{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"bioRxiv"}],"indexed":{"date-parts":[[2026,3,26]],"date-time":"2026-03-26T09:04:46Z","timestamp":1774515886663,"version":"3.50.1"},"posted":{"date-parts":[[2026,3,23]]},"group-title":"Bioinformatics","reference-count":54,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2026,3,23]],"date-time":"2026-03-23T00:00:00Z","timestamp":1774224000000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by-nc-nd\/4.0\/"}],"funder":[{"name":"Funda\u00e7\u00e3o para a Ci\u00eancia e Tecnologia","award":["UID\/00027\/2025"],"award-info":[{"award-number":["UID\/00027\/2025"]}]},{"name":"Funda\u00e7\u00e3o para a Ci\u00eancia e Tecnologia","award":["UIDB\/00326\/2025"],"award-info":[{"award-number":["UIDB\/00326\/2025"]}]},{"name":"Funda\u00e7\u00e3o para a Ci\u00eancia e Tecnologia","award":["UIDP\/00326\/2025"],"award-info":[{"award-number":["UIDP\/00326\/2025"]}]},{"name":"Funda\u00e7\u00e3o para a Ci\u00eancia e Tecnologia","award":["LA\/P\/0104\/2020"],"award-info":[{"award-number":["LA\/P\/0104\/2020"]}]},{"name":"Funda\u00e7\u00e3o para a Ci\u00eancia e Tecnologia","award":["2025.02850.BD"],"award-info":[{"award-number":["2025.02850.BD"]}]},{"name":"Funda\u00e7\u00e3o para a Ci\u00eancia e Tecnologia","award":["CEECINST\/00117\/2021\/CP2784\/CT0002"],"award-info":[{"award-number":["CEECINST\/00117\/2021\/CP2784\/CT0002"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,3,23]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:p>The application of artificial intelligence (AI) to functional magnetic resonance imaging (fMRI) has gained increasing attention due to its ability to model complex, high-dimensional brain data and capture nonlinear patterns of neural activity. However, deep learning architectures, such as Graph Neural Networks (GNNs), typically require large sample sizes to achieve stable convergence, limiting their applicability in neuroimaging contexts where data are often scarce. This challenge highlights the need for compact, data-efficient models that maintain predictive performance and interpretability.<\/jats:p>\n                <jats:p>Shallow neural networks (SNNs) have demonstrated robustness in low-sample settings but commonly rely on region-level features that treat brain areas independently, overlooking the brain\u2019s intrinsically network-based organization. To address this limitation, we propose a structurally constrained message-passing framework that integrates diffusion tensor imaging (DTI)-derived structural connectivity with region-level fMRI signals within a shallow architecture. This approach enables network-level modeling while preserving the stability and data efficiency of SNNs.<\/jats:p>\n                <jats:p>The method is evaluated on 30 subjects performing a Theory of Mind (ToM) task from the Human Connectome Project Young Adult dataset. A baseline SNN achieved global accuracies of 88.2% (fully connected), 80.0% (pruned), and 84.7% (retrained), while the proposed model achieved 87.1%, 77.6%, and 84.7%, respectively. Although structural constraints led to a more pronounced performance decrease after pruning, retraining restored accuracy to baseline levels, demonstrating that biological constraints can be incorporated without compromising predictive validity.<\/jats:p>\n                <jats:p>Model interpretability was assessed using SHAP (Shapley Additive Explanations). While the baseline model primarily identified isolated regions as key contributors, the proposed framework revealed distributed, structurally coherent networks as the main drivers of classification. These networks showed correspondence with established ToM regions, including the temporo-parietal junction, superior temporal sulcus, and inferior frontal gyrus. Importantly, the findings suggest that groups of moderately informative regions can collectively form highly relevant subnetworks.<\/jats:p>\n                <jats:p>Overall, the proposed framework achieves competitive performance in a limited dataset while incorporating graph-inspired message passing into a shallow architecture. Its explainability provides insight into how structurally constrained networks support stimulus-driven responses in ToM and demonstrates potential for investigating network dysfunction in disorders such as Alzheimer\u2019s disease, ADHD, autism spectrum disorder, bipolar disorder, mild cognitive impairment, and schizophrenia.<\/jats:p>","DOI":"10.64898\/2026.03.20.713136","type":"posted-content","created":{"date-parts":[[2026,3,23]],"date-time":"2026-03-23T23:15:31Z","timestamp":1774307731000},"source":"Crossref","is-referenced-by-count":0,"title":["Structurally Restricted Message-Passing within Shallow Architectures for Explainable Network-Level Brain Decoding on Small Cohorts"],"prefix":"10.64898","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-4101-2748","authenticated-orcid":false,"given":"Jos\u00e9 Diogo","family":"Marques dos Santos","sequence":"first","affiliation":[{"name":"Faculty of Engineering, University of Porto, Porto, Portugal"},{"name":"LIACC \u2013 Artificial Intelligence and Computer Science Laboratory, University of Porto, Porto, Portugal"}]},{"ORCID":"https:\/\/orcid.org\/0009-0008-3523-2246","authenticated-orcid":false,"given":"Maria Beatriz","family":"Ramos","sequence":"additional","affiliation":[{"name":"Faculty of Medicine, University of Porto, Porto, Portugal"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4709-1718","authenticated-orcid":false,"given":"Lu\u00eds Paulo","family":"Reis","sequence":"additional","affiliation":[{"name":"Faculty of Engineering, University of Porto, Porto, Portugal"},{"name":"LIACC \u2013 Artificial Intelligence and Computer Science Laboratory, University of Porto, Porto, Portugal"},{"name":"LASI \u2013 Intelligent Systems Associate Laboratory, Guimar\u00e3es, Portugal"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5567-944X","authenticated-orcid":false,"given":"Jos\u00e9 Paulo","family":"Marques dos Santos","sequence":"additional","affiliation":[{"name":"LIACC \u2013 Artificial Intelligence and Computer Science Laboratory, University of Porto, Porto, Portugal"},{"name":"Faculty of Medicine, University of Porto, Porto, Portugal"},{"name":"LASI \u2013 Intelligent Systems Associate Laboratory, Guimar\u00e3es, Portugal"},{"name":"University of Maia, Maia, Portugal"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3259-8815","authenticated-orcid":false,"given":"Bruno","family":"Direito","sequence":"additional","affiliation":[{"name":"LASI \u2013 Intelligent Systems Associate Laboratory, Guimar\u00e3es, Portugal"},{"name":"CISUC \u2013 Centre for Informatics and Systems, University of Coimbra, Coimbra, Portugal"}]}],"member":"54368","reference":[{"key":"2026032600451208000_2026.03.20.713136v1.1","doi-asserted-by":"publisher","DOI":"10.1016\/S0959-4388(00)00202-6"},{"key":"2026032600451208000_2026.03.20.713136v1.2","doi-asserted-by":"publisher","DOI":"10.1093\/ons\/opy253"},{"key":"2026032600451208000_2026.03.20.713136v1.3","unstructured":"Baliga, V. B. , Armstrong, M. S. , Press, E. R. , Bonnet-Lebrun, A.-S. , & Sciaini, M. (2023). pathviewr: Wrangle, Analyze, and Visualize Animal Movement Data (v. 1.1.7). In https:\/\/cran.r-project.org\/package=pathviewr"},{"key":"2026032600451208000_2026.03.20.713136v1.4","doi-asserted-by":"publisher","DOI":"10.1038\/nn.4502"},{"key":"2026032600451208000_2026.03.20.713136v1.5","doi-asserted-by":"publisher","DOI":"10.1093\/ons\/opy272"},{"key":"2026032600451208000_2026.03.20.713136v1.6","doi-asserted-by":"publisher","DOI":"10.1093\/ons\/opy268"},{"key":"2026032600451208000_2026.03.20.713136v1.7","doi-asserted-by":"publisher","DOI":"10.1093\/ons\/opy270"},{"key":"2026032600451208000_2026.03.20.713136v1.8","doi-asserted-by":"publisher","DOI":"10.1093\/ons\/opy271"},{"key":"2026032600451208000_2026.03.20.713136v1.9","doi-asserted-by":"publisher","DOI":"10.1093\/ons\/opy269"},{"key":"2026032600451208000_2026.03.20.713136v1.10","doi-asserted-by":"publisher","DOI":"10.1016\/j.tics.2008.02.010"},{"key":"2026032600451208000_2026.03.20.713136v1.11","doi-asserted-by":"publisher","DOI":"10.1007\/s44163-024-00161-0"},{"key":"2026032600451208000_2026.03.20.713136v1.12","doi-asserted-by":"publisher","DOI":"10.1093\/ons\/opy264"},{"key":"2026032600451208000_2026.03.20.713136v1.13","doi-asserted-by":"publisher","DOI":"10.1093\/ons\/opy266"},{"key":"2026032600451208000_2026.03.20.713136v1.14","doi-asserted-by":"publisher","DOI":"10.1093\/ons\/opy267"},{"key":"2026032600451208000_2026.03.20.713136v1.15","doi-asserted-by":"publisher","DOI":"10.1080\/17470910600683549"},{"key":"2026032600451208000_2026.03.20.713136v1.16","unstructured":"Doran, D. , Schulz, S. , & Besold, T. R. (2018). What Does Explainable AI Really Mean? A New Conceptualization of Perspectives. In T. R. Besold & O. Kutz (Eds.), Proceedings of the 1st International Workshop on Comprehensibility and Explanation in AI and ML, CEX 2017 (Vol. 2071). CEUR Workshop Proceedings."},{"key":"2026032600451208000_2026.03.20.713136v1.17","doi-asserted-by":"publisher","DOI":"10.3389\/fnins.2018.00525"},{"key":"2026032600451208000_2026.03.20.713136v1.18","doi-asserted-by":"publisher","DOI":"10.7554\/eLife.86327"},{"key":"2026032600451208000_2026.03.20.713136v1.19","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuroimage.2021.118543"},{"key":"2026032600451208000_2026.03.20.713136v1.20","doi-asserted-by":"publisher","DOI":"10.3389\/fnins.2022.906290"},{"key":"2026032600451208000_2026.03.20.713136v1.21","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuron.2006.05.001"},{"key":"2026032600451208000_2026.03.20.713136v1.22","doi-asserted-by":"publisher","DOI":"10.1038\/nature18933"},{"key":"2026032600451208000_2026.03.20.713136v1.23","unstructured":"Haykin, S . (2009). Neural Networks and Learning Machines (3rd ed.). Prentice Hall."},{"key":"2026032600451208000_2026.03.20.713136v1.24","doi-asserted-by":"publisher","DOI":"10.2307\/1416950"},{"key":"2026032600451208000_2026.03.20.713136v1.25","doi-asserted-by":"publisher","DOI":"10.1016\/j.cortex.2016.04.017"},{"key":"2026032600451208000_2026.03.20.713136v1.26","unstructured":"Limas, M. C. , Mer\u00e9, J. B. O. , Marcos, A. G. , Ascacibar, F. J. M. d. P. , Espinoza, A. V. P. , El\u00edas, F. A. , & Ramos, J. M. P. (2014). AMORE: A MORE flexible neural network package (0.2-15). In (Version 0.2-15) http:\/\/cran.nexr.com\/web\/packages\/AMORE\/index.html"},{"key":"2026032600451208000_2026.03.20.713136v1.27","unstructured":"Lundberg, S. M. , & Lee, S.-I. (2017). A unified approach to interpreting model predictions Advances in Neural Information Processing Systems 30 (NIPS 2017), Long Beach (CA), USA. https:\/\/proceedings.neurips.cc\/paper\/2017\/file\/8a20a8621978632d76c43dfd28b67767-Paper.pdf"},{"key":"2026032600451208000_2026.03.20.713136v1.28","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-07802-6_22"},{"key":"2026032600451208000_2026.03.20.713136v1.29","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-25891-6_31"},{"key":"2026032600451208000_2026.03.20.713136v1.30","doi-asserted-by":"publisher","DOI":"10.3390\/make7010017"},{"key":"2026032600451208000_2026.03.20.713136v1.31","doi-asserted-by":"publisher","DOI":"10.3389\/fnhum.2024.1305164"},{"issue":"1","key":"2026032600451208000_2026.03.20.713136v1.32","first-page":"17","article-title":"Graph Neural Networks in Brain Connectivity Studies: Methods, Challenges, and Future Directions","volume":"15","year":"2025","journal-title":"Brain Sciences"},{"key":"2026032600451208000_2026.03.20.713136v1.33","doi-asserted-by":"publisher","DOI":"10.1002\/wcs.29"},{"key":"2026032600451208000_2026.03.20.713136v1.34","doi-asserted-by":"publisher","DOI":"10.1016\/j.neulet.2023.137267"},{"key":"2026032600451208000_2026.03.20.713136v1.35","doi-asserted-by":"publisher","DOI":"10.1016\/j.neulet.2009.07.064"},{"key":"2026032600451208000_2026.03.20.713136v1.36","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuroimage.2008.11.007"},{"key":"2026032600451208000_2026.03.20.713136v1.37","doi-asserted-by":"publisher","DOI":"10.1017\/S0140525X00076512"},{"key":"2026032600451208000_2026.03.20.713136v1.38","doi-asserted-by":"publisher","DOI":"10.3390\/brainsci13040692"},{"key":"2026032600451208000_2026.03.20.713136v1.39","doi-asserted-by":"publisher","DOI":"10.1016\/j.biopsych.2022.05.031"},{"key":"2026032600451208000_2026.03.20.713136v1.40","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2020.2976199"},{"key":"2026032600451208000_2026.03.20.713136v1.41","doi-asserted-by":"publisher","DOI":"10.1093\/ons\/opy265"},{"key":"2026032600451208000_2026.03.20.713136v1.42","doi-asserted-by":"crossref","unstructured":"Saxe, R . (2006). Four Brain Regions for One Theory of Mind? In J. T. Cacioppo , P. S. Visser , & C. L. Pickett (Eds.), Social Neuroscience: People Thinking About Thinking People (pp. 83\u2013101). MIT Press.","DOI":"10.7551\/mitpress\/6304.003.0007"},{"key":"2026032600451208000_2026.03.20.713136v1.43","doi-asserted-by":"publisher","DOI":"10.1016\/j.neubiorev.2014.01.009"},{"key":"2026032600451208000_2026.03.20.713136v1.44","doi-asserted-by":"publisher","DOI":"10.1093\/brain\/awn279"},{"key":"2026032600451208000_2026.03.20.713136v1.45","unstructured":"Shrikumar, A. , Greenside, P. , & Kundaje, A . (2017, 06\u201311 Aug, 2017). Learning important features through propagating activation differences. 34th International Conference on Machine Learning, PMLR."},{"key":"2026032600451208000_2026.03.20.713136v1.46","doi-asserted-by":"publisher","DOI":"10.1093\/scan\/nsae033"},{"key":"2026032600451208000_2026.03.20.713136v1.47","doi-asserted-by":"publisher","DOI":"10.1016\/B978-0-12-416008-8.00027-9"},{"key":"2026032600451208000_2026.03.20.713136v1.48","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuroimage.2004.07.054"},{"key":"2026032600451208000_2026.03.20.713136v1.49","unstructured":"Van Essen, D. C. , & Glasser, M. F. (2016). The Human Connectome Project: Progress and Prospects. Cerebrum: the Dana Forum on Brain Science, 2016, cer\u201310\u201316."},{"key":"2026032600451208000_2026.03.20.713136v1.50","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuroimage.2013.05.041"},{"key":"2026032600451208000_2026.03.20.713136v1.51","doi-asserted-by":"publisher","DOI":"10.3390\/make6030098"},{"key":"2026032600451208000_2026.03.20.713136v1.52","doi-asserted-by":"publisher","DOI":"10.3390\/make3030032"},{"key":"2026032600451208000_2026.03.20.713136v1.53","doi-asserted-by":"publisher","DOI":"10.3389\/fninf.2018.00023"},{"key":"2026032600451208000_2026.03.20.713136v1.54","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuropsychologia.2010.05.012"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.64898\/2026.03.20.713136","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,26]],"date-time":"2026-03-26T07:45:22Z","timestamp":1774511122000},"score":1,"resource":{"primary":{"URL":"http:\/\/biorxiv.org\/lookup\/doi\/10.64898\/2026.03.20.713136"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,3,23]]},"references-count":54,"URL":"https:\/\/doi.org\/10.64898\/2026.03.20.713136","relation":{},"subject":[],"published":{"date-parts":[[2026,3,23]]},"subtype":"preprint"}}