CONNECTOME-BASED BRAIN FINGERPRINTS PREDICT ART THERAPY RESPONSE IN PARKINSON’S DISEASE

Ielo A^1,§, Genovese D^2,3,§, Falcó-Roget J⁴, Amico E^5,6, Norcini M⁷, Quartarone A¹, Ghilardi MF⁸ and Cacciola A^9,10 | ¹IRCCS Centro Neurolesi Bonino Pulejo, Messina, Italy; ²Institute of Neurology, Fondazione Policlinico Universitario A. Gemelli IRCCS, Rome, Italy; ³Department of Neuroscience, Università Cattolica del Sacro Cuore, Rome, Italy; ⁴Computational Neuroscience Group, Sano Centre for Computational Medicine, Kraków, Poland; ⁵School of Mathematics, University of Birmingham, Birmingham, UK; ⁶Centre for Human Brain Health, University of Birmingham, Birmingham, UK; ⁷NYU Langone Health, Department of Neurology, New York, USA; ⁸Department of Molecular, Cellular & Biomedical Sciences, School of Medicine, City University of New York, New York, USA; ⁹Department of Biomedical Sciences, Humanitas University, Pieve Emanuele, Milan, Italy; ¹⁰IRCCS Humanitas Research Hospital, Rozzano, Milan, Italy. ^§Shared co-first authorship

Art therapy has emerged as a promising non-pharmacological, complementary approach in Parkinson’s Disease (PD), engaging motor, cognitive, and emotional systems and promoting neuroplasticity. However, individual responses to art therapy are highly variable, and predictors of therapeutic efficacy remain largely unknown. Here, we propose a novel framework combining brain fingerprinting and machine learning to predict art therapy outcomes in PD. We mapped functional connectomes from restingstate functional MRI (fMRI) of PD patients before and after art therapy, assessed individual connectome-based fingerprints, and examined their spatial specificity. First, we found that functional connectomes derived from resting-state fMRI retained high levels of identifiability in both healthy controls and patients with PD, indicating that stable and subject-specific brain fingerprints are preserved even in the context of neurodegeneration. Second, we identified associations between fingerprint topography and cognitive domains relevant to PD. Although mean edge-wise reliability was comparable across conditions, the topography of reliability differed by group and session. In PD, reliable edges shifted toward posterior midline and occipital territories, with fewer stable connections in several associative systems. This pattern suggests a shift in reliability toward sensory-perceptual hubs and away from networks that support flexible control. Finally, leveraging network fingerprints, we computed subject-wise topology measures which served as input to a supervised classification framework designed to predict clinical responder versus non-responder status, based on changes in UPDRS-III scores. Among the tested classifiers, treebased methods provided the most robust predictive performance, with random forest achieving the highest performance, with an accuracy of 0.83 and a ROC-AUC of 0.80. Our results demonstrate that brain fingerprint-informed network measures capture interindividual variability in art therapy response, offering a datadriven, personalized approach to rehabilitation. This study provides the first evidence that functional connectome fingerprints can guide art therapy interventions, thus opening a new precision medicine framework in PD.