Biomedical Signal Processing and Control最新文献_第8页

LNDCNN: A lung nodule detection model based on improved YOLOv7 ndcnn：一种基于改进YOLOv7的肺结节检测模型

IF 4.9 2区医学

Biomedical Signal Processing and Control Pub Date : 2025-10-08 DOI: 10.1016/j.bspc.2025.108754

Lu Meng , Lijun Zhou

{"title":"LNDCNN: A lung nodule detection model based on improved YOLOv7","authors":"Lu Meng , Lijun Zhou","doi":"10.1016/j.bspc.2025.108754","DOIUrl":"10.1016/j.bspc.2025.108754","url":null,"abstract":"<div><div>The automatic detection of lung nodules is crucial for cancer screening. However, lung nodules typically occupy small areas in CT images with indistinct features, posing significant detection challenges. To address this, YOLOv7 is optimized and a lung nodule detection model is proposed. Firstly, considering the problem of feature loss caused by the PAFPN (Path Aggregation Feature Pyramid Network), an Enhanced Feature Pyramid Network (E_FPN) is introduced, which employs dilated convolutions in series to adjust the receptive field size and supplement contextual information. The dilation rate is designed to fully utilize fine-grained features. Then, an adaptive fusion strategy is proposed, enabling each detection head to adaptively integrate information from other detection heads, thereby achieving complementary utilization of detection head information. Finally, considering that the IoU (Intersection over Union) metric is very sensitive to the position change of small objects, the Gaussian Wasserstein distance metric is adopted to optimize the loss function and enhance target localization accuracy. Experiments are primarily conducted on the LUNA16 dataset. The optimized model, LNDCNN (Lung Nodule Detection Convolutional Neural Network), achieves an improvement of 11.42% in mAP<sub>50</sub>, 6.17% in mAP<span><math><msub><mrow></mrow><mrow><mn>50</mn><mo>:</mo><mn>90</mn></mrow></msub></math></span>, and a 14.4% increase in Recall compared to YOLOv7. At the same time, we validate the algorithm’s generalization ability on the ALIBABA TianChi dataset. LNDCNN demonstrates consistent improvements with 5.68% higher mAP<sub>50</sub>, 2.92% higher mAP<span><math><msub><mrow></mrow><mrow><mn>50</mn><mo>:</mo><mn>90</mn></mrow></msub></math></span>, and 8.61% higher Recall compared to YOLOv7, demonstrating promising generalization capability. The implementation of LNDCNN is publicly available at <span><span>https://github.com/zlj-zlj/LNDCNN</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108754"},"PeriodicalIF":4.9,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145265313","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

CBAM-LSTM-attention enabled human emotion recognition using EEG signals 基于cbam - lstm -注意力的人类情绪识别

IF 4.9 2区医学

Biomedical Signal Processing and Control Pub Date : 2025-10-08 DOI: 10.1016/j.bspc.2025.108767

Jingqi Le , Yanghui Wang , Yong Zhou , Sheng Zou

{"title":"CBAM-LSTM-attention enabled human emotion recognition using EEG signals","authors":"Jingqi Le , Yanghui Wang , Yong Zhou , Sheng Zou","doi":"10.1016/j.bspc.2025.108767","DOIUrl":"10.1016/j.bspc.2025.108767","url":null,"abstract":"<div><div>Human emotion recognition seeks to facilitate machines in comprehending human emotional states. EEG signals, being non-invasive measurements, are among the signals that most accurately represent human emotions. Nevertheless, the abundance of channels in EEG signals leads to heightened computational complexity and a greater risk of overfitting the model. In this research, eXtreme Gradient Boosting (XGBoost) was employed to choose the suitable EEG channel, and a unique model named convolutional block attention module − long short-term memory − attention module (CBAM-LSTM-Attention) was introduced for emotion recognition. The model combines residual blocks, LSTM networks, and attention mechanisms to efficiently incorporate important channel-spatial and temporal domain features of EEG signals. To achieve high prediction accuracy in emotion recognition, the following innovations can be implemented: (i) Integrating XGBoost algorithm to evaluate each channel based on power spectral density (PSD), and selecting the channel with the highest score as the input for the model. This approach reduces the computational complexity of the model and minimises the risk of overfitting. (ii) Introducing a channel-spatial attention module in the residual block to enhance the model’s ability to extract channel-spatial domain features in the convolutional block attention module (CBAM) model. (iii) Utilising a multi-head attention mechanism to improve the model’s capability to extract temporal domain features, enabling global feature perception and input to the LSTM layer for decoding. The results indicated that the proposed CBAM-LSTM-Attention model achieved 95.108 % accuracy for arousal and 94.862 % for valence on the DEAP dataset using single-channel data. Using multi-channel data, the model achieved 98.790 % for arousal and 97.249 %for valence. This suggests that the model effectively enables correct classification of human emotion recognition.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108767"},"PeriodicalIF":4.9,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145265546","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

Style transformation and distance map guided nucleus instance segmentation via multi task learning 基于多任务学习的风格转换和距离图引导的核实例分割

IF 4.9 2区医学

Biomedical Signal Processing and Control Pub Date : 2025-10-08 DOI: 10.1016/j.bspc.2025.108724

Deboch Eyob Abera , Nazar Zaki , Wenjian Qin

{"title":"Style transformation and distance map guided nucleus instance segmentation via multi task learning","authors":"Deboch Eyob Abera , Nazar Zaki , Wenjian Qin","doi":"10.1016/j.bspc.2025.108724","DOIUrl":"10.1016/j.bspc.2025.108724","url":null,"abstract":"<div><div>Instance segmentation of nuclei in histopathological images is hindered by three critical challenges: overlapping nuclei, domain shift caused by staining variability, and generalization across diverse multi-organ datasets. To address these issues, we propose a unified multi-task learning framework for nucleus instance segmentation that integrates style transformation and distance map-guided segmentation. Our architecture employs multi-dilated residual blocks and encoder–decoder attention gates to capture multi-scale features and preserve fine nuclear details, while a transformer in the bottleneck enhances contextual understanding and models long-range dependencies. The network incorporates dual heads for semantic segmentation and distance-map prediction, effectively addressing overlapping nuclei. Additionally, a histogram-based, reference-guided stain normalization module mitigates domain shift caused by staining variability, and when combined with our robust model architecture, it enhances the overall generalization ability across multi-organ datasets. Experimental results demonstrate our method’s superior performance over existing segmentation approaches. The source code is available at <span><span>https://github.com/eyob12/MTL-NucleusSeg</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108724"},"PeriodicalIF":4.9,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145265724","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

Malignant ventricular arrhythmias prediction based on cardiorespiratory signals analysis and machine learning approach 基于心肺信号分析和机器学习方法的恶性室性心律失常预测

IF 4.9 2区医学

Biomedical Signal Processing and Control Pub Date : 2025-10-08 DOI: 10.1016/j.bspc.2025.108743

Nikola N. Radovanović , Mirjana M. Platiša , Nadica Miljković

{"title":"Malignant ventricular arrhythmias prediction based on cardiorespiratory signals analysis and machine learning approach","authors":"Nikola N. Radovanović , Mirjana M. Platiša , Nadica Miljković","doi":"10.1016/j.bspc.2025.108743","DOIUrl":"10.1016/j.bspc.2025.108743","url":null,"abstract":"<div><div>This study aims to evaluate the potential of machine learning (ML) algorithms in predicting malignant ventricular arrhythmias based on RR intervals and respiratory signals in heart failure patients. A total of 26 distinct features are extracted from both signals and systematically categorized into three groups: cardiac, respiratory, and their interaction-based metrics. The data comprise 82 heart failure patients, all of whom underwent screening prior to cardioverter defibrillator implantation, ensuring a clinically relevant population for evaluation. Despite the absence of statistically significant differences in features identified by traditional analysis, the integration of advanced ML models, specifically Random Forest (RF) and eXtreme Gradient Boosting (XGBoost) ML algorithms, confirms a separation between patient groups. Moreover, the valuable application of Random Over-Sampling Examples (ROSE) as a data augmentation technique is successful in addressing limitations inherent in the small and imbalanced dataset. Our results suggest that the ML algorithms can detect subtle, nonlinear relationships embedded within the complex interplay of cardiac and respiratory dynamics. The best predictive outcomes are achieved through a strategic combination of pseudo-random oversampling and undersampling, as well as through application of ensemble learning algorithms. These findings underscore the promise of ML approaches—not only in improving the predictive potential of cardiorespiratory features, but also in illuminating the critical role of respiratory features and cardiorespiratory interactions in the early identification of malignant ventricular arrhythmias among heart failure patients.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108743"},"PeriodicalIF":4.9,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145265726","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

Multi-scale network for medical image segmentation integrated with edge perception 融合边缘感知的医学图像分割多尺度网络

IF 4.9 2区医学

Biomedical Signal Processing and Control Pub Date : 2025-10-08 DOI: 10.1016/j.bspc.2025.108820

Mingjun Wei , Qian Wu , Jinghao Jia , Weibin Chen , Ao Cai , Hui Li , Xiaochuan Sun , Jinyun Liu

{"title":"Multi-scale network for medical image segmentation integrated with edge perception","authors":"Mingjun Wei , Qian Wu , Jinghao Jia , Weibin Chen , Ao Cai , Hui Li , Xiaochuan Sun , Jinyun Liu","doi":"10.1016/j.bspc.2025.108820","DOIUrl":"10.1016/j.bspc.2025.108820","url":null,"abstract":"<div><div>Precise medical image segmentation plays a crucial role in early disease diagnosis, yet existing methods struggle with complex backgrounds and ambiguous boundaries. To overcome these issues, a multi-scale network integrated with edge perception (MENet) is proposed in this paper. Firstly, an edge-related module is introduced to extract and feedback edge features, enhancing the overall feature representation. Secondly, a frequency-domain enhancement module is developed to dynamically amplify critical frequency bands, improving lesion morphology modeling while preserving global contextual representations. Thirdly, a multi-scale feature fusion module is constructed to achieve effective integration of features across different levels by leveraging the channel attention mechanism. Finally, a multi-scale aggregation loss function is designed to supervise segmentation and edge detection tasks. Experiments are conducted on Synapse, ACDC, CVC-ClinicDB and BUSI datasets. MENet achieves 84.36%, 92.40%, 95.34% and 81.24% on mDice individually. HD95 is 14.75 mm on Synapse. mIoU is 91.23% on CVC-ClinicDB and 72.64% on BUSI. It can be demonstrated that MENet consistently outperforms traditional models, baseline variants, and recent methods in terms of segmentation accuracy.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108820"},"PeriodicalIF":4.9,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145265950","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

Multi-stage cascaded refinement with wavelet downsampling for retinal vessel segmentation 基于小波下采样的多级级联细化视网膜血管分割

IF 4.9 2区医学

Biomedical Signal Processing and Control Pub Date : 2025-10-08 DOI: 10.1016/j.bspc.2025.108824

Zikun Ling , Jiaojiao Yu , Qiankun Zuo , Bangjun Lei

{"title":"Multi-stage cascaded refinement with wavelet downsampling for retinal vessel segmentation","authors":"Zikun Ling , Jiaojiao Yu , Qiankun Zuo , Bangjun Lei","doi":"10.1016/j.bspc.2025.108824","DOIUrl":"10.1016/j.bspc.2025.108824","url":null,"abstract":"<div><div>The morphological changes of retinal vessels play a crucial role in assisting doctors with the diagnosis of ocular and cardiovascular diseases. Retinal vessels exhibit complex and variable shapes, and current vessel segmentation methods are ineffective at capturing the features of small vessels of varying sizes. This results in difficulties in segmenting small vessels and causes vessel segmentation to suffer from discontinuities. To address this challenge, we propose a multi-stage cascaded refinement with wavelet downsampling for retinal vessel segmentation. In our network, we introduce a multi-stage cascaded structure, which first employs multi-scale feature fusion in the early stages to extract vessel shape representations of different sizes, thereby enhancing the model’s ability to capture small vessel features. To further refine the feature representation, we embed a feature refinement module at the bottom of the network, utilizing a self-attention mechanism to capture the long-range distribution continuity of the vessels. This mechanism also helps to reduce information redundancy in densely distributed vessels. Additionally, we employ wavelet downsampling as the downsampling layer in the encoder, which effectively minimizes the loss of vessel detail information during the downsampling process.Experimental results on the public datasets DRIVE, CHASE_DB1, and STARE show that the proposed method achieves AUC scores of 0.9891, 0.9910, and 0.9928, and accuracy scores of 0.9707, 0.9764, and 0.9783, respectively. These results demonstrate the superiority of our method in retinal vessel segmentation, which can significantly assist in the early diagnosis and monitoring of ocular and cardiovascular diseases.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108824"},"PeriodicalIF":4.9,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145265443","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

MedVKAN: Efficient feature extraction with Mamba and KAN for medical image segmentation MedVKAN：使用曼巴和KAN进行医学图像分割的高效特征提取

IF 4.9 2区医学

Biomedical Signal Processing and Control Pub Date : 2025-10-08 DOI: 10.1016/j.bspc.2025.108821

Hancan Zhu, Jinhao Chen, Guanghua He

{"title":"MedVKAN: Efficient feature extraction with Mamba and KAN for medical image segmentation","authors":"Hancan Zhu, Jinhao Chen, Guanghua He","doi":"10.1016/j.bspc.2025.108821","DOIUrl":"10.1016/j.bspc.2025.108821","url":null,"abstract":"<div><div>Medical image segmentation has traditionally relied on convolutional neural networks (CNNs) and Transformer-based models. CNNs, however, are constrained by limited receptive fields, while Transformers face scalability challenges due to quadratic computational complexity. To overcome these issues, recent studies have explored alternative architectures. The Mamba model, a selective state-space design, achieves near-linear complexity and effectively captures long-range dependencies. Its vision-oriented variant, the Visual State Space (VSS) model, extends these strengths to image feature learning. In parallel, the Kolmogorov-Arnold Network (KAN) enhances nonlinear expressiveness by replacing fixed activation functions with learnable ones. Motivated by these advances, we propose the VSS-Enhanced KAN (VKAN) module, which integrates VSS with the Expanded Field Convolutional KAN (EFC-KAN) as a replacement for Transformer modules, thereby strengthening feature extraction. We further embed VKAN into a U-Net framework, resulting in MedVKAN, an efficient medical image segmentation model. Extensive experiments on five public datasets demonstrate that MedVKAN achieves state-of-the-art performance on four datasets and ranks second on the remaining one. These results underscore the effectiveness of combining Mamba and KAN while introducing a novel and computationally efficient feature extraction framework. The source code is available at: <span><span>https://github.com/beginner-cjh/MedVKAN</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108821"},"PeriodicalIF":4.9,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145265037","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 dual encoder–decoder multi-task 3D deep learning framework for the segmentation of focal cortical dysplasia lesions 用于局灶性皮质发育不良病变分割的双编码器-解码器多任务3D深度学习框架

IF 4.9 2区医学

Biomedical Signal Processing and Control Pub Date : 2025-10-08 DOI: 10.1016/j.bspc.2025.108764

S. Niyas , Chandrasekharan Kesavadas , Jeny Rajan

{"title":"A dual encoder–decoder multi-task 3D deep learning framework for the segmentation of focal cortical dysplasia lesions","authors":"S. Niyas , Chandrasekharan Kesavadas , Jeny Rajan","doi":"10.1016/j.bspc.2025.108764","DOIUrl":"10.1016/j.bspc.2025.108764","url":null,"abstract":"<div><div>Focal cortical dysplasia (FCD) is a congenital malformation of brain development that is the most common cause of intractable epilepsy in adults and children. Automating the identification and segmentation of FCD lesions from magnetic resonance imaging (MRI) volumes is useful for neuroradiologists in pre-surgical evaluations. The prevailing methods in FCD segmentation using two-dimensional (2D) convolutional neural networks (CNNs) largely overlook the potential of utilizing three-dimensional (3D) MRI volumes, thus neglecting the valuable inter-slice information inherent in the MRI volumes. We propose a novel 3D deep learning model employing a multi-view dual encoder–decoder architecture to precisely segment FCD lesions within MRI volumes. Our approach is based on a 3D CNN framework with integrated residual connections, serving as the backbone for the segmentation network. The model also incorporates various architecture-wise enhancements. Firstly, we integrate multi-view training, a concept drawn from the methodology employed by neuro-radiologists when examining 3D MRI volumes. Here, the model processes fluid-attenuated inversion recovery (FLAIR) MRI volumes and their corresponding cortical thickness maps. This information is channeled through a dual-encoder network, wherein the individual encoders are interlinked through a 3D attention mechanism. Additionally, the model implements a dual-decoder stage to facilitate dual-task learning, leveraging the distance map derived from the ground truth data. The model achieved Dice similarity coefficient (DSC) that were 4.8% and 2.3% higher compared to state-of-the-art 2D and 3D FCD segmentation methods, respectively.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108764"},"PeriodicalIF":4.9,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145265951","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

Enhancing the efficacy of hormone therapy in prostate cancer through Conditional Super-Twisting Sliding Mode Control and Redfox optimization 通过条件超扭滑模控制和红狐优化提高前列腺癌激素治疗的疗效

IF 4.9 2区医学

Biomedical Signal Processing and Control Pub Date : 2025-10-08 DOI: 10.1016/j.bspc.2025.108895

Atif Rehman , Rimsha Ghias , Nadia Sultan , Hammad Iqbal Sherazi , Sarra Ayouni , Sohail Khalid , Mujeeb Ur Rehman

{"title":"Enhancing the efficacy of hormone therapy in prostate cancer through Conditional Super-Twisting Sliding Mode Control and Redfox optimization","authors":"Atif Rehman , Rimsha Ghias , Nadia Sultan , Hammad Iqbal Sherazi , Sarra Ayouni , Sohail Khalid , Mujeeb Ur Rehman","doi":"10.1016/j.bspc.2025.108895","DOIUrl":"10.1016/j.bspc.2025.108895","url":null,"abstract":"<div><div>Tumor recurrence remains a significant challenge in prostate cancer hormone therapy, particularly because of the regeneration of androgen-independent cells following prolonged androgen deprivation. Intermittent androgen suppression has been suggested to delay relapse while minimizing side effects, thereby improving patients’ quality of life. This study compares several control strategies, including the Sliding Mode Control (SMC), Integral Sliding Mode Control (ISMC), and Conditional Super-Twisting Sliding Mode Control (CSTSMC), along with a novel Redfox optimization technique for parameter tuning. This study aimed to evaluate the potential of these strategies to enhance the effectiveness of hormone therapy in controlling tumor relapse. Our analysis reveals that the CSTSMC outperforms other control methods in terms of robustness, tracking accuracy, and adaptability, making it the most effective option for this application. Despite the benefits of controller-based tumor cell reduction, treatment resistance remains a major concern, underscoring the importance of precise drug scheduling. To enhance the controller performance, parameter tuning was performed using the Redfox optimization algorithm with the integral of time absolute error serving as the objective function. The system stability was ensured using a Lyapunov-based theoretical framework. The effectiveness of the designed controllers was assessed using MATLAB/Simulink simulations, followed by real-time validation through a hardware-in-the-loop setup employing the C2000 Delfino™ microcontroller and F28379D Launchpad board.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108895"},"PeriodicalIF":4.9,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145265314","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

ULD-Net: A U-shaped branch large kernel depthwise convolution volume network for 3D medical image segmentation ULD-Net：一种用于三维医学图像分割的u形分支大核深度卷积体网络

IF 4.9 2区医学

Biomedical Signal Processing and Control Pub Date : 2025-10-07 DOI: 10.1016/j.bspc.2025.108746

Weisheng Li, JunTong Ci, Feiyan Li, Guofeng Zeng, Zhaopeng Huang

{"title":"ULD-Net: A U-shaped branch large kernel depthwise convolution volume network for 3D medical image segmentation","authors":"Weisheng Li, JunTong Ci, Feiyan Li, Guofeng Zeng, Zhaopeng Huang","doi":"10.1016/j.bspc.2025.108746","DOIUrl":"10.1016/j.bspc.2025.108746","url":null,"abstract":"<div><div>Recently, the hierarchical Transformers model represented by Swin UNETR has achieved the most advanced performance in the field of 3D medical image segmentation. This improvement largely relies on the inherent advantages of Transformers such as large receptive fields, while the inductive bias inherent in convolution has not been fully utilized. Therefore, problems such as missing organ boundaries and incorrect organ types are prone to occur during segmentation. We found that large kernel (LK) depthwise convolution can not only simulate these characteristics of Transformers, but also solve the above problems to some extent. In this work, we propose a 3D medical image segmentation network ULD-Net, which simulates large receptive fields through LK depthwise convolution for robust volume segmentation. And in order to accurately and exhaustively obtain features in volume segmentation, we simultaneously used depthwise convolutions with different kernel sizes and utilized branch structures to balance them. Furthermore, we will improve the sparse MLP (sMLP) applied to 2D image recognition to 3D ULD sMLP (UMLP). Use UMLP with fewer parameters and better performance to replace MLP with feature scaling in the Swin Transformer block. At the same time, we have also made some minor adjustments and improvements in the micro design. On the BTCV, FLARE 2021 and AMOS 2022 abdominal multi-organ data sets, ULD-Net outperforms existing SOTA models.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108746"},"PeriodicalIF":4.9,"publicationDate":"2025-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145265277","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