IEEE Transactions on Biomedical Engineering最新文献

{"title":"Echo Flow-Induced Temporal Correlation Learning for Ultrasound Video Object Segmentation.","authors":"Dongfang Wang, Tao Zhou, Shangbing Gao, Jian Yang","doi":"10.1109/TBME.2025.3594704","DOIUrl":"https://doi.org/10.1109/TBME.2025.3594704","url":null,"abstract":"<p><strong>Objective: </strong>The segmentation of ultrasound video objects aims to delineate specific anatomical structures or areas of injury in sequential ultrasound imaging data. Current methods exhibit promising results, but struggle with key aspects of ultrasound video analysis. They insufficiently capture inter-frame object motion, resulting in unsatisfactory segmentation for dynamic or low-contrast scenarios. With the release of SAM2, video object segmentation has advanced significantly. However, its performance in ultrasound videos remains suboptimal due to its design bias toward natural videos and lack of consideration for ultrasound-specific characteristics. We propose a novel EchoSAM2 method to achieve more accurate object segmentation in ultrasound videos.</p><p><strong>Methods: </strong>We propose Echo Flow, which captures motion trends between frames to enhance the modeling of temporal relationships. It also helps suppress interference from non-object regions by leveraging object motion patterns. Furthermore, we propose an Echo Modulation Block (EMB) to seamlessly incorporate Echo Flow into the SAM2 framework, improving the quality of feature representation. To further optimize SAM2's performance during fine-tuning, we present a Gaussian Adapter specifically designed for ultrasound image characteristics.</p><p><strong>Results: </strong>Extensive experiments on three ultrasound video datasets confirm the effectiveness of our method, achieving state-of-the-art results. On the EUDP dataset, our model achieves a Dice of 85.49%, outperforming the second-best method by 3.19%. Models trained on HMC-QU and CAMUS achieve the best generalization when tested on each other's unseen test sets.</p><p><strong>Conclusion: </strong>The introduction of Echo Flow, along with other supporting modules, enhances both segmentation accuracy and the model's generalizability.</p><p><strong>Significance: </strong>Accurate segmentation of ultrasound video objects enhances diagnostic accuracy and consistency, thereby increasing overall clinical value.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144834991","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}

{"title":"Adversarial Debiasing for Equitable and Fair Detection of Acute Coronary Syndrome using 12-Lead ECG.","authors":"Rui Qi Ji, Nathan T Riek, Zeineb Bouzid, Karina Kraevsky-Phillips, Tanmay Gokhale, Jessica K Zegre-Hemsey, Gilles Clermont, Samir Saba, Christian Martin-Gill, Clifton W Callaway, Murat Akcakaya, Ervin Sejdic, Salah S Al-Zaiti","doi":"10.1109/TBME.2025.3597527","DOIUrl":"https://doi.org/10.1109/TBME.2025.3597527","url":null,"abstract":"<p><strong>Objective: </strong>Acute coronary syndrome (ACS) is a life-threatening condition requiring accurate diagnosis for better outcomes. However, variability in signs and symptoms among racial subgroups could cause disparities in diagnostic accuracy. In this study, we use machine learning models to diagnose ACS, focusing on mitigating disparities and ensuring fairness between Black and non-Black populations.</p><p><strong>Methods: </strong>We built on a state-of-the-art random forest classifier to compare three mitigation strategies. The first two approaches involved resampling or partitioning the data prior to training, while the third approach proposed an innovative framework called adversarial debiasing. To evaluate our model performance, we used the receiver operating characteristic (ROC) curve and an operating point at 80% specificity for clinical importance.</p><p><strong>Results: </strong>After mitigation with adversarial debiasing, the difference in sensitivities between the two subgroups decreased from 9.8% to 1.3%. Specifically, this approach achieved areas under the ROC of 0.810 and 0.817, and sensitivities of 70.1% and 71.4%, respectively for Black and non-Black subgroups.</p><p><strong>Conclusion: </strong>The proposed adversarial debiasing model outperformed the other two methods in both diagnostic accuracy and effectiveness in minimizing disparities.</p><p><strong>Significance: </strong>We expect this framework to achieve fair diagnostic models across diverse demographic populations globally and be generalizable to other outcomes.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144821361","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}

{"title":"ChatRadio-Valuer: A Chat Large Language Model for Generalizable Radiology Impression Generation on Multi-institution and Multi-system Data.","authors":"Tianyang Zhong, Wei Zhao, Yutong Zhang, Yi Pan, Peixin Dong, Zuowei Jiang, Hanqi Jiang, Yifan Zhou, Xiaoyan Kui, Youlan Shang, Lin Zhao, Li Yang, Yaonai Wei, Zhuoyi Li, Jiadong Zhang, Longtao Yang, Hao Chen, Huan Zhao, Yuxiao Liu, Ning Zhu, Yiwei Li, Yisong Wang, Jiaqi Yao, Jiaqi Wang, Ying Zeng, Lei He, Chao Zheng, Zhixue Zhang, Ming Li, Zhengliang Liu, Haixing Dai, Zihao Wu, Lu Zhang, Shu Zhang, Xiaoyan Cai, Xintao Hu, Shijie Zhao, Xi Jiang, Xin Zhang, Wei Liu, Xiang Li, Dajiang Zhu, Lei Guo, Dinggang Shen, Junwei Han, Tianming Liu, Jun Liu, Tuo Zhang","doi":"10.1109/TBME.2025.3597325","DOIUrl":"https://doi.org/10.1109/TBME.2025.3597325","url":null,"abstract":"<p><p>Achieving clinical level performance and widespread deployment for generating radiology impressions encounters a giant challenge for conventional artificial intelligence models tailored to specific diseases and organs. Concurrent with the increasing accessibility of radiology reports and advancements in modern general AI techniques, the emergence and potential of deployable radiology AI exploration have been bolstered. Here, we present ChatRadio-Valuer, the first general radiology diagnosis large language model for localized deployment within hospitals and being close to clinical use for multi-institution and multi-system diseases. ChatRadio-Valuer achieved 15 state-of-the-art results across five human systems and six institutions in clinical-level events (n=332,673) through rigorous and full-spectrum assessment, including engineering metrics, clinical validation, and efficiency evaluation. Notably, it exceeded OpenAI's GPT-3.5 and GPT-4 models, achieving superior performance in comprehensive disease diagnosis compared to the average level of radiology experts. Besides, ChatRadio-Valuer supports zero-shot transfer learning, greatly boosting its effectiveness as a radiology assistant, while ensuring adherence to privacy standards and being readily utilized for large-scale patient populations. Our expeditions suggest the development of localized LLMs would become an imperative avenue in hospital applications.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144821362","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}

{"title":"Zero-Shot Enhancement with Cross-Modal Applicability for Low-Light Vis-μOCT Images.","authors":"Shujiang Chen, Yanshuo Li, Hua Wei, Fuwang Wu, Weiye Song","doi":"10.1109/TBME.2025.3597643","DOIUrl":"https://doi.org/10.1109/TBME.2025.3597643","url":null,"abstract":"<p><strong>Objective: </strong>Optical coherence tomography (OCT) is a rapid and non-destructive imaging technique, but image brightness decreases when imaging deep tissues or under low power and short exposure due to insufficient backscattered light. This issue is more pronounced in visible-light micro-OCT (vis-μOCT), where shorter wavelengths increase scattering and limit penetration, restricting its application.</p><p><strong>Method: </strong>In this paper, we propose DifNIR, a novel framework for enhancing low-light OCT images. The framework begins with a preliminary denoising stage. Image enhancement is then performed using a neural implicit representation (NIR) network, in which pixel values are incorporated as auxiliary input to mitigate the oversmoothing effect of fully connected layers. To enable unsupervised learning, customdesigned loss functions is employed. The proposed method is validated through qualitative and quantitative comparisons on a self-collected en face image dataset. To further assess its generalizability, we also performed experiments on B-scan images and retinal images acquired from other OCT devices.</p><p><strong>Result: </strong>On the en face image dataset, Dif-NIR outperforms existing methods in terms of visual quality, SNR (58.99 dB), CNR (49.56 dB), and NIQE (9.0553). It also effectively generalizes to OCT B-scan images and retinal images acquired by other devices.</p><p><strong>Conclusion: </strong>The proposed network effectively mitigates unpredictable brightness degradation, producing clearer and better-illuminated images while exhibiting strong generalization capability.</p><p><strong>Significance: </strong>The network effectively reveals deeplayer information in OCT images and can be applied to expand its usage scenarios to cost-effective and high-speed imaging settings.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144821363","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}

{"title":"Hybrid Residual Neural Network for Obstructive Sleep Apnea Detection Using ECG Scalogram.","authors":"Xielan Tang, Bin Zhang, Yamei Li, Ming Guan, Nian Xiong, Hung-Chun Li, Mikhail Poluektov, Xuesheng Ma, Yuan Zhang, Mang I Vai","doi":"10.1109/TBME.2025.3597083","DOIUrl":"https://doi.org/10.1109/TBME.2025.3597083","url":null,"abstract":"<p><p>The electrocardiogram (ECG) has emerged as a viable alternative to polysomnography (PSG) for the detection of obstructive sleep apnea (OSA). Given the limited feature information of ECG signals in a single domain, this study proposes an adaptive threshold denoising synchrosqueezed wavelet transform (ADSWT) algorithm to extract high-resolution time-frequency domain features of ECG signals. The ADSWT algorithm enhances the time-frequency resolution, enabling better extraction of key features while minimizing noise interference, and generates an ECG scalogram as input for the network. Additionally, we propose a hybrid residual neural network (HRN-Net) for the automatic classification of OSA. The HRN-Net is designed with a dynamic feature extraction module to parse complex local features and enhance the network's generalization capability, as well as a dependency modeling module to capture the relationships between global contextual information for better understanding and interpretation of the ECG signals. All-night PSG recordings from both public and private datasets are used to validate the proposed framework. The results show that the framework achieves an accuracy of 0.942, a sensitivity of 0.926, a specificity of 0.959, and an F1 score of 0.942 on the public dataset, and an accuracy of 0.946, a sensitivity of 0.921, a specificity of 0.972, and an F1 score of 0.945 on the private dataset. These results indicate that the proposed framework offers high accuracy in automatic OSA classification and has significant potential to aid clinical decision-making.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144803998","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}

{"title":"ProBot: Robot-Ultrasound Probe for Transrectal and Transperineal Prostate Biopsy.","authors":"Dan Stoianovici, Sunghwan Lim, Kunio Hashiba, John Waddell, Doru Petrisor, Jakub Piwowarczyk, Arvin K George, Katarzyna J Macura, Misop Han","doi":"10.1109/TBME.2025.3595815","DOIUrl":"https://doi.org/10.1109/TBME.2025.3595815","url":null,"abstract":"<p><strong>Objective: </strong>Over the past decade the incidence of prostate cancer (PCa) has been on a steady increase. Efforts to improve PCa outcome include targeted biopsy with magnetic resonance imaging (MRI) and precision sampling. Several biopsy devices are currently available to guide the biopsy with MRIultrasound fusion. They commonly use generic handheld ultrasound probes retrofitted for fusion biopsy. Robotic probe manipulation has the potential to reduce the required training, skills, and outcome variability among urologists. Rather than retrofitting an ultrasound probe, we developed a novel cohesive ultrasound-robotic probe (ProBot) that enables a novel needle insertion path.</p><p><strong>Methods: </strong>Tissue-probe contact frequently deforms the prostate contributing to fusion errors. To minimize deformations, ProBot uses a side-fire probe and its only motion is a robotic rotation about its axis, which can't force the gland. An additional robotic angulation of the needle allows targeting any gland location. As such, only 2 degrees-of-freedom are required for prostate biopsy. However, the probe and robot must allow clearance to angle the needle, thus a special probe and RemoteCenter-of-Motion (RCM) robot kinematics were developed. We present their design, ProBot prototype, and adapters for transrectal and transperineal needle access.</p><p><strong>Results: </strong>Pre-clinical tests showed sub-millimeter targeting errors and validated the sterilization process. With IRB approval, transrectal prostate biopsy was successfully performed in two patients.</p><p><strong>Conclusion: </strong>ProBot is a novel device for prostate biopsy (robotic, simple, compact, precise) with a successful pilot clinical evaluation.</p><p><strong>Significance: </strong>A precision, skill independent biopsy device can impact the management of PCa. Additional trials are required.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144803999","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}

{"title":"Dose is a Critical Factor Affecting Treatment Volumes for Integrated Nanosecond Pulse Irreversible Electroporation (INSPIRE).","authors":"Michael B Sano, Jordan A Fong, Robert H Williamson, Jewels Darrow, Logan Reeg, Kyle G Mathews, Callie A Fogle, Nathan C Nelson, Alina C Iuga, David A Gerber","doi":"10.1109/TBME.2025.3597274","DOIUrl":"https://doi.org/10.1109/TBME.2025.3597274","url":null,"abstract":"<p><strong>Objective: </strong>The objective of this study was to investigate the effect of electrical dose on in vivo INSPIRE treatments which administer high voltage ultrashort alternating polarity electrical pulses with active temperature control.</p><p><strong>Methods: </strong>INSPIRE was administered to healthy swine liver in vivo via a percutaneous single applicator and grounding pad approach. Using 45°C temperature control, 6000V waveforms consisting of 750ns, 1000ns, or 2000ns bipolar pulses were administered to examine the effect of pulses approximately shorter than, equal to, and longer than the cell membrane charging time. Treatment volumes were assessed one week post treatment via computed tomography and cardiac safety was assessed via serum troponin analysis.</p><p><strong>Results: </strong>Pulse duration did not significantly affect treatment volumes, however, dose was found to be a critical factor affecting treatment outcomes. For 0.0025s doses, treatment volumes of 1.3±0.6cm³ (2.4x0.9cm) were created in 0.3 minutes. This increased to 12.8±4.8 cm³ (9.7 minutes, 3.9x2.5cm) for 0.04s doses. No significant changes in troponin levels were found.</p><p><strong>Conclusion: </strong>This study demonstrated the in vivo safety of high voltage INSPIRE treatments without cardiac synchronization. There is a strong dose dependent effect on treatment volumes. Optimal treatment efficiency was found for treatment doses between 0.01 and 0.02s with treatment times between 2-4 minutes.</p><p><strong>Significance: </strong>Single applicator INSPIRE treatments significantly simplify treatment planning and clinical implementation versus traditional two to six applicator approaches. This study demonstrates that INSPIRE protocols can rapidly produce large spherical treatment zones while reducing treatment times by an order of magnitude compared to existing electroporation approaches.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144803997","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}

{"title":"Passive Acoustic Dynamic Differentiation and Mapping (PADAM): A Time-Domain Passive Cavitation Localization and Classification Approach.","authors":"Nathan Caso, Krunal Patel, Tao Sun","doi":"10.1109/TBME.2025.3596596","DOIUrl":"https://doi.org/10.1109/TBME.2025.3596596","url":null,"abstract":"<p><strong>Objective: </strong>Passive cavitation imaging has explored various beamforming algorithms to optimize spatial resolution, suppress imaging artifacts, and maintain computational efficiency. These factors are crucial for the clinical translation of Focused Ultrasound (FUS) therapies, where precise cavitation localization and dose control are required to minimize off-target effects. Commonly used methods such as Delay-Sum-Integrate (DSI) and Robust Capon Beamforming (RCB) have shown utility, but are limited by either significant artifacts or the need for a nonphysical input parameter. To address these challenges, we aimed to develop a method that enhances resolution and introduces a physically grounded parameter for signal characterization, without compromising computational speed and robustness.</p><p><strong>Methods: </strong>This work introduces Passive Acoustic Dynamic Differentiation and Mapping (PADAM), which adapts the Multiple Signal Classification algorithm to the time domain to improve cavitation localization and classification. PADAM incorporates a physically meaningful input parameter that dynamically reflects the frequency richness of the received signal.</p><p><strong>Results: </strong>PADAM achieves up to a 6-fold improvement in lateral beamwidth compared to RCB, and a 4-fold reduction in mean-square artifact intensity reduction. Its input parameter provides a novel physical insight, enabling differentiation between stable and inertial cavitation based on spectral content. This reduces reliance on empirically tuned or arbitrary thresholds and simplifies integration into therapy workflows.</p><p><strong>Conclusion: </strong>With its ability to improve resolution, reduce artifacts, and provide computational efficiency, PADAM represents a promising advancement for precise cavitation localization and therapy monitoring.</p><p><strong>Significance: </strong>This work introduces PADAM, a time-domain passive cavitation imaging method that offers superior resolution and artifact reduction compared to DSI and RCB. Its physically intuitive input parameter enables dynamic differentiation between stable and inertial cavitation, enhancing precision in the monitoring and control of FUS therapy.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144798940","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}

{"title":"Adaptive algorithms for DPOAE level-map acquisition.","authors":"Philipp Graff, Stepan Kempa, Anthony W Gummer, Ernst Dalhoff, Katharina Bader","doi":"10.1109/TBME.2025.3596784","DOIUrl":"https://doi.org/10.1109/TBME.2025.3596784","url":null,"abstract":"<p><p>Distortion-product otoacoustic emissions (DPOAE) are intermodulation products stimulated by two tones and reflect the nonlinear mechanical processing within the inner ear by the so-called cochlear amplifier. Therefore, they are interpreted as a diagnostic measure of its functional state. Due to their small amplitudes and the correspondingly long averaging time, current clinical systems measure one DPOAE amplitude at approximately seven fixed frequencies, representing a one-dimensional (1D) scan. More advanced systems record input-output functions of the distortion product amplitudes, where both tones are varied in a predefined ratio, resulting in a two-dimensional (2D) scan. A three-dimensional (3D) scan, where both stimulus tones are varied independently, yields more detailed information about the cochlea, but at the cost of longer measurement times. In this study, we introduce an adaptive measurement method, that autonomously collects more DPOAE data with sufficient signal-to-noise ratio (SNR) and leads more often to the identification of the so-called \"individually optimal stimulus level path\" than the previously used static method, which uses predefined fixed stimulus levels. In six ears of three subjects, the adaptive method detected 23% more valid DPOAE data. A bivariate histogram of area and density of DPOAE amplitudes for the optimal path samples shows that the adaptive method balances these competing goals more effectively. This results in higher-quality DPOAE level maps. Thus, the adaptive method has proven to be a time-efficient approach to characterize cochlear function more comprehensively.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144798939","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}

{"title":"A Spatiotemporal Causal Model for Revealing Developmental Changes in Infants' Brain Effective Connectivity Networks During the First Year of Life.","authors":"Meiliang Liu, Chao Yu, Xiaoxiao Yang, Yunfang Xu, Huiwen Dong, Zijin Li, Zhengye Si, Xinyue Yang, Junhao Huang, Ziyuan Shi, Kuiying Yin, Zhiwen Zhao","doi":"10.1109/TBME.2025.3596893","DOIUrl":"https://doi.org/10.1109/TBME.2025.3596893","url":null,"abstract":"<p><strong>Objective: </strong>The infant's brain undergoes significant structural and physiological transformations during the first year of life. Although extensive research has explored brain development during this critical period, most studies relied on traditional functional connectivity rather than effective connectivity (EC).</p><p><strong>Methods: </strong>We proposed a novel spatiotemporal Granger causality model for discovering causal relationships from time series. Different from existing deep learning-based models that infer Granger causality from the first-layer weights of the neural network, our model learns the spatiotemporal feature from time series and imposes a sparsity-inducing penalty on the feature to extract Granger causality.</p><p><strong>Results: </strong>Numerical experiments showed that our model achieved an average Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.879 and an Area Under the Precision-Recall Curve (AUPRC) of 0.828 on simulated datasets, outperforming baselines by up to 5.7% in AUROC and 3.9% in AUPRC. When applied to resting-state EEG from healthy infants, the model revealed distinct age-related patterns in EC networks. Specifically, right-hemispheric lateralization was observed between 3 and 9 months, which subsequently shifted toward hemispheric symmetry between 9 and 12 months. The development pattern of EC followed a posterior-to-anterior trajectory, starting in the occipital and advancing through the parietal, temporal, and frontal regions, consistent with the known myelination process. Network property analysis revealed that, as infants grew, their brain networks exhibited higher clustering coefficients and global efficiency along with reduced shortest path lengths.</p><p><strong>Conclusion/significance: </strong>Our study is the first to uncover changes in infants' brain EC network during the first year of life using the causal discovery approach, offering novel insights for investigating developmental trajectories during early infancy.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144798938","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}