Multimodal visual system analysis as a biomarker of visual hallucinations in Parkinson's disease.

Visual hallucinations (VH) are present in up to 75% of Parkinson's disease (PD) patients. However, their neural bases and participation of the visual system in VH are not well-understood in PD. Seventy-four participants, 12 PD with VH (PDVH), 35 PD without VH (PDnoVH) and 27 controls underwent a battery of primary visual function and visual cognition tests, retinal optical coherence tomography and structural and resting-state functional brain MRI. We quantified cortical thickness with Freesurfer and functional connectivity (FC) of Visual (VIS), Fronto-Parietal (FP), Ventral Attention (VAN) and Dorsal Attention (DAN) networks with CONN toolbox. Group comparisons were performed with MANCOVA. Area Under the Curve (AUC) was computed to assess the ability of visual variables to differentiate PDVH and PDnoVH. There were no significant PDVH vs PDnoVH differences in disease duration, motor manifestations, general cognition or dopamine agonist therapy (DA) use. Compared to PDnoVH and HC, and regardless of DA use, PDVH showed significantly reduced contrast sensitivity, visuoperceptive and visuospatial abilities, increased retina photoreceptor layer thickness, reduced cortical thickness mostly in right visual associative areas, decreased between-network VIS-VAN and VAN-DAN connectivity and increased within-network DAN connectivity. The combination of clinical and imaging variables that best discriminated PDVH and PDnoVH (highest AUC), where within-network DAN FC, photoreceptor layer thickness and cube analysis test from Visual Object and Space Perception Battery (accuracy of 81.8%). Compared to PDnoVH, PDVH have specific functional and structural abnormalities within the visual system, which can be quantified non-invasively and could potentially constitute biomarkers for VH in PD.