{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T15:02:38Z","timestamp":1784818958316,"version":"3.55.0"},"reference-count":46,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T00:00:00Z","timestamp":1784764800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Neurol."],"abstract":"<jats:sec>\n                    <jats:title>Objective<\/jats:title>\n                    <jats:p>Unlike our previously published Graphics Processing Unit-based (GPU) Dynamic Pathway-Enhanced Attention Network (DPEA-Net) which is optimized for radiotherapy planning, state-of-the-art GPU-dependent brain tumor segmentation models cannot be deployed in primary care clinics, community hospitals, or intraoperative Magnetic Resonance Imaging (MRI) settings where dedicated GPU hardware is unavailable. This hardware barrier leads to delayed detection, missed urgent referrals, and prolonged triage in resource-limited environments. Existing lightweight models still suffer from slow Central Processing Unit (CPU) inference and inadequate boundary quality for rapid clinical decision-making.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>We propose Gated Multi-scale Attention Network (GMANet), an ultra-lightweight 0.20M-parameter 3D CNN explicitly engineered for CPU-only real-time glioma triage. To achieve extreme deployability under strict computational constraints, we introduce three synergistic modules: a gated multi-scale dilated (GMSD) block for efficient multi-scale feature extraction, a full-dimension attention (FDA) module for 3D anatomical continuity, and a learnable transposed upsampling (LTU) module for CPU-optimized boundary recovery. The moderate reduction in enhancing tumor (ET) precision represents a deliberate trade-off to prioritize rapid whole tumor (WT) and tumor core (TC) assessment for initial screening.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>On BraTS2019 and BraTS2021, GMANet achieves WT Dice of 92.45% and 90.93%, TC Dice of 86.28% and 84.72%, and ET Dice of 78.42% and 76.68%, respectively, with only 20.10 Giga Floating-Point Operations (GFLOP). On a standard laptop CPU, inference time is 8.49 s per volume.The ET Dice exceeds the commonly adopted threshold for brain tumor preliminary screening in research practice. This performance renders the model potentially applicable to preliminary triage, pending further prospective clinical validation.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>GMANet trades marginal ET precision for extreme CPU deployability and sub-10-second inference, acting as a potential complementary tool to our GPU-based DPEA-Net. GMANet is designed to support a two-stage clinical pathway: fast CPU screening followed by GPU precision assessment, which requires further implementation and clinical evaluation. Importantly, GMANet is not an incremental improvement of our previously published GPU-based DPEA-Net but a complementary model targeting a distinct clinical scenario where GPUs are unavailable.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fneur.2026.1894796","type":"journal-article","created":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T14:19:35Z","timestamp":1784816375000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["GMANet: a CPU-only ultra-lightweight 3D CNN for real-time glioma triage in resource-limited environments"],"prefix":"10.3389","volume":"17","author":[{"given":"Caijian","family":"Hua","sequence":"first","affiliation":[{"name":"School of Computer Science, Sichuan University of Science and Engineering","place":["Zigong, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xuerong","family":"Jing","sequence":"additional","affiliation":[{"name":"School of Computer Science, Sichuan University of Science and Engineering","place":["Zigong, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Liuying","family":"Li","sequence":"additional","affiliation":[{"name":"Traditional Chinese Medicine Department, Zigong First People's Hospital","place":["Zigong, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xia","family":"Zhou","sequence":"additional","affiliation":[{"name":"Traditional Chinese Medicine Department, Zigong First People's Hospital","place":["Zigong, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,7,23]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"110406","DOI":"10.1016\/j.bspc.2026.110406","article-title":"Bridging transformers and state-space models: an efficient hybrid framework for 3D brain tumor segmentation","volume":"122","author":"Rahim","year":"2026","journal-title":"Biomed Signal Process Control"},{"key":"B2","doi-asserted-by":"publisher","first-page":"1474461","DOI":"10.3389\/fneur.2024.1474461","article-title":"Utilizing machine-learning techniques on MRI radiomics to identify primary tumors in brain metastases","volume":"15","author":"Yang","year":"2025","journal-title":"Front Neurol"},{"key":"B3","doi-asserted-by":"publisher","first-page":"20","DOI":"10.21037\/jmai-2025-104","article-title":"The financial challenges of artificial intelligence integration in healthcare: lessons from the past and considerations for the future","volume":"9","author":"Akula","year":"2026","journal-title":"J Med Artif Intell"},{"key":"B4","doi-asserted-by":"publisher","first-page":"102127","DOI":"10.1016\/j.jmir.2025.102127","article-title":"Artificial intelligence in medical imaging: utilization, challenges, and practitioner perceptions in Rwanda","volume":"57","author":"Mukandayisenga","year":"2026","journal-title":"J Med Imaging Radiat Sci"},{"key":"B5","doi-asserted-by":"publisher","first-page":"559","DOI":"10.1093\/nop\/npab040","article-title":"A multidisciplinary neuro-oncological triage panel reduces the time to referral and treatment for patients with a brain tumor","volume":"8","author":"De Swart","year":"2021","journal-title":"Neuro-oncol Pract"},{"key":"B6","doi-asserted-by":"publisher","first-page":"1581422","DOI":"10.3389\/fneur.2025.1581422","article-title":"Artificial intelligence in the task of segmentation and classification of brain metastases images: current challenges and future opportunities","volume":"16","author":"Hu","year":"2025","journal-title":"Front Neurol"},{"key":"B7","doi-asserted-by":"publisher","first-page":"1838","DOI":"10.3390\/s25061838","article-title":"Advancing precision: a comprehensive review of MRI segmentation datasets from brats challenges (2012-2025)","volume":"25","author":"Bonato","year":"2025","journal-title":"Sensors"},{"key":"B8","doi-asserted-by":"publisher","first-page":"116000","DOI":"10.1016\/j.knosys.2026.116000","article-title":"Explainable brain tumor segmentation via attention-guided hybrid CNN-Transformer-Mamba network","volume":"343","author":"Tehsin","year":"2026","journal-title":"Knowl-Based Syst"},{"key":"B9","doi-asserted-by":"publisher","first-page":"1832698","DOI":"10.3389\/fneur.2026.1832698","article-title":"DPEA-Net: a clinically-oriented lightweight 3D CNN for glioma segmentation in multiparametric MRI","volume":"17","author":"Hua","year":"2026","journal-title":"Front Neurol"},{"key":"B10","first-page":"234","article-title":"U-net: convolutional networks for biomedical image segmentation","volume-title":"International conference on medical image computing and computer-assisted intervention","author":"Ronneberger","year":"2015"},{"key":"B11","doi-asserted-by":"publisher","first-page":"200275","DOI":"10.1016\/j.mrl.2026.200275","article-title":"A hybrid deep learning framework based on VGG19 and U-Net for accurate brain tumor segmentation in MRI images","volume":"6","author":"Aslan","year":"2026","journal-title":"Magn Reson Lett"},{"key":"B12","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1511.07122","article-title":"Multi-scale context aggregation by dilated convolutions","author":"Yu","year":"2015","journal-title":"arXiv Preprint arXiv:151107122"},{"key":"B13","doi-asserted-by":"publisher","first-page":"203","DOI":"10.17694\/bajece.1617698","article-title":"Performance comparison of deep learning models in brain tumor classification","volume":"13","author":"Aslan","year":"2025","journal-title":"Balkan J Electr Comput Eng"},{"key":"B14","doi-asserted-by":"publisher","first-page":"106911","DOI":"10.1016\/j.bspc.2024.106911","article-title":"TDPC-Net: multi-scale lightweight and efficient 3D segmentation network with a 3D attention mechanism for brain tumor segmentation","volume":"99","author":"Li","year":"2025","journal-title":"Biomed Signal Process Control"},{"key":"B15","doi-asserted-by":"publisher","first-page":"538","DOI":"10.3390\/jcm12020538","article-title":"HMNet: hierarchical multi-scale brain tumor segmentation network","volume":"12","author":"Zhang","year":"2023","journal-title":"J Clin Med"},{"key":"B16","doi-asserted-by":"publisher","first-page":"223","DOI":"10.1007\/978-3-032-16370-7_20","article-title":"GAMMA-Net: Gated Attention Multi-scale Modular Architecture","author":"Arora","year":"2025","journal-title":"International conference on medical image computing and computer-assisted intervention"},{"key":"B17","doi-asserted-by":"publisher","first-page":"104296","DOI":"10.1016\/j.bspc.2022.104296","article-title":"MBANet: a 3D convolutional neural network with multi-branch attention for brain tumor segmentation from MRI images","volume":"80","author":"Cao","year":"2023","journal-title":"Biomed Signal Process Control"},{"key":"B18","doi-asserted-by":"publisher","first-page":"107221","DOI":"10.1016\/j.bspc.2024.107221","article-title":"An attention-fused architecture for brain tumor diagnosis","volume":"101","author":"Hekmat","year":"2025","journal-title":"Biomed Signal Process Control"},{"key":"B19","doi-asserted-by":"publisher","first-page":"98","DOI":"10.1097\/JCMA.0000000000001188","article-title":"Comparison of machine learning algorithms for automatic prediction of Alzheimer disease","volume":"88","author":"Aslan","year":"2025","journal-title":"J Chin Med Assoc"},{"key":"B20","first-page":"7132","article-title":"Squeeze-and-excitation networks","volume-title":"Proceedings of the IEEE conference on computer vision and pattern recognition","author":"Hu","year":"2018"},{"key":"B21","first-page":"3","article-title":"Cbam: convolutional block attention module","volume-title":"Proceedings of the European conference on computer vision (ECCV)","author":"Woo","year":"2018"},{"key":"B22","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1007\/s10278-026-01948-6","article-title":"Attention-driven deep learning for hydrocephalus: preliminary benchmarking of multiscale 3D attention against conventional 3D CNN models using ventricular enlargement on CT","author":"Sekkat","year":"2026","journal-title":"J Imag Inform Med"},{"key":"B23","doi-asserted-by":"publisher","first-page":"2931556","DOI":"10.1155\/int\/2931556","article-title":"Explainability-aligned reliability-weighted fuzzy ensemble for automated cervical cancer classification","volume":"2026","author":"Alpsalaz","year":"2026","journal-title":"Int J Intell Syst"},{"key":"B24","doi-asserted-by":"publisher","first-page":"80","DOI":"10.1038\/s41698-024-00575-0","article-title":"Artificial intelligence in neuro-oncology: advances and challenges in brain tumor diagnosis, prognosis, and precision treatment","volume":"8","author":"Khalighi","year":"2024","journal-title":"NPJ Precis Oncol"},{"key":"B25","doi-asserted-by":"publisher","first-page":"72","DOI":"10.1007\/s10586-024-04880-y","article-title":"Detection of road extraction from satellite images with deep learning method","volume":"28","author":"Aslan","year":"2025","journal-title":"Cluster Comput"},{"key":"B26","doi-asserted-by":"publisher","first-page":"170117","DOI":"10.1038\/sdata.2017.117","article-title":"Advancing the cancer genome atlas glioma MRI collections with expert segmentation labels and radiomic features","volume":"4","author":"Bakas","year":"2017","journal-title":"Sci Data"},{"key":"B27","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1811.02629","article-title":"Identifying the best machine learning algorithms for brain tumor segmentation, progression assessment, and overall survival prediction in the BRATS challenge","author":"Bakas","year":"2018","journal-title":"arXiv Preprint"},{"key":"B28","first-page":"424","article-title":"3D U-Net: learning dense volumetric segmentation from sparse annotation","volume-title":"International conference on medical image computing and computer-assisted intervention","author":"\u00c7i\u00e7ek","year":"2016"},{"key":"B29","doi-asserted-by":"crossref","first-page":"565","DOI":"10.1109\/3DV.2016.79","article-title":"V-net: fully convolutional neural networks for volumetric medical image segmentation","volume-title":"2016 Fourth international conference on 3D vision (3DV)","author":"Milletari","year":"2016"},{"key":"B30","first-page":"770","article-title":"Deep residual learning for image recognition","volume-title":"Proceedings of the IEEE conference on computer vision and pattern recognition","author":"He","year":"2016"},{"key":"B31","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1804.03999","article-title":"Attention u-net: learning where to look for the pancreas","author":"Oktay","year":"2018","journal-title":"arXiv Preprint arXiv:180403999"},{"key":"B32","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1802.06955","article-title":"Recurrent residual convolutional neural network based on u-net (r2u-net) for medical image segmentation","author":"Alom","year":"2018","journal-title":"arXiv Preprint arXiv:180206955"},{"key":"B33","first-page":"3","article-title":"Unet++: a nested u-net architecture for medical image segmentation","volume-title":"International workshop on deep learning in medical image analysis","author":"Zhou","year":"2018"},{"key":"B34","doi-asserted-by":"publisher","first-page":"74","DOI":"10.1016\/j.neunet.2019.08.025","article-title":"MultiResUNet: rethinking the U-Net architecture for multimodal biomedical image segmentation","volume":"121","author":"Ibtehaz","year":"2020","journal-title":"Neural Netw"},{"key":"B35","first-page":"1055","article-title":"Unet 3+: a full-scale connected unet for medical image segmentation","volume-title":"ICASSP 2020-2020 IEEE international conference on acoustics, speech and signal processing (ICASSP)","author":"Huang","year":"2020"},{"key":"B36","doi-asserted-by":"publisher","first-page":"6625688","DOI":"10.1155\/2021\/6625688","article-title":"R2AU-Net: attention recurrent residual convolutional neural network for multimodal medical image segmentation","volume":"2021","author":"Zuo","year":"2021","journal-title":"Security Commun Netw"},{"key":"B37","first-page":"109","article-title":"Transbts: multimodal brain tumor segmentation using transformer","volume-title":"International conference on medical image computing and computer-assisted Intervention","author":"Wang","year":"2021"},{"key":"B38","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2102.04306","article-title":"Transunet: transformers make strong encoders for medical image segmentation","author":"Chen","year":"2021","journal-title":"arXiv Preprint arXiv:210204306"},{"key":"B39","first-page":"205","article-title":"Swin-unet: Unet-like pure transformer for medical image segmentation","volume-title":"European conference on computer vision","author":"Cao","year":"2022"},{"key":"B40","doi-asserted-by":"publisher","first-page":"108005","DOI":"10.1016\/j.compbiomed.2024.108005","article-title":"ETUNet: exploring efficient transformer enhanced UNet for 3D brain tumor segmentation","volume":"171","author":"Zhang","year":"2024","journal-title":"Comput Biol Med"},{"key":"B41","doi-asserted-by":"publisher","first-page":"737","DOI":"10.1109\/JBHI.2020.2998146","article-title":"HDC-Net: hierarchical decoupled convolution network for brain tumor segmentation","volume":"25","author":"Luo","year":"2020","journal-title":"IEEE J Biomed Health Inform"},{"key":"B42","doi-asserted-by":"publisher","first-page":"110460","DOI":"10.1016\/j.compbiomed.2025.110460","article-title":"FFLUNet: feature fused lightweight UNet for brain tumor segmentation","volume":"194","author":"Kundu","year":"2025","journal-title":"Comput Biol Med"},{"key":"B43","doi-asserted-by":"publisher","first-page":"109353","DOI":"10.1016\/j.compbiomed.2024.109353","article-title":"LATUP-Net: a lightweight 3D attention U-Net with parallel convolutions for brain tumor segmentation","volume":"184","author":"Alwadee","year":"2025","journal-title":"Comput Biol Med"},{"key":"B44","doi-asserted-by":"publisher","first-page":"111925","DOI":"10.1016\/j.engappai.2025.111925","article-title":"lightweight segmentation model based on dilated multi-scale residual attention U-Net for brain tumor segmentation","volume":"160","author":"Zhang","year":"2025","journal-title":"Eng Appl Artif Intell"},{"key":"B45","doi-asserted-by":"publisher","first-page":"102727","DOI":"10.1016\/j.compmedimag.2026.102727","article-title":"LCBTS-Net: a lightweight cascaded 3D brain tumor segmentation network in magnetic resonance imaging","volume":"129","author":"Koc","year":"2026","journal-title":"Comput Med Imag Graph"},{"key":"B46","unstructured":"Digital Health Divide: Only 1 in 2 Countries in Europe and Central Asia Have Policies to Improve Digital Health Literacy\n          \n          2023"}],"container-title":["Frontiers in Neurology"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fneur.2026.1894796\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T14:19:40Z","timestamp":1784816380000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fneur.2026.1894796\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,23]]},"references-count":46,"alternative-id":["10.3389\/fneur.2026.1894796"],"URL":"https:\/\/doi.org\/10.3389\/fneur.2026.1894796","relation":{},"ISSN":["1664-2295"],"issn-type":[{"value":"1664-2295","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,23]]},"article-number":"1894796"}}