IEEE Transactions on Biomedical Engineering最新文献

{"title":"RePaint High-Density Surface Electromyography Signal Using Denoising Diffusion Probabilistic Model.","authors":"Yihui Zhao, Jiawei Liao, Xia Fang, Hai Wang, Ning Jiang, Jiayuan He","doi":"10.1109/TBME.2025.3604527","DOIUrl":"https://doi.org/10.1109/TBME.2025.3604527","url":null,"abstract":"<p><strong>Objective: </strong>High-density surface electromyography (HD-sEMG) has emerged as a powerful tool for myoelectric control and activation pattern analysis. However, signal loss due to poor electrode contact and channel corruption remains a significant challenge, limiting the reliability and practical applications of HD-sEMG signals. Conventional interpolation methods fail to effectively reconstruct corrupted signals, especially when multiple adjacent channels are affected.</p><p><strong>Methods: </strong>This paper proposes a novel HD-sEMG signal reconstruction approach based on the denoising diffusion probabilistic model (DDPM) with a repaint strategy. By leveraging a U-Net structure with spatiotemporal embedding modules that effectively learn the spatial and temporal characteristics of HD-sEMG signals, the proposed method achieves high-fidelity signal reconstruction without requiring prior knowledge of corruption patterns.</p><p><strong>Results: </strong>Experimental evaluations are conducted on 6 corruption patterns with varying ratios (from 12.5% to 50%) using self-collected datasets (including an amputated subject) and a benchmark dataset. Results demonstrate that the proposed approach consistently outperforms interpolation methods (linear: 0.038$pm$0.033, cubic: 0.038$pm$0.032), generative adversarial net (GAN) (0.049$pm$0.041), and variational autoencoder (VAE) (0.068$pm$0.046) in terms of $nRMSE$ ($p < 0.001$), achieving the lowest error of 0.027$pm$0.027 averaged across all corruption ratios. For $PSNR$, the proposed approach achieves the highest mean value (35.81$pm$ 17.95dB) compared to interpolation methods (linear: 33.89$pm$26.85, cubic: 33.88$pm$ 26.88dB), GAN (31.08$pm$ 19.14dB), and VAE (26.98$pm$ 18.94dB) ($p < 0.001$). Furthermore, the proposed method maintained robust classification accuracy, achieving statistically equivalent performance to ground truth at the lower corruption ratio.</p><p><strong>Significance: </strong>The proposed HD-sEMG signal reconstruction approach offers a new solution for enhancing the fidelity and reliability of HD-sEMG signal acquisition.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144952333","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 Resource-Efficient Cardiac Arrhythmia Detection Using Nonlinear Dynamics in Optimized Delay State Networks.","authors":"Basab Bijoy Purkayastha, Shovan Barma, Manob Jyoti Saikia","doi":"10.1109/TBME.2025.3605297","DOIUrl":"https://doi.org/10.1109/TBME.2025.3605297","url":null,"abstract":"<p><p>In this study, a novel methodology is proposed, combining Reconstructed Phase Space (RPS) analysis with an optimized Delay State Network (DSN) to enhance the detection and classification of cardiac arrhythmias. Traditional methods often fail to capture subtle temporal phase drifts indicative of arrhythmias or require extensive computational resources and handcrafted features, limiting their effectiveness for early diagnosis and real-time applicability. The proposed approach reconstructs the nonlinear dynamics of cardiac signals and leverages the entire Phase Space Structure (PSS) as direct input to the DSN. The optimized DSN employs a single nonlinear node with delayed feedback to emulate multiple virtual nodes, reducing hardware demands by over an order of magnitude compared to conventional reservoirs or LSTMs. To accurately capture ECG dynamics, the framework integrates delay and embedding optimization, while PCA and Ridge Embedding manage dimensionality within the DSN. The functionality of the DSN model is further optimized by incorporating shared memory and multiprocessing frameworks, enabling scalable and efficient handling of large datasets. The methodology was validated on three benchmark datasets, demonstrating its generalizability across diverse cardiac conditions. Experimental results achieved $99.3%$ accuracy, with sensitivity and specificity of $99.1%$ and $99.7%$, respectively. Edge deployment on a Raspberry Pi 5 demonstrated inference within $1.2-4.8$ seconds for $10s-60s$ ECG segments, with peak memory usage of 2.57 GB observed for 60s segments, and power consumption remaining below 2.5 W. The proposed framework provides a robust, scalable, and accurate solution for arrhythmia classification and broader time-series-based diagnostics in resource-constrained environments.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144952195","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":"Diffusion-QSM: diffusion model with timetravel and resampling refinement for quantitative susceptibility mapping.","authors":"Ming Zhang, Chunlei Liu, Yuyao Zhang, Hongjiang Wei","doi":"10.1109/TBME.2025.3605587","DOIUrl":"https://doi.org/10.1109/TBME.2025.3605587","url":null,"abstract":"<p><strong>Objective: </strong>Quantitative susceptibility mapping (QSM) is a useful magnetic resonance imaging technique. We aim to propose a deep learning (DL)-based method for QSM reconstruction that is robust to data perturbations.</p><p><strong>Methods: </strong>We developed Diffusion-QSM, a diffusion model-based method with a time-travel and resampling refinement module for high-quality QSM reconstruction. First, the diffusion prior is trained unconditionally on high-quality QSM images, without requiring explicit information about the measured tissue phase, thereby enhancing generalization performance. Subsequently, during inference, the physical constraints from the QSM forward model and measurement are integrated into the output of the diffusion model to guide the sampling process toward realistic image representations. In addition, a time-travel and resampling module is employed during the later sampling stage to refine the image quality, resulting in an improved reconstruction without significantly prolonging the time.</p><p><strong>Results: </strong>Experimental results show that Diffusion-QSM outperforms traditional and unsupervised DL methods for QSM reconstruction using simulation, in vivo and ex vivo data and shows better generalization capability than supervised DL methods when processing out-of-distribution data.</p><p><strong>Conclusion: </strong>Diffusion-QSM successfully unifies data-driven diffusion priors and subjectspecific physics constraints, enabling generalizable, high-quality QSM reconstruction under diverse perturbations, including image contrast, resolution and scan direction.</p><p><strong>Significance: </strong>This work advances QSM reconstruction by bridging the generalization gap in deep learning. The excellent quality and generalization capability underscore its potential for various realistic applications.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144952276","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":"Kinematic parameter estimation using workspace manifold mapping.","authors":"Eric R Peltola, Eunsuk Chong, Xiaoyu Wang, Veronica J Santos","doi":"10.1109/TBME.2025.3604794","DOIUrl":"https://doi.org/10.1109/TBME.2025.3604794","url":null,"abstract":"<p><strong>Objective: </strong>This work proposes a method to estimate the kinematic parameters of multi-joint systems where direct measurement is infeasible, such as joints of the hand.</p><p><strong>Methods: </strong>Our novel data-driven estimation method uses \"workspace manifold mapping\" that relies on a unique geometry that arises in exponential representations of 3D motions of two rigid bodies connected via one- or two-degree-of-freedom (DOF) joints. We describe and verify our \"Generative Topographic Mapping algorithm with kinematic constraints\" (GTM-KC) using simulated data and motion capture data for a 2-DOF bio-inspired mechanical linkage. We compare the performance of GTM-KC to several benchmark algorithms.</p><p><strong>Results: </strong>Upon applying GTM-KC to motion capture data of a bio-inspired linkage, the mean estimates of the 2-DOF joint axis orientations deviated from ground truth by 2.5° with a standard deviation of 3.4° for one axis and by 2.4° with a standard deviation of 2.7° for the second axis.</p><p><strong>Conclusion: </strong>Our GTM-KC method can be used to estimate the orientations of revolute joint axes that link the 3D kinematics of two rigid bodies, and either outperforms or is equivalent to existing methods in terms of accuracy, precision, and reliable convergence to a solution. Notably, the GTM-KC method outperforms existing methods in terms of robustness to initial conditions.</p><p><strong>Significance: </strong>Workspace manifold mapping provides improved kinematic parameter estimation as compared to existing benchmark methods, and can be applied to any 1- or 2-DOF kinematic relationship without loss of generality.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144952344","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 Decade of Rapid Serial Visual Presentation Paradigm in Brain-Computer Interface for Target Detection: Current Status and Trends.","authors":"Meng Xu, Baiwen Zhang, Lijian Zhang, Dan Wang, Yuanfang Chen","doi":"10.1109/TBME.2025.3603945","DOIUrl":"https://doi.org/10.1109/TBME.2025.3603945","url":null,"abstract":"<p><strong>Objective: </strong>Electroencephalography (EEG)-based Rapid Serial Visual Presentation (RSVP) has steadily gained attention since 2015 as a paradigm to enhance image target detection in brain-computer interfaces (BCIs) used with healthy individuals.</p><p><strong>Methods: </strong>We reviewed the literature using Scopus and Web of Science as primary databases, covering publications from 2015 to 2024. After literature screening and filtering, a total of 86 papers on RSVP-BCI studies were analyzed over this decadelong period. The research categorizes RSVP into three dimensions: public datasets, paradigm encoding, and decoding methods, while exploring eight mode combinations involving target types, subject groups, and different modalities.</p><p><strong>Results: </strong>Our literature search revealed a scarcity of studies addressing diverse target types across different subject groups or modality combinations, indicating a promising direction for future RSVP-BCI development. Future efforts should prioritize inclusivity across all age groups, the design of user-friendly stimulus interfaces, and the development of advanced algorithms, with the goal of creating a more widely accessible RSVP-BCI system.</p><p><strong>Conclusion: </strong>We have provided a comprehensive review of advances over the past decade in RSVP-based target detection, including datasets, encoding design, and decoding methods and potential applications.</p><p><strong>Significance: </strong>The present work aims to articulate prospective trajectories for the continued advancement of the RSVP community.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144952265","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":"Influence of Shear Waves on Transcranial Ultrasound Propagation in Cortical Brain Regions.","authors":"Ya Gao, Beat Werner, Beatrice Lauber, Yiming Chen, Giovanni Colacicco, Daniel Razansky, Hector Estrada","doi":"10.1109/TBME.2025.3603836","DOIUrl":"https://doi.org/10.1109/TBME.2025.3603836","url":null,"abstract":"<p><strong>Objective: </strong>Transcranial ultrasound applications require accurate simulations to predict intracranial acoustic pressure fields. The current gold standard typically consists of calculating a longitudinal ultrasound wave propagation using a fluid skull model, which is based on full head CT images for retrieving the skull's geometry and elastic constants. Although this approach has extensively been validated for deep brain targets and routinely used in transcranial ultrasound ablation procedures, its accuracy in shallow cortical regions remains unexplored. In this study, we explore the shear wave effects associated with transcranial focused ultrasound propagation, both numerically and experimentally. The intracranial acoustic pressure was measured at different incidence angles at the parietal and frontal regions in an ex vivo human skull. The fluid-like skull model was then compared to the solid model comprising both longitudinal and shear waves. The results consistently show a larger error and variability for both models when considering an oblique incidence, reaching a maximum of 170% mean deviation of the focal area when employing the fluid skull model. Statistical assessments further revealed that ignoring shear waves results in an average ∼40% overestimation of the intracranial acoustic pressure and inability to obtain an accurate intracranial acoustic pressure distribution. Moreover, the solid model has a more stable performance, even when small variations in the skull-transducer relative position are introduced. Our results could contribute to the refinement of the transcranial ultrasound propagation modeling methods thus help improving the safety and outcome of transcranial ultrasound therapy in the cortical brain areas.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144952321","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":"Displacement-guided Focused Ultrasound of the Median Nerve Modulates Somatosensory Evoked Potentials in Humans.","authors":"Erica P McCune, Samuel G Blackman, Hermes A S Kamimura, Ethan V Bendau, Talia D Sachs, Seongyeon Kim, Stephen A Lee, Christopher J Winfree, Elisa E Konofagou","doi":"10.1109/TBME.2025.3602291","DOIUrl":"https://doi.org/10.1109/TBME.2025.3602291","url":null,"abstract":"<p><strong>Objective: </strong>This study investigated the attenuation of somatosensation in healthy subjects (n = 18) and subjects with carpal tunnel syndrome (CTS) (n = 6) via displacement-guided median nerve focused ultrasound (FUS). Somatosensory evoked potentials (SSEPs), pain, stiffness, and numbness were used as markers of sensation.</p><p><strong>Methods: </strong>Electroencephalography (EEG) was used to evaluate changes in electrically-induced SSEPs when paired with FUS. Subjects underwent 1,000 (healthy population) or 500 (CTS population) surface electrical pulses at a maximum rate of 1 Hz to the median nerve to invoke a thumb twitch. Half of these pulses were paired with 5 ms of upstream median nerve sonication from a 1.1 MHz transducer at 1.8 or 2.9 MPa derated peak-positive pressure.</p><p><strong>Results: </strong>On-nerve sonication reduced the amplitude of SSEPs at both pressures. Reductions in beta and low-gamma frequency power occurred in healthy subjects and increases in alpha power occurred in CTS subjects. CTS subjects who received on-nerve sonication reported an average pain reduction of 40.6%. Off-nerve sonication did not reduce SSEP amplitudes or CTS symptoms. Median nerve displacement was correlated with SSEP reductions.</p><p><strong>Conclusion: </strong>Sonication of the median nerve can attenuate SSEPs, potentially in a displacement dose-driven manner. Symptom reduction in CTS subjects may last 1-3 days post-procedure. Nerve targeting is essential to ensure therapeutic efficacy.</p><p><strong>Significance: </strong>This study is the first to demonstrate that peripheral nerve FUS modulates somatosensory evoked potentials in humans. FUS attenuated neuropathic pain and stiffness related to median nerve entrapment, illustrating the potential analgesic effects of the procedure.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144952266","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":"Non-invasive measurement of in vivo corneal steady-state biomechanical properties via controllable negative pressure inflation.","authors":"Zimeng Zhou, Honghao Wang, Zuowei Wang, Ying Zhang, Haijun Lv, Zhuoyu Zhang, Zhengwen Gao, Xiuli Liu, Xiaohua Lv, Tingwei Quan, Shangbin Chen, Shaoqun Zeng","doi":"10.1109/TBME.2025.3603048","DOIUrl":"https://doi.org/10.1109/TBME.2025.3603048","url":null,"abstract":"<p><strong>Objective: </strong>To measure the steady-state corneal biomechanical properties related to postoperative corneal ectasia, keratoconus, glaucoma and other ophthalmic diseases, we propose a novel in vivo measurement method.</p><p><strong>Methods: </strong>By precisely manipulating ambient negative pressure via a suction device to achieve controlled in vivo corneal inflation, we analyzed the coupling relationship between corneal deformation response and negative pressure loading. The displacement corresponding to the corneal initial configuration under 0 mmHg was extrapolated. The inverse finite element analysis (FEA) technique was employed to calculate the segmental Young's modulus of the cornea under different steady-state intraocular pressure (IOP) conditions.</p><p><strong>Results: </strong>In vivo experiments on 3-month-old rabbit eyes demonstrated a Young's modulus of 0.2-0.3 MPa at physiological IOP (10-12 mmHg), along with an increasing trend of corneal stiffness across a pressure range of 0-85 mmHg.</p><p><strong>Conclusion: </strong>The proposed non-invasive measurement method exhibits stability and consistency in parameter inversion under different IOPs, indicating its significant clinical potential.</p><p><strong>Significance: </strong>Providing a new approach for corneal biomechanical assessment is beneficial for the diagnosis and treatment of ophthalmic diseases.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144952322","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":"Distinct Neural Information Shared in Spastic Muscle through Decoding Motoneuron Activity.","authors":"Tian Xie, Chuyao Jian, Yujian Lei, Yan Leng, Chuhuai Wang, Rong Song","doi":"10.1109/TBME.2025.3600863","DOIUrl":"https://doi.org/10.1109/TBME.2025.3600863","url":null,"abstract":"<p><strong>Objective: </strong>We aimed to assess whether spinal motoneurons received and sent the same neural information during involuntary and voluntary activation following stroke.</p><p><strong>Methods: </strong>High-density surface electromyography (HD-sEMG) signals of biceps brachii muscle were recorded, while 14 stroke survivors and 10 age-matched controls performed passive stretch and active contraction. Populational motor unit (MU) activity was extracted from HD-sEMG recordings with decomposition algorithms. The MU discharge rate, discharge variability and spatial distribution of MU action potentials (MUAP) were used to detect neural drive to muscles. The cross-correlation analysis between MU discharge timings was performed to detect common synaptic input (CSI) to motoneurons.</p><p><strong>Results: </strong>In stroke survivors, involuntary activation exhibited higher discharge rate and lower discharge variability than voluntary activation (18.41 ± 2.05 Hz vs. 14.99 ± 1.50 Hz, p = 0.000, d = 1.392; 0.04 ± 0.02 vs. 0.12 ± 0.04, p = 0.000, d = 1.775). The opposite MUAP distribution patterns were observed between involuntary and voluntary activation (lateral-medial: 3.91 ± 1.01 vs. 4.85 ± 0.57, p = 0.000, d = 1.304; distal-proximal: 5.16 ± 0.80 vs. 4.01 ± 0.73, p = 0.001, d = 1.222). CSI was lower in involuntary activation than voluntary activation (0.42 ± 0.07 vs. 0.55 ± 0.08, p = 0.004). The discharge variability was significantly positively correlated with CSI.</p><p><strong>Conclusion: </strong>Our decoding results demonstrated that information flow between supraspinal and spinal cord was unbalanced following stroke.</p><p><strong>Significance: </strong>The determinants of MU discharge in reflex activity depend more on intrinsic properties of spinal motoneurons than on brain.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144952295","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":"Meta-learning with Unlabeled Query Updating and Consistency Learning for Few-shot OCT Image Classification.","authors":"Ziting Yin, Bo Wu, Weifang Zhu, Dehui Xiang, Xinjian Chen, Tao Peng, Qing Peng, Fei Shi","doi":"10.1109/TBME.2025.3602687","DOIUrl":"https://doi.org/10.1109/TBME.2025.3602687","url":null,"abstract":"<p><strong>Objective: </strong>Deep neural networks are widely used in the field of optical coherence tomography (OCT) to screen some common retinal diseases. However, for rare diseases with fewer cases for model training, it is challenging to achieve automatic diagnosis using traditional deep learning. Meta-learning based few-shot learning can be used to address the problem of insufficient training data.</p><p><strong>Methods: </strong>We propose a novel algorithm for few-shot OCT image classification, where meta-learning is used to fine-tune the pre-trained model and obtain good initialization for task generalization. Unsupervised learning based on query data is for the first time introduced in meta-learning. Cross-set consistency learning is proposed to reduce the gap between meta-knowledge learned from support and query data. Data mixup is also integrated to generate virtual samples and enhance data variety.</p><p><strong>Results: </strong>A lightweight subset was constructed based on a public OCT dataset and extensive experiments were performed. The classification accuracy of the proposed method was higher than existing few-shot learning methods. To show the generalization of the proposed method, experiments were also performed on a histological image dataset, and superior performance was also achieved.</p><p><strong>Conclusion: </strong>The proposed strategies help the model to fully utilize the limited data and to explore hidden information, improving its generalization to unseen tasks.</p><p><strong>Significance: </strong>The proposed method has great value in training deep learning models for diagnosis of rare diseases.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":""},"PeriodicalIF":4.5,"publicationDate":"2025-08-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144952375","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}