Biomedical Physics & Engineering Express最新文献

{"title":"Wavelet-Based Energy Mapping Analysis Framework for Magnetocardiography.","authors":"Kun Yang, Zhidan Zhang, Hai Wang, Xiangyan Kong, Peizhi Miao, Zhenfei Chen","doi":"10.1088/2057-1976/aeb67c","DOIUrl":"https://doi.org/10.1088/2057-1976/aeb67c","url":null,"abstract":"<p><p>Magnetocardiography (MCG) measures the magnetic fields produced by the cardiac electrical activity, exhibiting remarkable temporal and spatial resolution. However, its clinical utility has been constrained by limitations in analytical methods and the lack of extensive clinical validation. In this work, a comprehensive MCG analysis framework based on time-frequency domain energy map was proposed. To extract true MCG signals, raw multi-channel recordings were denoised by magnetic compensation and digital filter. The wavelet transform was then applied to decompose the filtered MCG signals into six levels corresponding to specific frequency bands, and energy maps of three segments further outlined and highlighted spatial features. Experimental results suggested a potential variation in the similarity coefficient of the energy maps associated with MI. This study introduces a novel approach for analyzing frequency-dependent energy map of the MCG, offering valuable insights into myocardial abnormalities diagnosis.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149227055","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Adaptive electrocardiogram (ECG) signal denoising using particle swarm optimization-based non-local means (NLM) filtering.","authors":"Kiran Jash, Nanda Dulal Jana, Swarup Kumar Laha","doi":"10.1088/2057-1976/aeb129","DOIUrl":"10.1088/2057-1976/aeb129","url":null,"abstract":"<p><p>An ECG signal is usually corrupted by AWGN (additive white Gaussian noise), baseline wandering (BW), power line interference (PLI), and motion artifacts (EM), which may impact the reliable diagnosis of arrhythmias. This study proposes a hybrid ECG denoising framework that integrates particle swarm optimization with non-local means filtering to suppress diverse ECG artifacts while preserving signal morphology. In the first experiment, ECG signals from MIT-BIH Arrhythmia Database are artificially contaminated using AWGN and PLI to measure their denoising capability. In the second experiment, real-world BW, muscle artifacts, and electrode EM from the MIT-BIH noise stress test database are used to evaluate the robustness of the proposed framework under practical noise conditions. The proposed method shows improved denoising performance in SNR, PRD, and RMSE for both experiments. Morphology preservation analysis represents minimal distortion of the waveforms and preservation of important ECG characteristics. Furthermore, a one-dimensional ResNet architecture is used for arrhythmia classification, demonstrating good classification performance on the denoised ECG signals with only a minor reduction compared with the noise-free signals. Overall, the proposed framework effectively suppresses diverse ECG artifacts while preserving diagnostically relevant information for arrhythmia classification.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149126718","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Fabrication, testing, and clinical validation of an epoxy resin-based range shifter for pencil beam scanning proton therapy.","authors":"Kantaram Darekar, Bhushankumar J Patil, Akshay Ashok Bhavke, Umesh Bharat Gayake, Vysakh Raveendran, Rohidas Namdeo Punde, Prakash Hari Shinde, Sanjay D Dhole, Chithambara Prabu Arunasalam, Lalit Chaudhari, Siddhartha Laskar, Reena D Phurailatpam","doi":"10.1088/2057-1976/aeb678","DOIUrl":"https://doi.org/10.1088/2057-1976/aeb678","url":null,"abstract":"<p><strong>Background and objectives: </strong>Range shifters (RS) are used to treat shallow targets in pencil beam scanning (PBS) proton therapy. This study aims to validate an in-house-fabricated epoxy resin RS through mechanical and dosimetric measurements, evaluate the Monte Carlo (MC) algorithm dose‑calculation accuracy of the treatment planning system (TPS) beam model incorporating this device, and assess its clinical feasibility for PBS proton therapy applications.</p><p><strong>Methods: </strong>The RS was fabricated using a silicone mould and machined to the required dimensions after curing. It was further evaluated through material characterization, mechanical testing, and dosimetric validation. Material characterization included determining the elemental composition, mass density, and mean excitation energy, which were incorporated into the beam modelling. Mechanical validation comprised dimensional verification, fitment testing, and assessment of material integrity and uniformity. Dosimetric validation was performed by measuring water-equivalent thickness (WET), range parameters, spot size, absolute dose, and fluence, and comparing them with TPS-MC calculations.</p><p><strong>Results: </strong>The cured epoxy resin exhibited normalized elemental composition of H:8.0%, C:65.0%, N:0.1%, and O:26.9%, with mass density 1.15 g/cm³ and mean excitation energy 72.5 eV. The fabricated RS matched the reference design within ±1 mm and demonstrated satisfactory fitment, reproducibility, material integrity and uniformity. The measured WET agreed with TPS estimated values within 0.06 g/cm², while differences in range parameters were <1 mm. Measured spot sizes and the 9-spot matrix showed excellent agreement with TPS estimates across different air-gap settings. Absolute dose measurements agreed with TPS-MC calculations within 3%, and patient-specific fluence measurements showed excellent agreement with TPS calculations under stringent gamma evaluation criteria.</p><p><strong>Conclusion: </strong>This study presents the first comprehensive clinical validation and evaluation of cured solid epoxy resin as an RS material for PBS proton therapy applications. The in-house-fabricated, cured epoxy resin-based RS demonstrated excellent agreement with TPS-MC calculations and experimental measurements, establishing its clinical feasibility.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149227021","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"easyPET.3D Measurement of Positron Emitter Maps in Proton-Irradiated PMMA Phantoms.","authors":"Ana Catarina Monteiro Magalhães, Pedro Manuel Mendes Correia, Martin Grossmann, David Meer, Joao F C A Veloso, Ana Silva","doi":"10.1088/2057-1976/aeb679","DOIUrl":"https://doi.org/10.1088/2057-1976/aeb679","url":null,"abstract":"<p><p>Accurate range verification is a critical component of proton therapy, ensuring that the high dose gradients characteristic of Bragg peaks are delivered precisely to the target volume. This work presents the first experimental evaluation of the easyPET.3D, a compact and cost-effective preclinical PET scanner, as a potential tool for quality control assessement. Homogeneous PMMA phantoms were irradiated at the Paul Scherrer Institute Gantry 1 facility using proton beam energies ranging from 138 MeV to 195 MeV. The resulting 3D β+ activity distributions were reconstructed and analyzed to assess the system's spatial accuracy, stability, and sensitivity. For the intermediate energy of 160 MeV, the measured activity profile showed a high degree of agreement with Monte Carlo (GATE) simulations, with a difference of 0.24 mm in the R_50%position. The system exhibited robust performance under varying irradiation conditions. Doubling the number of delivered protons relative to a clinically relevant value resulted in a deviation of only 0.4 mm in the calculated range, confirming the method's reproducibility. Furthermore, the system successfully resolved distinct beam paths in a dual-energy irradiation mode, distinguishing tracks separated vertically by 10 mm and quantifying longitudinal range differences.Sensitivity tests indicated that the system is capable of detecting longitudinal phantom displacements as small as 1mm. The relationship between expected and measured displacements followed a linear trend with a slope of 1.13, indicating a systematic overestimation of 13%, and an offset of 0.45mm, reflecting a small baseline bias. While further optimization of detector sensitivity and reconstruction algorithms is required to reduce acquisition times for routine clinical workflows, these findings validate the potential of easyPET.3D as a promising, accessible solution for proton range verification.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149227049","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"ADHDeepNet: An Explainable Framework for EEG-Based ADHD Detection From Raw Signals.","authors":"Ali Amini, Mohammad Alijanpour Shalmani, Behnam Latifi, Ali Motie Nasrabadi","doi":"10.1088/2057-1976/aeb67a","DOIUrl":"https://doi.org/10.1088/2057-1976/aeb67a","url":null,"abstract":"<p><p>Attention Deficit Hyperactivity Disorder (ADHD) is a common brain disorder in children that can persist into adulthood, affecting social, academic, and career life. Early diagnosis is crucial for managing these impacts on patients and the healthcare system but is often labor-intensive and time-consuming. This paper presents a novel method to improve ADHD diagnosis precision and timeliness by leveraging Deep Learning (DL) approaches and electroencephalogram (EEG) signals. We introduce ADHDeepNet, a deep learning model that utilizes comprehensive temporal-spatial characterization, attention modules, and explainability techniques optimized for EEG signals. ADHDeepNet integrates feature extraction and refinement processes to enhance ADHD diagnosis. The model was trained and validated on a dataset of 121 participants (61 ADHD, 60 Healthy Controls), employing nested cross-validation for robust performance. The proposed two-stage methodology uses a 10-fold cross-subject validation strategy. Initially, each iteration optimizes the model's hyper-parameters with inner 2-fold cross-validation. Then, Additive Gaussian Noise (AGN) with various standard deviations and magnification levels is applied for data augmentation. ADHDeepNet achieved 100% sensitivity and 99.17% accuracy in classifying ADHD/HC subjects. To clarify model explainability and identify key brain regions and frequency bands for ADHD diagnosis, we analyzed the learned weights and activation patterns of the model's primary layers. Additionally, t-distributed Stochastic Neighbor Embedding (t-SNE) visualized high-dimensional data, aiding in interpreting the model's decisions. This study highlights the potential of deep learning and EEG in enhancing ADHD diagnosis accuracy and efficiency.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149227012","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Fuzzy-Enhanced Broad Learning Framework with Minority-Class Ensemble for ECG Arrhythmia Classification.","authors":"Qi Feng, Ao-Hua Yu, Du-Cheng Shi, Xiao-Li Wang","doi":"10.1088/2057-1976/aeb67b","DOIUrl":"https://doi.org/10.1088/2057-1976/aeb67b","url":null,"abstract":"<p><p>Existing Broad Learning System (BLS)-based Electrocardiogram (ECG) arrhythmia classification methods still face limitations in multi-source feature representation, sample reliability modeling, and minority-class recognition. To address these issues, this study proposes a combinational improvement framework built upon BLS, integrating uncertainty aware gated fuzzy feature enhancement and a minority-class ensemble strategy for ECG arrhythmia classification. First, heartbeat waveform, QRS complex structural information, and RR rhythm features are integrated to construct a hybrid ECG representation that jointly captures morphological and rhythm characteristics. The BLS is then employed for high-dimensional nonlinear feature mapping, upon which an uncertainty-aware gated fuzzy enhancement mechanism evaluates sample reliability through robust class prototype estimation and fuzzy membership calculation. A confidence-based gating strategy is further introduced to achieve adaptive feature enhancement, strengthening discriminative information from reliable samples while reducing noise interference caused by ambiguous boundary samples. Finally, a minority-class-aware ensemble classification strategy is developed, which employs eXtreme Gradient Boosting (XGBoost) as the primary classifier. One-versus-Rest(OvR) classifiers, a complementary Logistic Regression (LR) model, and class-wise probability calibration are incorporated to refine classification decisions and alleviate majority class bias caused by class imbalance. Experimental results demonstrate that the proposed method achieves overall accuracies of 99.04% and 92.58% (Macro-F1 of 93.39% and 61.60%) under the intra-patient and inter-patient evaluation protocols, respectively. The framework trades a moderate training-time and memory footprint (an ensemble of tree models over ANOVA-selected BLS features) for these minority-class gains, while inference remains a CPU-friendly single forward pass through the frozen pipeline. It improves minority-class recognition relative to the compared methods under severe class imbalance, while the recognition of fusion beats under the inter-patient protocol remains challenging(five-seed mean F1 of 6.72%±2.69%)and is explicitly discussed as a limitation.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149227026","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"B-spline Kolmogorov-Arnold networks for interpretable semi-supervised ECG classification.","authors":"M R Baidillah, M Rahmawati, S Oktamuliani, R Febryarto, P Busono, I M Astawa, S Karim, R Riyanto, C Achmad","doi":"10.1088/2057-1976/aeb0af","DOIUrl":"10.1088/2057-1976/aeb0af","url":null,"abstract":"<p><p>This study investigates how Kolmogorov-Arnold Networks (KANs) behave when combined with semi-supervised learning for single-lead ECG analysis. Three architectures are compared: a standard CNN-BiLSTM-Attention model (accurate but uninterpretable), a minimal KAN with a simple CNN front-end (ablation), and the proposed SE-MSC-KAN-CNN. The proposed model uses multi-scale depthwise convolutions, Squeeze-and-Excitation attention, BiLSTM, and a B-spline KAN head. Interpretability is evaluated in three ways: visualising learned B-spline functions, applying Grad-CAM to highlight important signal regions, and measuring how the splines change when fewer labels are used (spline drift). On the PhysioNet 2017 dataset (8528 recordings, four classes), the proposed model reaches 85.2% supervised accuracy (<i>F</i>1 = 0.851), outperforming the CNN baseline (83.1%) and matching much larger black-box models. With only 90% of the labels, it still achieves 85.1%. After training, we applied post-training TFLite INT8 conversion. Only the CNN-front-end KAN converted (measured size 422 KB). The proposed SE-MSC-KAN-CNN did not convert, because Bidirectional LSTM and custom attention lack TFLite kernels; its float32 footprint is about 9.8 MB (theoretical INT8 size about 806 KB). Classification metrics in this paper are float32; INT8 test accuracy was not measured. The spline drift metric summarises how KAN coefficients change under semi-supervised training; use as an on-device out-of-distribution detector is proposed but not validated here. These results show that competitive accuracy and inspectable KAN maps can be obtained in one hybrid architecture. Wearable deployment of the proposed SE-MSC-KAN-CNN is not yet achieved: TFLite INT8 conversion failed, and the present graph still requires float32 inference.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149103893","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Wavelet-based adaptive image fusion of CT and MRI with local neighborhood variance for nasopharyngeal carcinoma.","authors":"Xinjian Qin, Haojiang Li, Shanshan Liu, Pengxing Huang, Lizhi Liu, Hong-Bo Chen, Xin Yang","doi":"10.1088/2057-1976/aeb551","DOIUrl":"https://doi.org/10.1088/2057-1976/aeb551","url":null,"abstract":"<p><p>Multimodal imaging of nasopharyngeal carcinoma (NPC) requires complementary representation of CT-derived bony anatomy and MRI-derived soft-tissue contrast. Existing fusion methods may fail to preserve both structural detail and soft-tissue information in a transparent and reproducible manner. We present a wavelet-based fusion framework that employs local neighborhood variance for adaptive weighting. This interpretable variance-driven strategy selectively integrates CT bony structures and MRI soft-tissue textures according to regional information content. The neighborhood configuration was selected through an ablation analysis of 232 clinical cases. Quantitative evaluation showed improved image-quality performance over the selected comparison methods across entropy, spatial frequency, average gradient, standard deviation, and edge information preservation factor, with respective gains of approximately 5.5%, 2.4%, 6.7%, 5.1%, and 3.2% relative to the best-performing baseline for each metric. These findings support the proposed method as an image fusion methodology for preserving complementary multimodal information; direct effects on target delineation, workflow efficiency, or dosimetric outcomes require future clinical validation.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149206490","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Patient-specific dosimetric impact of online adaptive radiotherapy for locally advanced cervical cancer.","authors":"Jasmine Penney, Krista Chytyk-Praznik","doi":"10.1088/2057-1976/aeb0b0","DOIUrl":"10.1088/2057-1976/aeb0b0","url":null,"abstract":"<p><p><i>Background:</i>When cervical cancer patients are treated with external beam radiation therapy, patient-specific anatomical changes, particularly bladder filling, alter the position and shape of the target volume throughout treatment. Online adaptive radiotherapy (oART) accounts for these interfractional changes by generating a treatment plan based on the patient's daily anatomy.<i>Purpose:</i>This study evaluated the dosimetric impact of oART and investigated the relationship between patient-specific anatomical changes and the dosimetric of benefit of adaptation using standard cone-beam CT (CBCT) images for locally advanced cervical cancer (LACC).<i>Methods:</i>Ten patients previously treated for LACC at our institution were identified. A reference treatment plan incorporating a 5 mm clinical target volume (CTV)-to-planning target volume (PTV) margin was generated within the planning system simulator. Three CBCT images (Fractions 1, 13, and 25) were used to retrospectively simulate an adaptive treatment session. For each CBCT, the reference plan was recalculated on the daily anatomy to generate a scheduled plan, and an adapted plan was generated for comparison based on target coverage and organ at risk (OAR) dose metrics. The relationship between interfraction anatomical variation and the dosimetric benefit of adaptation was evaluated.<i>Results:</i>Adapted plans improved target coverage while generally maintaining or reducing OAR dose relative to the scheduled plans. However, the magnitude of benefit from adaptation varied between patients. Moderate associations were observed between interfraction anatomical variation of the bladder volume change and improvements in target coverage.<i>Conclusion:</i>The dosimetric benefit of oART was patient-specific and was partially associated with interfraction anatomical variation. These findings support an individualized approach to identify patient-specific characteristics that may help determine what patients would benefit the most from adaptive radiotherapy.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149103471","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A multi-atlas ensemble learning framework for autism spectrum disorder diagnosis.","authors":"Shengbin Liang, Zhe Zhang, Wencai Du, Feng Gu","doi":"10.1088/2057-1976/aeb128","DOIUrl":"10.1088/2057-1976/aeb128","url":null,"abstract":"<p><p><i>Objective.</i>Clinical autism spectrum disorder (ASD) relies on subjective scales and neuroimaging, with traditional approaches suffering from high subjectivity and insufficient feature extraction, increasing misdiagnoses risk.<i>Approach.</i>To improve the diagnostic accuracy, we propose a multimodal ASD prediction network (APN) that integrates resting-state functional magnetic resonance imaging with non-imaging data. Algorithmically, we first introduce a novel cumulative distribution divergence score (CDDS) feature selection method based on the multi-atlas functional connectivity (FC), which quantifies statistical distributional differences between positive and negative samples, effectively addressing the limitations of the traditional<i>F</i>-score in overlapping or weakly discriminative features. Building upon this, we adapt existing deep learning (DL) paradigms to construct a deep neural architecture (variational autoencoder-denoising autoencoder (VAEA)), combining a variational autoencoder (VAE), a denoising autoencoder, and a multilayer perceptron. The combination of these modules is non-trivial: the novel CDDS first purifies the high-dimensional feature space, enabling the VAEA to learn robust latent probabilistic representations via unsupervised pre-training without suffering from the curse of dimensionality. This is followed by the supervised fine-tuning to optimize parameters. We further incorporate demographic factors including age, full-scale intelligence quotient, and gender via a late-fusion strategy to prevent them from being overshadowed by imaging features, thereby enhancing the discriminability. FC features from three atlases including AAL, CC400, and Dosenbach160 are weighted and integrated to achieve the effective multi-source fusion.<i>Main results.</i>We conduct 10-fold cross-validations on the Autism Brain Imaging Data Exchange (ABIDE) dataset, and the proposed APN achieves 81.63% accuracy, 81.89% sensitivity and 81.46% specificity, significantly outperforming state-of-the-art methods. Interpretability analysis identifies hippocampus, amygdala, and cingulate gyrus as strongly ASD-associated regions, offering neural mechanistic insights.<i>Significance.</i>This study provides an efficient, objective, and interpretable multimodal DL framework for ASD assisted diagnosis, with substantial clinical application and neuroscience value.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149126696","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}