{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,31]],"date-time":"2026-08-31T06:15:38Z","timestamp":1788156938746,"version":"build-2803163510"},"reference-count":61,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,8,31]],"date-time":"2026-08-31T00:00:00Z","timestamp":1788134400000},"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. Artif. Intell."],"abstract":"<jats:p>\n                    Deep neural networks for medical image classification rely almost exclusively on fixed activation functions such as ReLU. We introduce NeuroPlast, a parametric activation function whose four differentiable components (a shifted sigmoid modeling NMDA-type voltage gating, a Gaussian plateau inspired by AMPA receptor scaling, an excitatory rectifier, and an inhibitory leak) are combined through six learnable parameters. Two mixing strategies are evaluated: a\n                    <jats:italic>static<\/jats:italic>\n                    variant with fixed learned weights, and a\n                    <jats:italic>metaplastic<\/jats:italic>\n                    variant whose mixing coefficients adapt per sample via a lightweight squeeze-excite gate conditioned on channel statistics. NeuroPlast is embedded within NADN Ultra, a 13.5 M-parameter residual convolutional neural network (CNN) with convolutional block attention modules (CBAM), trained entirely from scratch through a two-phase knowledge distillation (KD) pipeline. The teacher is a fine-tuned EfficientNet-B0; the student combines logit-level KD with optional feature-level alignment losses. Across four medical imaging benchmarks (Brain Tumor MRI, 7,200 images, 4 classes; Chest X-ray Pneumonia, 5,856 images, 2 classes; Skin Cancer HAM10000, 10,015 images, 7 classes; COVID-19 Radiography, 10,848 images, 2 classes), evaluated under five-fold stratified cross-validation with 95% confidence intervals, the static-KD variant reaches 99.4\u201399.5% accuracy on COVID-19 X-ray across three seeds, on par with pretrained EfficientNet-B0 (99.18%, within replication noise) and above ResNet-18 (98.69%). It closes 65% of the accuracy gap on Brain Tumor MRI (98.29% vs. 99.13%), reaches 92.5% on Chest X-ray under a uniform class-balancing rule, and 79.5% on Skin Cancer under lesion-grouped cross-validation. The metaplastic variant achieves 99.28% on COVID-19 X-ray, still above pretrained baselines, but does not consistently outperform the static version, a negative result analyzed through ablation experiments. On three tabular medical datasets spanning three orders of magnitude in sample size (569 to 253,680), NeuroPlast matches five established activations within \u00b11.7 percentage points; at 253 K samples all activations converge within 0.09 pp. Our findings indicate that knowledge distillation is the primary enabler for from-scratch architectures to approach pretrained-level performance on medical images; under an identical distillation pipeline, NeuroPlast adds a small but consistent gain over ReLU, GELU, Swish, Mish, and PReLU, leading on all four imaging benchmarks by margins below one point.\n                  <\/jats:p>","DOI":"10.3389\/frai.2026.1849571","type":"journal-article","created":{"date-parts":[[2026,8,31]],"date-time":"2026-08-31T05:39:32Z","timestamp":1788154772000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["NeuroPlast: a learnable activation function evaluated under knowledge distillation for medical image classification"],"prefix":"10.3389","volume":"9","author":[{"given":"Fabrice","family":"Vaussenat","sequence":"first","affiliation":[{"name":"Department of Electrical Engineering, \u00c9cole de technologie sup\u00e9rieure","place":["Montr\u00e9al, QC, Canada"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Abhiroop","family":"Bhattacharya","sequence":"additional","affiliation":[{"name":"Department of Electrical Engineering, \u00c9cole de technologie sup\u00e9rieure","place":["Montr\u00e9al, QC, Canada"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Julie","family":"Payette","sequence":"additional","affiliation":[{"name":"Department of Electrical Engineering, \u00c9cole de technologie sup\u00e9rieure","place":["Montr\u00e9al, QC, Canada"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Thibaut","family":"Desmettre","sequence":"additional","affiliation":[{"name":"Department of Acute Medicine, H\u00f4pital Universitaire de Gen\u00e8ve","place":["Gen\u00e8ve, Switzerland"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Alireza","family":"Saidi","sequence":"additional","affiliation":[{"name":"Institut de recherche Robert-Sauv\u00e9 en sant\u00e9 et en s\u00e9curit\u00e9 du travail (IRSST)","place":["Montr\u00e9al, QC, Canada"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ghyslain","family":"Gagnon","sequence":"additional","affiliation":[{"name":"Department of Electrical Engineering, \u00c9cole de technologie sup\u00e9rieure","place":["Montr\u00e9al, QC, Canada"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sylvain G.","family":"Cloutier","sequence":"additional","affiliation":[{"name":"Department of Electrical Engineering, \u00c9cole de technologie sup\u00e9rieure","place":["Montr\u00e9al, QC, Canada"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,8,31]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"126","DOI":"10.1016\/S0166-2236(96)80018-X","article-title":"Metaplasticity: the plasticity of synaptic plasticity","volume":"19","author":"Abraham","year":"1996","journal-title":"Trends Neurosci"},{"key":"B2","article-title":"\u201cLearning activation functions to improve deep neural networks,\u201d","volume-title":"Proceedings of the 3rd International Conference on Learning Representations","author":"Agostinelli","year":"2015"},{"key":"B3","doi-asserted-by":"publisher","first-page":"14","DOI":"10.1016\/j.neunet.2021.01.026","article-title":"A survey on modern trainable activation functions","volume":"138","author":"Apicella","year":"2021","journal-title":"Neural Netw"},{"key":"B4","doi-asserted-by":"publisher","first-page":"32","DOI":"10.1523\/JNEUROSCI.02-01-00032.1982","article-title":"Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex","volume":"2","author":"Bienenstock","year":"1982","journal-title":"J. Neurosci"},{"key":"B5","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"},{"key":"B6","doi-asserted-by":"publisher","first-page":"132665","DOI":"10.1109\/ACCESS.2020.3010287","article-title":"Can AI help in screening viral and COVID-19 pneumonia?","volume":"8","author":"Chowdhury","year":"2020","journal-title":"IEEE Access"},{"key":"B7","article-title":"\u201cFast and accurate deep network learning by exponential linear units (ELUs),\u201d","volume-title":"Proceedings of the 4th International Conference on Learning Representations","author":"Clevert","year":"2016"},{"key":"B8","doi-asserted-by":"publisher","first-page":"431","DOI":"10.1038\/s41583-023-00705-w","article-title":"The neuroconnectionist research programme","volume":"24","author":"Doerig","year":"2023","journal-title":"Nat. Rev. Neurosci"},{"key":"B9","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2010.11929","article-title":"An image is worth 16x16 words: transformers for image recognition at scale","author":"Dosovitskiy","year":"2020","journal-title":"arXiv"},{"key":"B10","doi-asserted-by":"publisher","first-page":"92","DOI":"10.1016\/j.neucom.2022.06.111","article-title":"Activation functions in deep learning: a comprehensive survey and benchmark","volume":"503","author":"Dubey","year":"2022","journal-title":"Neurocomputing"},{"key":"B11","doi-asserted-by":"publisher","first-page":"115","DOI":"10.1038\/nature21056","article-title":"Dermatologist-level classification of skin cancer with deep neural networks","volume":"542","author":"Esteva","year":"2017","journal-title":"Nature"},{"key":"B12","article-title":"\u201cDeep sparse rectifier neural networks,\u201d","volume-title":"Proceedings of the 14th International Conference on Artificial Intelligence and Statistics (AISTATS)","author":"Glorot","year":"2011"},{"key":"B13","volume-title":"Deep Learning","author":"Goodfellow","year":"2016"},{"key":"B14","doi-asserted-by":"publisher","first-page":"1789","DOI":"10.1007\/s11263-021-01453-z","article-title":"Knowledge distillation: a survey","volume":"129","author":"Gou","year":"2021","journal-title":"Int. J. Comput. Vis"},{"key":"B15","doi-asserted-by":"publisher","first-page":"245","DOI":"10.1016\/j.neuron.2017.06.011","article-title":"Neuroscience-inspired artificial intelligence","volume":"95","author":"Hassabis","year":"2017","journal-title":"Neuron"},{"key":"B16","first-page":"574","article-title":"\u201cUNETR: Transformers for 3d medical image segmentation,\u201d","volume-title":"Proceedings of the IEEE\/CVF Winter Conference on Applications of Computer Vision","author":"Hatamizadeh","year":"2022"},{"key":"B17","first-page":"1026","article-title":"\u201cDelving deep into rectifiers: surpassing human-level performance on imagenet classification,\u201d","volume-title":"Proceedings of the IEEE International Conference on Computer Vision","author":"He","year":"2015"},{"key":"B18","first-page":"770","article-title":"\u201cDeep residual learning for image recognition,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"He","year":"2016"},{"key":"B19","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1606.08415","article-title":"Gaussian error linear units (GELUs)","author":"Hendrycks","year":"2016","journal-title":"arXiv"},{"key":"B20","doi-asserted-by":"publisher","first-page":"337","DOI":"10.1038\/nrn.2016.37","article-title":"Synaptic AMPA receptor composition in development, plasticity and disease","volume":"17","author":"Henley","year":"2016","journal-title":"Nat. Rev. Neurosci"},{"key":"B21","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1503.02531","article-title":"Distilling the knowledge in a neural network","author":"Hinton","year":"2015","journal-title":"arXiv"},{"key":"B22","first-page":"7132","article-title":"\u201cSqueeze-and-excitation networks,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Hu","year":"2018"},{"key":"B23","first-page":"4700","article-title":"\u201cDensely connected convolutional networks,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Huang","year":"2017"},{"key":"B24","first-page":"646","article-title":"\u201cDeep networks with stochastic depth,\u201d","volume-title":"European Conference on Computer Vision","author":"Huang","year":"2016"},{"key":"B25","doi-asserted-by":"publisher","first-page":"109136","DOI":"10.1016\/j.jcp.2019.109136","article-title":"Adaptive activation functions accelerate convergence in deep and physics-informed neural networks","volume":"404","author":"Jagtap","year":"2020","journal-title":"J. Comput. Phys"},{"key":"B26","doi-asserted-by":"publisher","first-page":"45769","DOI":"10.1038\/s41598-025-28636-9","article-title":"A hybrid CNN-ViT framework with cross-attention fusion and data augmentation for robust brain tumor classification","volume":"15","author":"Jayaraman","year":"2025","journal-title":"Sci. Rep"},{"key":"B27","doi-asserted-by":"publisher","first-page":"100804","DOI":"10.1016\/j.patter.2023.100804","article-title":"Leakage and the reproducibility crisis in machine-learning-based science","volume":"4","author":"Kapoor","year":"2023","journal-title":"Patterns"},{"key":"B28","doi-asserted-by":"publisher","first-page":"1122","DOI":"10.1016\/j.cell.2018.02.010","article-title":"Identifying medical diagnoses and treatable diseases by image-based deep learning","volume":"172","author":"Kermany","year":"2018","journal-title":"Cell"},{"key":"B29","doi-asserted-by":"publisher","first-page":"436","DOI":"10.1038\/nature14539","article-title":"Deep learning","volume":"521","author":"LeCun","year":"2015","journal-title":"Nature"},{"key":"B30","first-page":"2980","article-title":"\u201cFocal loss for dense object detection,\u201d","volume-title":"Proceedings of the IEEE International Conference on Computer Vision","author":"Lin","year":"2017"},{"key":"B31","doi-asserted-by":"publisher","first-page":"e537","DOI":"10.1136\/bmj.m3164","article-title":"Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension","volume":"2","author":"Liu","year":"2020","journal-title":"Lancet Digit. Health"},{"key":"B32","first-page":"10012","article-title":"\u201cSwin transformer: Hierarchical vision transformer using shifted windows,\u201d","volume-title":"Proceedings of the IEEE\/CVF International Conference on Computer Vision","author":"Liu","year":"2021"},{"key":"B33","doi-asserted-by":"publisher","first-page":"1452","DOI":"10.1007\/s11760-025-05041-3","article-title":"Comparative analysis of CNN, vision transformer, and hybrid architectures for white blood cancer classification","volume":"19","author":"Manal\u0131","year":"2025","journal-title":"Signal Image Video Proc"},{"key":"B34","doi-asserted-by":"publisher","DOI":"10.5244\/C.34.191","article-title":"Mish: A self regularized non-monotonic activation function","author":"Misra","year":"2020","journal-title":"arXiv"},{"key":"B35","article-title":"\u201cPad\u00e9 activation units: End-to-end learning of flexible activation functions in deep networks,\u201d","volume-title":"Proceedings of the 8th International Conference on Learning Representations","author":"Molina","year":"2020"},{"key":"B36","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2103.10158","article-title":"\u201cTrivialAugment: tuning-free yet state-of-the-art data augmentation,\u201d","author":"M\u00fcller","year":"2021","journal-title":"Proceedings of the IEEE\/CVF International Conference on Computer Vision"},{"key":"B37","unstructured":"Nickparvar\n              M.\n            \n          \n          Brain Tumor MRI Dataset\n          \n          2023"},{"key":"B38","doi-asserted-by":"publisher","first-page":"383","DOI":"10.1038\/nrn3504","article-title":"NMDA receptor subunit diversity: impact on receptor properties, synaptic plasticity and disease","volume":"14","author":"Paoletti","year":"2013","journal-title":"Nat. Rev. Neurosci"},{"key":"B39","doi-asserted-by":"publisher","first-page":"76","DOI":"10.3389\/fninf.2017.00076","article-title":"Reproducibility vs. replicability: a brief history of a confused terminology","volume":"11","author":"Plesser","year":"2018","journal-title":"Front. Neuroinform"},{"key":"B40","article-title":"\u201cTransfusion: understanding transfer learning for medical imaging,\u201d","author":"Raghu","year":"2019","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B41","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1711.05225","article-title":"CheXNet: Radiologist-level pneumonia detection on chest X-rays with deep learning","author":"Rajpurkar","year":"2017","journal-title":"arXiv"},{"key":"B42","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1710.05941","article-title":"Searching for activation functions","author":"Ramachandran","year":"2017","journal-title":"arXiv"},{"key":"B43","doi-asserted-by":"publisher","first-page":"199","DOI":"10.1038\/s42256-021-00307-0","article-title":"Common pitfalls and recommendations for using machine learning to detect and prognosticate for covid-19 using chest radiographs and ct scans","volume":"3","author":"Roberts","year":"2021","journal-title":"Nat. Mach. Intellig"},{"key":"B44","article-title":"\u201cFitNets: Hints for thin deep nets,\u201d","volume-title":"Proceedings of the 3rd International Conference on Learning Representations (ICLR)","author":"Romero","year":"2015"},{"key":"B45","first-page":"6105","article-title":"\u201cEfficientNet: rethinking model scaling for convolutional neural networks,\u201d","volume-title":"Proceedings of the 36th International Conference on Machine Learning","author":"Tan","year":"2019"},{"key":"B46","doi-asserted-by":"publisher","first-page":"105605","DOI":"10.1016\/j.bspc.2023.105605","article-title":"Htc-Net: a hybrid CNN-transformer framework for medical image segmentation","volume":"88","author":"Tang","year":"2024","journal-title":"Biomed. Signal Process. Control"},{"key":"B47","article-title":"\u201cMean teachers are better role models: Weight-averaged consistency targets improve semi-supervised learning results,\u201d","author":"Tarvainen","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B48","article-title":"\u201cContrastive representation distillation,\u201d","volume-title":"Proceedings of the 8th International Conference on Learning Representations (ICLR)","author":"Tian","year":"2020"},{"key":"B49","doi-asserted-by":"publisher","first-page":"24002","DOI":"10.1002\/jcu.24002","article-title":"Melanoma skin classification using the hybrid approach residual network-vision transformer for cancer diagnosis","volume":"53","author":"Toure","year":"2025","journal-title":"J. Clini. Ultrasound"},{"key":"B50","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1038\/sdata.2018.161","article-title":"The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions","volume":"5","author":"Tschandl","year":"2018","journal-title":"Scientific Data"},{"key":"B51","doi-asserted-by":"publisher","first-page":"422","DOI":"10.1016\/j.cell.2008.10.008","article-title":"The self-tuning neuron: Synaptic scaling of excitatory synapses","volume":"135","author":"Turrigiano","year":"2008","journal-title":"Cell"},{"key":"B52","doi-asserted-by":"publisher","first-page":"91","DOI":"10.1186\/1471-2105-7-91","article-title":"Bias in error estimation when using cross-validation for model selection","volume":"7","author":"Varma","year":"2006","journal-title":"BMC Bioinformat"},{"key":"B53","first-page":"5998","article-title":"\u201cAttention is all you need,\u201d","author":"Vaswani","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B54","doi-asserted-by":"publisher","first-page":"1348","DOI":"10.3390\/s26041348","article-title":"Early drowsiness detection via second-order derivative analysis of heart rate variability: a non-contact ECG approach with machine learning","volume":"26","author":"Vaussenat","year":"2026","journal-title":"Sensors"},{"key":"B55","first-page":"109881","article-title":"UCTNet: Uncertainty-guided CNN-transformer hybrid networks for medical image segmentation","volume":"145","author":"Wang","year":"2024","journal-title":"Pattern Recognit"},{"key":"B56","first-page":"3","article-title":"\u201cCBAM: Convolutional block attention module,\u201d","volume-title":"Proceedings of the European Conference on Computer Vision (ECCV)","author":"Woo","year":"2018"},{"key":"B57","first-page":"6023","article-title":"\u201cCutMix: regularization strategy to train strong classifiers with localizable features,\u201d","volume-title":"Proceedings of the IEEE\/CVF International Conference on Computer Vision","author":"Yun","year":"2019"},{"key":"B58","doi-asserted-by":"publisher","first-page":"1597","DOI":"10.1038\/s41467-023-37180-x","article-title":"Catalyzing next-generation Artificial Intelligence through NeuroAI","volume":"14","author":"Zador","year":"2023","journal-title":"Nat. Commun"},{"key":"B59","doi-asserted-by":"publisher","first-page":"e1003330","DOI":"10.1371\/journal.pcbi.1003330","article-title":"Synaptic plasticity in neural networks needs homeostasis with a fast rate detector","volume":"9","author":"Zenke","year":"2013","journal-title":"PLoS Comput. Biol"},{"key":"B60","article-title":"\u201cmixup: beyond empirical risk minimization,\u201d","volume-title":"Proceedings of the 6th International Conference on Learning Representations (ICLR)","author":"Zhang","year":"2018"},{"key":"B61","doi-asserted-by":"publisher","first-page":"1614046","DOI":"10.3389\/fmed.2025.1614046","article-title":"A dual attention and multi-scale fusion network for diabetic retinopathy image analysis","volume":"12","author":"Zhang","year":"2025","journal-title":"Front. Med"}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1849571\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,31]],"date-time":"2026-08-31T05:39:39Z","timestamp":1788154779000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1849571\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,31]]},"references-count":61,"alternative-id":["10.3389\/frai.2026.1849571"],"URL":"https:\/\/doi.org\/10.3389\/frai.2026.1849571","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,8,31]]},"article-number":"1849571"}}