OBJECTIVE: Visual disability in Parkinson's disease (PD) is not fully explained by retinal neurodegeneration. We aimed to delineate the brain substrate of visual dysfunction in PD and its association with retinal thickness. METHODS: Forty-two PD patients and 29 controls underwent 3-Tesla MRI, retinal spectral-domain optical coherence tomography, and visual testing across four domains. Voxel-level associations between gray matter volume and visual outcomes were used to define a visual impairment region (visualROI). Functional connectivity of the visualROI with brain networks was analyzed. Covariance analysis of brain regions associated with retinal thinning (retinalROI) was conducted using hierarchical clustering to develop a model of retinal and brain neurodegeneration linked to disease progression. RESULTS: The amygdala was the primary component of the visualROI, comprising 32.3% and 14.6% of its left and right volumes. Functional connectivity analysis revealed significant disruptions between the visualROI and medial/lateral visual networks in PD. Covariance analysis identified three clusters within retinalROI: (1) the thalamic nucleus, (2) the amygdala and lateral/occipital visual regions, and (3) frontal regions, including the anterior cingulate cortex and frontal attention networks. Hierarchical clustering suggested a two-phase progression: early amygdala damage (Braak 1-3) disrupting visual network connections, followed by retinal and frontal atrophy (Braak 4-5) exacerbating visual dysfunction. INTERPRETATION: Our findings support a novel, amygdala-centric two-phase model of visual dysfunction in PD. Early amygdala degeneration disrupts visual pathways, while advanced-stage disconnection between the amygdala and frontal regions and retinal neurodegeneration contributes to further visual disability.