Biomedical Signal Processing and Control最新文献

{"title":"SSA-LSTM-based locomotion mode recognition algorithm for the control of powered hip disarticulation prostheses","authors":"Qiaoling Meng ,&nbsp;Zhenkun Sun ,&nbsp;Jing Zhao ,&nbsp;Vincenzo Parenti Castelli ,&nbsp;Hongliu Yu","doi":"10.1016/j.bspc.2025.108583","DOIUrl":"10.1016/j.bspc.2025.108583","url":null,"abstract":"<div><h3>Background</h3><div>Accurate recognition of locomotion modes is essential for the effective control of lower limb prosthetics, enabling amputees to navigate various terrains with ease. Despite advancements, current prosthetics lack adaptive capabilities for complex movements, necessitating intelligent systems that can discern user intentions from sensory inputs.</div></div><div><h3>Objective</h3><div>This paper introduces the SSA-LSTM algorithm, which integrates the Sparrow Search Algorithm (SSA) with Long Short-Term Memory (LSTM) networks to enhance the stability and accuracy of motion pattern recognition in powered hip disarticulation prostheses.</div></div><div><h3>Methods</h3><div>A comprehensive dataset was constructed, capturing gait characteristics of both healthy individuals and amputees across various motion modes, including level walking, stair climbing, and ramp navigation. The SSA-LSTM algorithm optimizes the LSTM’s initial state, thereby improving convergence and learning efficiency. Its performance was bench-marked against established methods, including Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), ensemble learning, and LSTM.</div></div><div><h3>Results</h3><div>The SSA-LSTM model achieved superior recognition accuracy, averaging over 99 % for healthy subjects and 96.4 % for hip disarticulation amputees. This model demonstrated faster convergence, underscoring the SSA’s role in enhancing the LSTM’s learning capabilities.</div></div><div><h3>Conclusion</h3><div>The SSA-LSTM model, through its integration of SSA optimization, represents a significant advancement in locomotion mode recognition. This research contributes to the development of intelligent prosthetics by providing a more precise and responsive control mechanism, which is crucial for enhancing the mobility and independence of amputees.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108583"},"PeriodicalIF":4.9,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145219703","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":"Population-invariant measurement of IMU-based gait biomarkers for automated Parkinson’s disease diagnosis","authors":"Xiangzhi Liu,&nbsp;Hanyi Huang,&nbsp;Jiaxing Li,&nbsp;Haozhou Zeng,&nbsp;Xiangliang Zhang,&nbsp;Tao Liu","doi":"10.1016/j.bspc.2025.108678","DOIUrl":"10.1016/j.bspc.2025.108678","url":null,"abstract":"<div><div>Accurate and scalable Parkinson’s disease (PD) screening currently demands extensive clinician time and costly imaging or laboratory resources. Gait analyze have emerged as a promising digital biomarker, yet cohort-specific variability often obscures disease signals and undermines cross-group performance. We present a population-invariant IMU signal measurement framework that extracts robust gait biomarkers for PD diagnosis. Using two shank-mounted inertial measurement units (IMUs), our method applies Multivariate Singular Spectrum Analysis (MSSA) to five consecutive gait cycles, systematically isolates and removes cohort-confounding modes, and then reconstructs purified gait signals. Statistical validation via the Bhattacharyya distance demonstrates a marked reduction in inter-cohort variance while preserving diagnostic features. Evaluated on a diverse population of 127 subjects—spanning young healthy, middle-aged healthy, older healthy, middle-aged PD, and older PD groups—this lightweight, low-cost pipeline achieves 94.5 % cross-validated diagnostic accuracy. By delivering universal gait biomarkers that transcend age and demographic differences, our approach minimizes cohort bias, enhances generalizability, and paves the way toward automated, precision-diagnostic tools for Parkinson’s disease.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108678"},"PeriodicalIF":4.9,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145219696","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":"HSI-MSSAF net: A dual-stream network for nasal tumor tissue diagnosis using hyperspectral spectral-spatial features","authors":"Yunze Li ,&nbsp;Fangying Liu ,&nbsp;Yanhai Zhang ,&nbsp;Guanghui Liu ,&nbsp;Jinlin Deng ,&nbsp;Qize Lv ,&nbsp;Yifei Liu ,&nbsp;Haomiao Zhao ,&nbsp;Wei Li ,&nbsp;Xin Feng","doi":"10.1016/j.bspc.2025.108691","DOIUrl":"10.1016/j.bspc.2025.108691","url":null,"abstract":"<div><div>Accurate differentiation of benign and malignant nasal cavity lesions is clinically critical due to their lack of distinct morphological specificity. Hyperspectral imaging (HSI), emerging as a novel modality, deciphers spatial-spectral multidimensional signatures of pathological tissues, thereby delivering novel data dimensions for diagnostic precision beyond conventional histomorphological limitations. This study employs HSI technology and a multi-scale spatial-spectral attention fusion network (HSI-MSSAF net), which combines residual networks, Transformer network architecture, and multi-scale attention mechanisms. This approach efficiently extracts and integrates spatial-spectral features from different scales and channels.Experimental results show that the proposed method achieves remarkable performance in differentiating benign and malignant nasal tumors, with a classification accuracy of 91.8%, precision of 0.91, recall of 0.92, F1 score of 0.92, AUC of 0.98, and Matthews correlation coefficient of 0.84. The results indicate that the proposed model effectively leverages sample data to learn comprehensive joint feature representations. The novel methodology introduced herein offers a complementary strategy that may mitigate certain limitations inherent to conventional medical diagnostic techniques, thereby underscoring the potential of high-precision diagnostic approaches in facilitating the classification and prognostic evaluation of complex nasal tumors. By establishing a dedicated hyperspectral nasal tumor database and implementing advanced network architectures, this approach demonstrates potential for clinical integration, contingent upon further in vivo validation.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108691"},"PeriodicalIF":4.9,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145219635","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":"MDCT-Unet: A dual-encoder network combining multi-scale dilated convolutions with Transformer for medical image segmentation","authors":"Zhengshi Chen ,&nbsp;Juanjuan He ,&nbsp;Xiaoming Liu ,&nbsp;Jinshan Tang","doi":"10.1016/j.bspc.2025.108709","DOIUrl":"10.1016/j.bspc.2025.108709","url":null,"abstract":"<div><div>Precise medical image segmentation is crucial in clinical diagnosis and pathological analysis. Most segmentation methods are based on U-shaped convolutional neural networks (U-Net). Although U-Net performs well in medical image segmentation, as a method based on CNN, its main drawback lies in the difficulty of establishing long-range pixels dependencies and has a constrained receptive field, which restricts segmentation accuracy. Many models address this issue by incorporating Transformer models into U-Net architectures to better capture long-range dependencies. However, these methods often suffer from simple feature fusion techniques and limited receptive fields for local features. To address these challenges, we propose a dual-encoder framework, named MDCT-Unet, which combines Swin-Transformer and CNN for enhanced medical image segmentation. This framework introduces a novel dynamic feature fusion module to better integrate of local and global features. By combining channel and spatial attention mechanisms and inducing competition between them, we enhance the coupling of these two types of features, ensuring richer information representation. In addition, to better extract multi-scale local features from medical images, we design a dilated convolution encoder (DCE) as the CNN branch of our model. By incorporating dilated convolutions with varying receptive fields, DCE captures rich local features at multiple scales, thereby enhancing the model’s ability to segment challenging regions such as boundaries and small organs. We conducted extensive experiments on four datasets: Synapse, ISIC2018, CHASEDB1, and MMWHS. The experimental results show that our method outperforms most current medical image segmentation methods quantitatively and qualitatively.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108709"},"PeriodicalIF":4.9,"publicationDate":"2025-09-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145219636","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":"Assessing the impact of signal quality on heart rate detection from long-term clinical wrist PPG under varying cardiac rhythms","authors":"Yangyang Zhao ,&nbsp;Olli Lahdenoja ,&nbsp;Jonas Sandelin ,&nbsp;Sepehr Seifizarei ,&nbsp;Arman Anzanpour ,&nbsp;Joonas Lehto ,&nbsp;Joel Nuotio ,&nbsp;Jussi Jaakkola ,&nbsp;Arto Relander ,&nbsp;Tuija Vasankari ,&nbsp;Juhani Airaksinen ,&nbsp;Tuomas Kiviniemi ,&nbsp;Matti Kaisti ,&nbsp;Tero Koivisto","doi":"10.1016/j.bspc.2025.108688","DOIUrl":"10.1016/j.bspc.2025.108688","url":null,"abstract":"<div><div>Reliable heart rate (HR) detection is essential for long-term cardiac monitoring, particularly in hospitalized patients with complex conditions. Due to its optical and non-invasive nature, photoplethysmography (PPG) is inherently susceptible to motion artifacts and noise. These challenges intensify under arrhythmic conditions such as atrial fibrillation (AF), where signal distortions may blur the boundary between poor-quality segments and pathological rhythms, potentially impairing downstream tasks like HR estimation. This study developed a signal quality assessment (SQA) algorithm designed for this high-risk clinical population and evaluated its robustness through HR estimation. We collected 24-hour synchronous PPG and electrocardiogram (ECG) recordings from 49 hospitalized cardiac patients, with all PPG segments manually annotated for quality. External validation was conducted using the MIMIC-IV dataset. To avoid dependence on specific segment lengths or classifier types, we assessed SQA performance using seven machine learning models and four segmentation lengths. The SQA framework was then applied to HR estimation to evaluate clinical utility. We implemented a Standard Deviation of Successive Differences (SDSD)-based peak filtering method and compared it with an autocorrelation-based approach under different cardiac rhythm conditions. Threshold tuning in both SQA classification and SDSD filtering was conducted to explore the balance between data usability and reliable HR estimation. The proposed model achieved an AUROC of 96.1% (Sinus Rhythm (SR) ＋ AF), with 90.6% on MIMIC-IV. Predicted SQA labels closely matched manual annotations, with mean absolute error (MAE) differences of 0.08 bpm (SR＋AF), 0.25 bpm (SR), 0.62 bpm (AF), and 0.53 bpm (MIMIC-IV). SDSD reduced MAE by 46.57% for SR＋AF, 41.67% for SR, and 49.69% for AF, further demonstrating the effectiveness of integrating SQA into HR estimation workflows.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108688"},"PeriodicalIF":4.9,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145157653","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":"Consistent assessment of heart rate recovery across exercise intensities","authors":"Byungmun Kang ,&nbsp;DaeEun Kim ,&nbsp;Sungjoon Yoon ,&nbsp;Dongwoo Kim ,&nbsp;Hwang-Jae Lee ,&nbsp;Dokwan Lee ,&nbsp;YoonMyung Kim","doi":"10.1016/j.bspc.2025.108690","DOIUrl":"10.1016/j.bspc.2025.108690","url":null,"abstract":"<div><div>Accurate assessment of heart rate (HR) recovery is important for evaluating cardiorespiratory function and endurance capacity. Conventional approaches – such as 1 min or 2 min HR decline exponential fits – often prioritize fitting precision, yet can show substantial variability across individuals and exercise intensities, limiting their broader applicability. In this study, we examined a Decay Time Constant (Decay TC) derived from a first-order differential model applied to HR data scaled between exercise termination and 2 min post-exercise. Thirty-five healthy adults performed robot-resisted knee-up exercises at three intensities (72, 84, and 96 RPM), with HR continuously monitored via a wireless chest sensor. Normalization based on the first-order model reduced the influence of differing starting HR values and recovery slopes, enabling the Decay TC to reflect recovery characteristics that remained relatively consistent across intensities. Correlation analysis – performed overall and by sex and age group – showed that this Decay TC maintained more stable relationships with submaximal VO₂max than conventional HR recovery indicators, with moderate-to-strong correlations (<span><math><mrow><mo>|</mo><mi>R</mi><mo>|</mo></mrow></math></span> up to 0.93). Multivariable regression confirmed it as a significant predictor, but the aim was not to maximize VO₂max prediction accuracy or optimize curve-fitting, but rather to provide a simple, interpretable measure that captures an individual’s consistent recovery profile as a potential physiological signature. These findings suggest that the Decay TC obtained from scaled HR data offers a practical metric for characterizing HR recovery dynamics, with potential for integration into endurance assessment protocols and wearable health monitoring systems.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108690"},"PeriodicalIF":4.9,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145157777","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":"The efficacy prediction of subcutaneous immunotherapy for pediatric allergic Rhinitis: Application of machine learning methods","authors":"Danjing Wang ,&nbsp;Qingqing Lv ,&nbsp;Hao Yao ,&nbsp;Yi Chen ,&nbsp;Jiahui Yu ,&nbsp;Xiaohong Jin ,&nbsp;Huiling Chen ,&nbsp;Weixi Zhang","doi":"10.1016/j.bspc.2025.108704","DOIUrl":"10.1016/j.bspc.2025.108704","url":null,"abstract":"<div><div>Allergen immunotherapy (AIT) is an effective treatment for allergic rhinitis (AR); however, some patients do not respond optimally. This study aims to identify the factors influencing the efficacy of subcutaneous specific immunotherapy (SCIT) for AR in children and to develop a predictive model for treatment outcomes using machine learning techniques. Data were collected from 272 children aged 4–15 years with AR, excluding those with asthma, who underwent more than three years of mite SCIT at two hospitals in southern Zhejiang. Patients were categorized into effective and ineffective groups based on the improvement in the Combined Symptom Medication Score (CSMS) before and after treatment. The data were split into a training set and a testing set, and the bIRSCA algorithm was applied to identify optimal feature subsets in the training set. The selected features were then used in a support vector machine (SVM) model to assess performance on the testing set, with ten-fold cross-validation applied to evaluate the model. The final results were based on the average performance metrics across ten iterations. The bIRSCA-SVM model identified key biomarkers, including the sIgE/tIgE (Der p) ratio, sIgE (Der p), eosinophil count, eosinophil ratio, and the sIgE/tIgE (Der f) ratio, as significant predictors of therapeutic efficacy. The model achieved an accuracy of 88.992 %, sensitivity of 99.736 %, and specificity of 86.872 %, outperforming other models. In conclusion, a positive response to SCIT is associated with baseline levels of the identified biomarkers. The bIRSCA-SVM model provides an effective and accurate method for predicting the efficacy of mite SCIT in children with allergic rhinitis.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108704"},"PeriodicalIF":4.9,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145219701","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":"Spatio-temporal representation learning with selective state space models for EEG-based depression detection","authors":"Yutao Dou ,&nbsp;Tao Xing ,&nbsp;Xiongjun Zhao ,&nbsp;Xianliang Chen ,&nbsp;Jiansong Zhou ,&nbsp;Shaoliang Peng","doi":"10.1016/j.bspc.2025.108707","DOIUrl":"10.1016/j.bspc.2025.108707","url":null,"abstract":"<div><div>Electroencephalogram signals, as key cognitive biomarrs, capture subtle brain activities and are crucial for diagnosing mental disorders like depression. However, owing to the multi-channel and high sampling rate of acquisition devices, EEG data exhibit high dimensionality and long sequences. Most existing studies focus on analyzing individual domains, such as temporal, structural, or state features, making it challenging to effectively capture and represent the correlations among these features across multiple channels while avoiding the loss of key information in long sequential data. In addition, variations in patients’ conditions result in differences in detection times, further increasing the complexity of data processing. To tackle these challenges, we propose the TSS-SSM framework, which combines spatio-temporal representation learning of temporal, structural, and state correlations with selective state-space model to effectively handle the complex features of EEG signals. First, by segmenting EEG signals into adaptive time slices and using multiple GCNs, we effectively extracted structural relationships between brain regions. The integration of LSTM networks and attention mechanisms enabled us to model the historical states of EEG segments and retain critical information from past states in continuous time sequences. Then, by integrating SSM and a selection mechanism, our model highlights important brain activity events and prevents them from being overlooked in long sequences. Experimental results on the public MODMA dataset and the real-world dataset from Xiangya Hospital demonstrate that TSS-SSM achieved significant performance improvements, with ACC values of 0.9481 and 0.8836, respectively, and its effectiveness was further validated through extensive ablation studies.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108707"},"PeriodicalIF":4.9,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145219585","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":"Research on stack-based LSTM adaptive ECG mapping method","authors":"Yuyu Shen ,&nbsp;Yingwei Li ,&nbsp;Yujia Wang ,&nbsp;Tingting Lu ,&nbsp;Ruihua Cao ,&nbsp;Huiquan Wang ,&nbsp;Feng Cao","doi":"10.1016/j.bspc.2025.108727","DOIUrl":"10.1016/j.bspc.2025.108727","url":null,"abstract":"<div><div>Sudden cardiovascular diseases impose a significant burden on individuals’ health and life, with high mortality and disability rates. Most portable electrocardiogram collection devices collect ECG signals in vector ECG format, which differs from standard ECG signals, making it difficult for doctors to diagnose diseases. To address this issue, we designed a human-engineered “palm” rapid ECG collection system based on flexible sensing materials. Additionally, we implemented an individualized adaptive ECG mapping algorithm using a stack LSTM network to map non-standard ECG signals collected by the portable ECG collection front-end to standard signals. To evaluate the performance of our approach, we conducted a comparative analysis experiment on ECG data collected from 30 participants. Our results show that the correlation between the “palm” rapid ECG graph obtained using our proposed mapping algorithm and the standard 12-lead ECG graph was 97.45 %, with a RMSE of 0.09 mV. These findings indicate that our approach has significant implications for optimizing signal analysis of wearable ECG collection devices.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108727"},"PeriodicalIF":4.9,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145219702","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":"ECGtizer: An open-source, fully automated pipeline for digitization and signal recovery from paper electrocardiograms","authors":"Alex Lence ,&nbsp;Ahmad Fall ,&nbsp;Samuel David Cohen ,&nbsp;Federica Granese ,&nbsp;Jean-Daniel Zucker ,&nbsp;Joe-Elie Salem ,&nbsp;Edi Prifti","doi":"10.1016/j.bspc.2025.108710","DOIUrl":"10.1016/j.bspc.2025.108710","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Electrocardiograms (ECGs) are essential for diagnosing cardiac pathologies, yet traditional paper-based ECG storage poses significant challenges for automated analyses. Despite the growing interest in leveraging AI for ECG analysis, there remains a lack of accessible, fully automated tools for digitizing paper-based ECGs. Existing solutions are often incomplete, behind paywalls, or not suited for large-scale use. To address this gap, we present <span>ECGtizer</span>: an open-source, fully automated tool that enables high-fidelity digitization of paper ECGs, ensuring long-term preservation of clinical data and unlocking their potential for modern AI-driven analysis.</div></div><div><h3>Methods:</h3><div><span>ECGtizer</span> employs automated lead detection, three different pixel-based signal extraction algorithms, and a deep learning-based signal reconstruction module. We evaluated <span>ECGtizer</span> on two datasets: a real-life cohort from the COVID-19 pandemic (JOCOVID) and a publicly available dataset (PTB-XL). Performance was compared with two existing methods: the fully automated <span>ECGminer</span> and the semi-automated <span>PaperECG</span>, which requires human intervention. The tools’ digitization performance was assessed in terms of signal recovery, the fidelity of clinically relevant feature measurement and downstream AI classification tasks on a third dataset (GENEREPOL).</div></div><div><h3>Results:</h3><div>Results show that <span>ECGtizer</span> outperforms state-of-the-art methods, with its <span>ECGtizer</span> <span><math><msub><mrow></mrow><mrow><mtext>Frag</mtext></mrow></msub></math></span> algorithm delivering superior signal recovery performance. While <span>PaperECG</span> demonstrated better outcomes than <span>ECGminer</span>, it also requires human input.</div></div><div><h3>Conclusions:</h3><div><span>ECGtizer</span> enhances the usability of historical ECG data and supports advanced AI-based diagnostic methods, making it a valuable addition to the field of AI in ECG analysis.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":"112 ","pages":"Article 108710"},"PeriodicalIF":4.9,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145157660","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}