<p>Objective: To characterize white matter (WM) network abnormalities across clinical subtypes of Parkinson’s disease (PD) and to evaluate whether baseline connectomic measures are associated with subsequent clinical progression. Methods: A total of 88 PD patients and 38 age- and gender-matched healthy controls (HCs) were enrolled in the study at baseline, with a follow-up after an average of 2.75 ± 0.59 years. Cluster analysis was performed to classify patients into different clinical subtypes. Diffusion tensor imaging (DTI) scan was performed to construct WM networks, followed by graph theoretical analyses to identify changes in WM structural networks among PD clusters. Regression analysis was used to explore the relationship between baseline WM network metrics and longitudinal clinical progression. Results: Two primary PD subtypes were identified: mild PD (<i>n</i> = 51) and moderate PD (<i>n</i> = 36). Within the mild PD, two further subtypes were defined: mild motor-predominant PD (<i>n</i> = 25) and mild-diffuse PD (<i>n</i> = 26). At the global level, both mild and moderate PD subtypes exhibited significantly reduced local efficiency compared with HCs. At the regional level, moderate PD showed altered nodal efficiency in the frontal and parietal cortices, paracentral lobule, hippocampus, insula, precuneus, and cerebellum lobule X relative to mild PD. Additionally, mild-diffuse PD exhibited disrupted nodal efficiency in the middle frontal gyrus, superior temporal gyrus, and inferior parietal lobule compared with mild motor-predominant PD. Moreover, lower baseline nodal efficiency was associated with greater worsening of motor symptoms, non-motor symptoms, and cognitive decline during follow-up. Conclusion: WM network disruptions differ across PD subtypes. DTI-based graph theoretical analysis provides an objective characterization of subtype-specific network alterations and may offer prognostic value for predicting clinical progression.</p>