of the left ventricle, the center of the left ventricle base and the center of the right ventricle) were manually annotated and used as the gold standard. A 3D CNN incorporating residual modules and fully connected layers was developed to predict the coordinates of three anchor points. A composite loss function was designed to guide the model training.</div></div><div><h3>Results</h3><div>The predicted coordinates demonstrated a significant correlation with the gold standard (ICCs > 0.75, p < 0.05). Across 17 segments, the average normalized root mean square error (NRMSE) was below 0.082, and the average relative difference was less than 8.69%. No significant differences in pharmacokinetic parameters were observed between manual annotation and the proposed method (all p > 0.05). An NRMSE of 0.053 was achieved on the simulated pseudo image.</div></div><div><h3>Conclusions</h3><div>The 3D CNN-based method for anchor point detection demonstrated performance comparable to the manual approach, providing a novel and effective solution for the reorientation of <sup>11</sup>C-acetate cardiac PET image.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"119 ","pages":"Article 109814"},"PeriodicalIF":4.9,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192614","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

Validation of eleven federated learning strategies for multi-contrast image-to-image MRI data synthesis from heterogeneous sources 11种联合学习策略的验证，用于从异构来源合成多对比度图像到图像MRI数据

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2026-06-15 Epub Date: 2026-02-13 DOI: 10.1016/j.bspc.2026.109649

Jan Fiszer , Dominika Ciupek , Maciej Malawski , Tomasz Pieciak

Deep learning (DL)-based image synthesis has recently gained enormous interest in medical imaging, allowing for generating multi-contrast data and therefore, the recovery of missing samples from interrupted or artefact-distorted acquisitions. However, the accuracy of DL models heavily relies on the representativeness of the training datasets naturally characterized by their distributions, experimental setups or preprocessing schemes. These complicate generalizing DL models across multi-site heterogeneous datasets while maintaining the confidentiality of the data. One of the possible solutions is to employ federated learning (FL), which enables the collaborative training of a DL model in a decentralized manner, demanding the involved sites to share only the characteristics of the models without transferring their sensitive medical data. The paper presents a DL-based magnetic resonance (MR) data translation in a FL way. We introduce a new aggregation strategy called FedBAdam that couples two methods with complementary strengths by incorporating momentum in the aggregation scheme and skipping the batch normalization layers. The work comprehensively validates 11 FL-based strategies for an image-to-image multi-contrast MR translation, considering healthy and tumorous brain scans from five different institutions. Our study has revealed that the FedBAdam achieves superior results in terms of mean squared error and structural similarity index compared with standard FL-based aggregation techniques, such as FedAvg or FedProx, and is on par with or superior to personalised methods, while exhibiting more stable convergence in a multi-site, multi-vendor, heterogeneous environment. The FedBAdam has prevented the overfitting of the model and gradually reached the optimal model parameters, exhibiting no oscillations.

基于深度学习（DL）的图像合成最近在医学成像领域引起了极大的兴趣，它允许生成多对比度数据，因此可以从中断或伪影失真的采集中恢复缺失的样本。然而，深度学习模型的准确性在很大程度上依赖于训练数据集的代表性，这些训练数据集的自然特征是它们的分布、实验设置或预处理方案。这使得跨多站点异构数据集泛化DL模型变得复杂，同时还要保持数据的机密性。一种可能的解决方案是采用联邦学习（FL），它能够以分散的方式对DL模型进行协作训练，要求相关站点仅共享模型的特征，而不传输其敏感的医疗数据。本文提出了一种基于dl的磁共振（MR）数据的FL转换方法。我们引入了一种新的聚合策略，称为FedBAdam，它通过在聚合方案中加入动量并跳过批处理规范化层，将两种具有互补优势的方法结合在一起。考虑到来自五个不同机构的健康和肿瘤脑部扫描，该工作全面验证了11种基于fl的图像到图像多对比度MR翻译策略。我们的研究表明，与标准的基于fl的聚合技术（如fedag或FedProx）相比，FedBAdam在均方误差和结构相似性指数方面取得了更好的结果，并且与个性化方法相当或优于个性化方法，同时在多站点，多供应商，异构环境中表现出更稳定的收敛性。FedBAdam防止了模型的过拟合，并逐渐达到最优模型参数，没有振荡。

{"title":"Validation of eleven federated learning strategies for multi-contrast image-to-image MRI data synthesis from heterogeneous sources","authors":"Jan Fiszer , Dominika Ciupek , Maciej Malawski , Tomasz Pieciak","doi":"10.1016/j.bspc.2026.109649","DOIUrl":"10.1016/j.bspc.2026.109649","url":null,"abstract":"<div><div>Deep learning (DL)-based image synthesis has recently gained enormous interest in medical imaging, allowing for generating multi-contrast data and therefore, the recovery of missing samples from interrupted or artefact-distorted acquisitions. However, the accuracy of DL models heavily relies on the representativeness of the training datasets naturally characterized by their distributions, experimental setups or preprocessing schemes. These complicate generalizing DL models across multi-site heterogeneous datasets while maintaining the confidentiality of the data. One of the possible solutions is to employ federated learning (FL), which enables the collaborative training of a DL model in a decentralized manner, demanding the involved sites to share only the characteristics of the models without transferring their sensitive medical data. The paper presents a DL-based magnetic resonance (MR) data translation in a FL way. We introduce a new aggregation strategy called FedBAdam that couples two methods with complementary strengths by incorporating momentum in the aggregation scheme and skipping the batch normalization layers. The work comprehensively validates 11 FL-based strategies for an image-to-image multi-contrast MR translation, considering healthy and tumorous brain scans from five different institutions. Our study has revealed that the FedBAdam achieves superior results in terms of mean squared error and structural similarity index compared with standard FL-based aggregation techniques, such as FedAvg or FedProx, and is on par with or superior to personalised methods, while exhibiting more stable convergence in a multi-site, multi-vendor, heterogeneous environment. The FedBAdam has prevented the overfitting of the model and gradually reached the optimal model parameters, exhibiting no oscillations.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"119 ","pages":"Article 109649"},"PeriodicalIF":4.9,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192791","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

MD-SIRNet: Multi-domain representations for EEG-based speech imagery recognition with deep learning MD-SIRNet：基于脑电图的深度学习语音图像识别的多域表示

IF 4.9 2区医学 Q1 ENGINEERING, BIOMEDICAL

Biomedical Signal Processing and Control

Pub Date : 2026-06-15 Epub Date: 2026-02-11 DOI: 10.1016/j.bspc.2026.109817

Liang Dong , Hengyi Shao , Zhejun Zhang , Yingqi Zhu , Shaoting Guo , Lin Zhang , Lei Li

Speech imagery (SI) recognition from Electroencephalography (EEG), enhances the foundation of the brain-computer interface (BCI). Although some existing research has been proposed to solve the high variability and low signal-to-noise ratio of multi-channel EEG signals, the spatial–temporal-frequency information is still underutilized to improve the performance of SI recognition. We propose MD-SIRNet to obtain multi-domain features efficiently and precisely. MD-SIRNet decomposes spatial multi-channel EEG data into four sets of intrinsic mode functions (IMFs) using Multivariate Variational Mode Decomposition (MVMD). To emphasize the main features, the IMFs are summed and then transformed into time–frequency representation (TFR) images after extracting high-precision time–frequency features using the Synchrosqueezed Wavelet Transform (SSWT). TFR images are fed into the Tuned-CNN model. MD-SIRNet is validated on two publicly available EEG datasets, compared with five methods by accuracy. The results of MD-SIRNet achieve an accuracy improvement of 2.23%, 0.45%, 1.59%, 4.45%, and 5.56% for long words, long-short words, short words, vowels, and command words. The code and model are available at https://github.com/buptantEEG/MD-SIRNet.

从脑电图（EEG）中识别语音图像，增强了脑机接口（BCI）基础。虽然已有研究针对多通道脑电信号的高变异性和低信噪比提出了一些解决方案，但在提高SI识别性能方面，仍未充分利用脑电信号的时空频率信息。为了高效、精确地获取多域特征，我们提出了MD-SIRNet。MD-SIRNet利用多变量变分模态分解（Multivariate Variational mode Decomposition， MVMD）将空间多通道脑电数据分解为四组固有模态函数（IMFs）。为了突出图像的主要特征，利用同步压缩小波变换（SSWT）提取高精度时频特征后，对图像进行汇总，并将其转换为时频表示（TFR）图像。TFR图像被输入到调谐cnn模型中。MD-SIRNet在两个公开可用的EEG数据集上进行了验证，并与五种方法进行了准确率比较。MD-SIRNet对长词、长短词、短词、元音和命令词的准确率分别提高了2.23%、0.45%、1.59%、4.45%和5.56%。代码和模型可在https://github.com/buptantEEG/MD-SIRNet上获得。

{"title":"MD-SIRNet: Multi-domain representations for EEG-based speech imagery recognition with deep learning","authors":"Liang Dong , Hengyi Shao , Zhejun Zhang , Yingqi Zhu , Shaoting Guo , Lin Zhang , Lei Li","doi":"10.1016/j.bspc.2026.109817","DOIUrl":"10.1016/j.bspc.2026.109817","url":null,"abstract":"<div><div>Speech imagery (SI) recognition from Electroencephalography (EEG), enhances the foundation of the brain-computer interface (BCI). Although some existing research has been proposed to solve the high variability and low signal-to-noise ratio of multi-channel EEG signals, the spatial–temporal-frequency information is still underutilized to improve the performance of SI recognition. We propose MD-SIRNet to obtain multi-domain features efficiently and precisely. MD-SIRNet decomposes spatial multi-channel EEG data into four sets of intrinsic mode functions (IMFs) using Multivariate Variational Mode Decomposition (MVMD). To emphasize the main features, the IMFs are summed and then transformed into time–frequency representation (TFR) images after extracting high-precision time–frequency features using the Synchrosqueezed Wavelet Transform (SSWT). TFR images are fed into the Tuned-CNN model. MD-SIRNet is validated on two publicly available EEG datasets, compared with five methods by accuracy. The results of MD-SIRNet achieve an accuracy improvement of 2.23%, 0.45%, 1.59%, 4.45%, and 5.56% for long words, long-short words, short words, vowels, and command words. The code and model are available at <span><span>https://github.com/buptantEEG/MD-SIRNet</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"119 ","pages":"Article 109817"},"PeriodicalIF":4.9,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192797","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