Diffusion-Informed Joint Segmentation Enhances Detection of Thalamic Atrophy in Parkinson's Disease

The thalamus is a critical subcortical hub that relays sensorimotor information and regulates higher-order cognitive processes. Accurate delineation of thalamic nuclei is essential for elucidating disease mechanisms and tracking clinical progression. In this study, we compared two segmentation approaches implemented in FreeSurfer: the conventional structural method and a joint framework that integrates diffusion tensor imaging. Magnetic resonance imaging (MRI) data from 24 healthy controls (HC), 27 patients with cognitively normal Parkinson's disease (PD-CN), and 33 Parkinson's disease patients with mild cognitive impairment (PD-MCI) were analyzed. Segmentation methods were compared in HC to assess their effect on volume estimates. Group comparisons were then conducted separately for each method to evaluate sensitivity in detecting disease-related volumetric differences. Finally, nuclei with significant group effects in joint segmentation were tested for associations with Addenbrooke's Cognitive Examination-Revised (ACE-R) scores. Joint segmentation yielded systematically lower thalamic volume estimates than the structural method, with significant differences across hemispheres and nuclei in HC. Group-wise analyses revealed that joint segmentation, but not structural segmentation, detected significant atrophy in the right thalamus of PD-MCI patients. At the nuclei group level, joint segmentation showed greater sensitivity, identifying bilateral anterolateral and posterior nuclei as significantly reduced in PD-MCI relative to HC. Moreover, volumes of these nuclei correlated positively with ACE-R scores. These results highlight that methodological choices critically shape the detection of thalamic pathology in PD, suggesting that incorporating diffusion MRI into segmentation may improve sensitivity to cognitively relevant subnuclear changes and support early diagnosis and disease monitoring.