{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"bioRxiv"}],"indexed":{"date-parts":[[2026,10,11]],"date-time":"2026-10-11T16:03:34Z","timestamp":1791734614124,"version":"4.3.4"},"posted":{"date-parts":[[2025,11,18]]},"group-title":"Neuroscience","reference-count":50,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2025,11,18]],"date-time":"2025-11-18T00:00:00Z","timestamp":1763424000000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"award":["R01 MH118500"],"award-info":[{"award-number":["R01 MH118500"]}],"id":[{"id":"https:\/\/ror.org\/04xeg9z08","id-type":"ROR","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000893","name":"Simons Foundation","doi-asserted-by":"crossref","award":["Simons Foundation Autism Research Initiative (SFARI Grant 946867)"],"award-info":[{"award-number":["Simons Foundation Autism Research Initiative (SFARI Grant 946867)"]}],"id":[{"id":"10.13039\/100000893","id-type":"DOI","asserted-by":"crossref"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2025,12,3]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:p>Parkinson\u2019s Disease (PD) is a progressive neurodegenerative disorder affecting approximately 1% of the population over 65. Clinical diagnosis typically depends on tracking gradually developing motor symptoms as the disease progresses, underscoring the need for early detection methods to aid intervention while symptoms are still minor. Inexpensive and widely available imaging modalities such as T1-weighted MRI (T1w MRI) have potential for early PD diagnosis but lack established systematic biomarkers of PD pathology. In this study, a 3D convolutional neural network (3D CNN) was trained on 100 predominately early-state PD and 100 control T1w MRIs from Parkinson\u2019s Progression Markers Initiative (PPMI), achieving a classification accuracy of 84.5%. Misclassified subjects were majority unmedicated and particularly early PD (&lt; 3 years since first symptoms). To interrogate the biological basis behind the model\u2019s decisions, novel explainability methods were applied to generate regional saliency maps from both PD and control classifications. Regional saliency across subjects correlated best with cognitive and motor scores in nigrostriatal and other subcortical regions, as well as in temporal and insular cortices, indicating changes in these areas were best connected with symptom progression. The model was also sensitive to changes in the left frontal cortex across many subjects, which exhibited the greatest raw saliency magnitude. Pairwise saliency correlation was most pronounced between areas within the same functional network, suggesting the CNN was sensitive to network level changes in structural MRI. These findings demonstrate the potential of explainable 3D CNNs to identify network and regional biomarkers of early PD from T1w MRI.<\/jats:p>","DOI":"10.1101\/2025.11.17.688918","type":"posted-content","created":{"date-parts":[[2025,11,18]],"date-time":"2025-11-18T17:40:26Z","timestamp":1763487626000},"source":"Crossref","is-referenced-by-count":2,"title":["Explainable 3D CNNs link regional and network level disruption in early Parkinson\u2019s MRIs to symptom progression"],"prefix":"10.64898","author":[{"ORCID":"https:\/\/orcid.org\/0009-0003-9198-8471","authenticated-orcid":false,"given":"Ezekiel","family":"Moroze","sequence":"first","affiliation":[{"name":"Computational Neuroscience and Vision Laboratory, Department of Psychological and Brain Sciences, Boston University, Boston, MA, 02215 United States"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8834-5104","authenticated-orcid":false,"given":"Basilis","family":"Zikopoulos","sequence":"additional","affiliation":[{"name":"Graduate Program for Neuroscience, Boston University, Boston, MA, United States"},{"name":"Center for Systems Neuroscience, Boston University, Boston, MA, United States"},{"name":"Human Systems Neuroscience Laboratory, Department of Health Sciences, Program in Human Physiology, Boston University, Boston, MA, 02215 United States"},{"name":"Department of Anatomy & Neurobiology, Boston University School of Medicine, Boston, MA, United States"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2792-6770","authenticated-orcid":false,"given":"Arash","family":"Yazdanbakhsh","sequence":"additional","affiliation":[{"name":"Computational Neuroscience and Vision Laboratory, Department of Psychological and Brain Sciences, Boston University, Boston, MA, 02215 United States"},{"name":"Graduate Program for Neuroscience, Boston University, Boston, MA, United States"},{"name":"Center for Systems Neuroscience, Boston University, Boston, MA, United States"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026010104584107000_2025.11.17.688918v2.1","doi-asserted-by":"publisher","DOI":"10.1016\/S1474-4422(06)70471-9"},{"key":"2026010104584107000_2025.11.17.688918v2.2","doi-asserted-by":"crossref","first-page":"10","DOI":"10.1136\/jnnp.62.1.10","article-title":"Prevalence of parkinsonism and Parkinson\u2019s disease in Europe: the EUROPARKINSON Collaborative Study. European Community Concerted Action on the Epidemiology of Parkinson\u2019s disease. Journal of Neurology","volume":"62","year":"1997","journal-title":"Neurosurgery & Psychiatry"},{"key":"2026010104584107000_2025.11.17.688918v2.3","doi-asserted-by":"crossref","first-page":"3117","DOI":"10.1093\/brain\/awad064","article-title":"Impaired dopamine release in Parkinson\u2019s disease","volume":"146","year":"2023","journal-title":"Brain"},{"key":"2026010104584107000_2025.11.17.688918v2.4","doi-asserted-by":"publisher","DOI":"10.1148\/radiol.2021203341"},{"key":"2026010104584107000_2025.11.17.688918v2.5","doi-asserted-by":"publisher","DOI":"10.1001\/jama.2019.22360"},{"key":"2026010104584107000_2025.11.17.688918v2.6","first-page":"17","article-title":"Nucleus accumbens atrophy in Parkinson\u2019s disease (Mavridis\u2019 atrophy): 10 years later","volume":"11","year":"2022","journal-title":"Am J Neurodegener Dis"},{"key":"2026010104584107000_2025.11.17.688918v2.7","doi-asserted-by":"publisher","DOI":"10.1093\/brain\/awu036"},{"key":"2026010104584107000_2025.11.17.688918v2.8","doi-asserted-by":"publisher","DOI":"10.1016\/j.jns.2016.11.046"},{"key":"2026010104584107000_2025.11.17.688918v2.9","unstructured":"Adachi, M. , Hosoya, T. , Haku, T. , Yamaguchi, K. & Kawanami, T . Evaluation of the Substantia Nigra in Patients with Parkinsonian Syndrome Accomplished Using Multishot Diffusion-Weighted MR Imaging. AJNR Am J Neuroradiol vol. 20 (1999)."},{"key":"2026010104584107000_2025.11.17.688918v2.10","doi-asserted-by":"crossref","first-page":"915","DOI":"10.1007\/s00702-017-1717-8","article-title":"Magnetic resonance imaging for the diagnosis of Parkinson\u2019s disease","volume":"124","year":"2017","journal-title":"J Neural Transm"},{"key":"2026010104584107000_2025.11.17.688918v2.11","unstructured":"Seifert, K. D. & Wiener, J. I. The Impact of DaTscan on the Diagnosis and Management of Movement Disorders: A Retrospective Study. Am J Neurodegener Dis vol. 2 www.AJND.us\/ (2013)."},{"key":"2026010104584107000_2025.11.17.688918v2.12","doi-asserted-by":"publisher","DOI":"10.7861\/clinmedicine.16-4-371"},{"key":"2026010104584107000_2025.11.17.688918v2.13","doi-asserted-by":"crossref","first-page":"tgac044","DOI":"10.1093\/texcom\/tgac044","article-title":"Cerebral cortical thickness and cognitive decline in Parkinson\u2019s disease","volume":"4","year":"2023","journal-title":"Cereb Cortex Commun"},{"key":"2026010104584107000_2025.11.17.688918v2.14","doi-asserted-by":"publisher","DOI":"10.1001\/archneurol.2011.725"},{"key":"2026010104584107000_2025.11.17.688918v2.15","doi-asserted-by":"publisher","DOI":"10.1002\/brb3.573"},{"key":"2026010104584107000_2025.11.17.688918v2.16","unstructured":"Patel, S. B. , Goh, V. , Fitzgerald, J. J. & Antoniades, C. A. Multi-Cohort Development and Validation of 2D and 3D Deep Learning Models for MRI-Based Parkinson\u2019s Disease Classification: Comparative Analysis of ConvKANs, CNNs, and GCNs. https:\/\/arxiv.org\/abs\/2407.17380 (2024)."},{"key":"2026010104584107000_2025.11.17.688918v2.17","doi-asserted-by":"publisher","DOI":"10.1109\/IJCNN55064.2022.9892850"},{"key":"2026010104584107000_2025.11.17.688918v2.18","doi-asserted-by":"publisher","DOI":"10.1093\/jamia\/ocy017"},{"key":"2026010104584107000_2025.11.17.688918v2.19","doi-asserted-by":"crossref","first-page":"1460","DOI":"10.1002\/acn3.644","article-title":"The Parkinson\u2019s progression markers initiative (PPMI) \u2013 establishing a PD biomarker cohort","volume":"5","year":"2018","journal-title":"Ann Clin Transl Neurol"},{"key":"2026010104584107000_2025.11.17.688918v2.20","doi-asserted-by":"publisher","DOI":"10.1002\/mds.27361"},{"key":"2026010104584107000_2025.11.17.688918v2.21","doi-asserted-by":"publisher","DOI":"10.1016\/j.pneurobio.2011.09.005"},{"key":"2026010104584107000_2025.11.17.688918v2.22","doi-asserted-by":"crossref","unstructured":"Chakraborty, S. , Aich, S. & Kim, H. C . Detection of Parkinson\u2019s disease from 3T T1 weighted MRI scans using 3D convolutional neural network. Diagnostics 10, (2020).","DOI":"10.3390\/diagnostics10060402"},{"key":"2026010104584107000_2025.11.17.688918v2.23","doi-asserted-by":"crossref","first-page":"bpae080","DOI":"10.1093\/biomethods\/bpae080","article-title":"Deep learning and transfer learning for brain tumor detection and classification","volume":"9","year":"2024","journal-title":"Biol Methods Protoc"},{"key":"2026010104584107000_2025.11.17.688918v2.24","doi-asserted-by":"publisher","DOI":"10.2196\/53207"},{"key":"2026010104584107000_2025.11.17.688918v2.25","doi-asserted-by":"publisher","DOI":"10.1186\/s12880-022-00793-7"},{"key":"2026010104584107000_2025.11.17.688918v2.26","doi-asserted-by":"crossref","first-page":"34195","DOI":"10.1007\/s11042-020-09054-7","article-title":"Revisiting spatial dropout for regularizing convolutional neural networks","volume":"79","year":"2020","journal-title":"Multimedia Tools Appl"},{"key":"2026010104584107000_2025.11.17.688918v2.27","unstructured":"Glorot, X. & Bengio, Y. Understanding the Difficulty of Training Deep Feedforward Neural Networks. http:\/\/www.iro.umontreal. (2010)."},{"key":"2026010104584107000_2025.11.17.688918v2.28","doi-asserted-by":"crossref","unstructured":"He, K. , Zhang, X. , Ren, S. & Sun, J. Deep Residual Learning for Image Recognition. http:\/\/image-net.org\/challenges\/LSVRC\/2015\/ (2016).","DOI":"10.1109\/CVPR.2016.90"},{"key":"2026010104584107000_2025.11.17.688918v2.29","doi-asserted-by":"crossref","first-page":"16791","DOI":"10.1038\/s41598-023-43715-5","article-title":"Effects of MRI scanner manufacturers in classification tasks with deep learning models","volume":"13","year":"2023","journal-title":"Sci Rep"},{"key":"2026010104584107000_2025.11.17.688918v2.30","doi-asserted-by":"publisher","DOI":"10.1101\/2024.09.23.24313179"},{"key":"2026010104584107000_2025.11.17.688918v2.31","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2017.74"},{"key":"2026010104584107000_2025.11.17.688918v2.32","unstructured":"Mathworks. Grad-CAM Reveals the Why Behind Deep Learning Decisions. MATLAB & Simulink https:\/\/www.mathworks.com\/help\/deeplearning\/ug\/gradcam-explains-why.html."},{"key":"2026010104584107000_2025.11.17.688918v2.33","doi-asserted-by":"publisher","DOI":"10.1126\/science.abb4588"},{"key":"2026010104584107000_2025.11.17.688918v2.34","doi-asserted-by":"publisher","DOI":"10.5281\/zenodo.15591482"},{"key":"2026010104584107000_2025.11.17.688918v2.35","doi-asserted-by":"publisher","DOI":"10.1152\/jn.00338.2011"},{"key":"2026010104584107000_2025.11.17.688918v2.36","doi-asserted-by":"crossref","unstructured":"Kahali, S. , Raichle, M. E. & Yablonskiy, D. A . The Role of the Human Brain Neuron\u2013Glia\u2013Synapse Composition in Forming Resting-State Functional Connectivity Networks. Brain Sci 11, (2021).","DOI":"10.1101\/2021.07.01.449170"},{"key":"2026010104584107000_2025.11.17.688918v2.37","doi-asserted-by":"publisher","DOI":"10.1002\/hbm.22499"},{"key":"2026010104584107000_2025.11.17.688918v2.38","doi-asserted-by":"publisher","DOI":"10.3389\/fneur.2013.00090"},{"key":"2026010104584107000_2025.11.17.688918v2.39","doi-asserted-by":"publisher","DOI":"10.1177\/1545968320940985"},{"key":"2026010104584107000_2025.11.17.688918v2.40","doi-asserted-by":"publisher","DOI":"10.1111\/j.1532-5415.2005.53221.x"},{"key":"2026010104584107000_2025.11.17.688918v2.41","doi-asserted-by":"crossref","first-page":"536","DOI":"10.1002\/mdc3.12476","article-title":"Differences in MDS-UPDRS Scores Based on Hoehn and Yahr Stage and Disease Duration","volume":"4","year":"2017","journal-title":"Mov Disord Clin Pract"},{"key":"2026010104584107000_2025.11.17.688918v2.42","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuroimage.2018.06.027"},{"key":"2026010104584107000_2025.11.17.688918v2.43","doi-asserted-by":"publisher","DOI":"10.1093\/brain\/awm142"},{"key":"2026010104584107000_2025.11.17.688918v2.44","doi-asserted-by":"crossref","first-page":"175884","DOI":"10.1016\/j.ejphar.2023.175884","article-title":"Dopamine agonists versus levodopa monotherapy in early Parkinson\u2019s disease for the potential risks of motor complications: A network meta-analysis","volume":"954","year":"2023","journal-title":"Eur J Pharmacol"},{"key":"2026010104584107000_2025.11.17.688918v2.45","doi-asserted-by":"crossref","first-page":"102745","DOI":"10.1016\/j.nicl.2021.102745","article-title":"Subcortical grey matter changes associated with motor symptoms evaluated by the Unified Parkinson\u2019s disease Rating Scale (part III): A longitudinal study in Parkinson\u2019s disease","volume":"31","year":"2021","journal-title":"Neuroimage Clin"},{"key":"2026010104584107000_2025.11.17.688918v2.46","doi-asserted-by":"publisher","DOI":"10.1016\/S0010-9452(80)80006-2"},{"key":"2026010104584107000_2025.11.17.688918v2.47","doi-asserted-by":"publisher","DOI":"10.1212\/WNL.0000000000209699"},{"key":"2026010104584107000_2025.11.17.688918v2.48","doi-asserted-by":"crossref","first-page":"104401","DOI":"10.1016\/j.imavis.2022.104401","article-title":"A review on 2D instance segmentation based on deep neural networks","volume":"120","year":"2022","journal-title":"Image Vis Comput"},{"key":"2026010104584107000_2025.11.17.688918v2.49","doi-asserted-by":"crossref","unstructured":"Ragguett, R. M. , Eagleson, R. & de Ribaupierre, S . Evaluating normalized registration and preprocessing methodologies for the analysis of brain MRI in pediatric patients with shunt-treated hydrocephalus. Front Neurosci 18, (2024).","DOI":"10.3389\/fnins.2024.1405363"},{"key":"2026010104584107000_2025.11.17.688918v2.50","doi-asserted-by":"publisher","DOI":"10.1155\/2015\/450341"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.1101\/2025.11.17.688918","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,15]],"date-time":"2026-01-15T10:03:07Z","timestamp":1768471387000},"score":1,"resource":{"primary":{"URL":"http:\/\/biorxiv.org\/lookup\/doi\/10.1101\/2025.11.17.688918"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,11,18]]},"references-count":50,"URL":"https:\/\/doi.org\/10.1101\/2025.11.17.688918","relation":{},"subject":[],"published":{"date-parts":[[2025,11,18]]},"subtype":"preprint"}}