Emerging pathological and neurophysiological evidence has highlighted the cerebellum's involvement in Parkinson's disease (PD). This study aimed to explore the potential of cerebellum-derived magnetic resonance imaging (MRI) radiomics in distinguishing PD patients from healthy controls (HC). A retrospective analysis was conducted using three-dimensional-T1 MRI data (n= 374) from the Parkinson's Progression Markers Initiative (PPMI) dataset (n= 204) and an independent in-house cohort (n= 170). Radiomic features (n= 883) were extracted from the cerebellar gray and white matter of each individual. Three machine learning models were developed: a cerebellar gray matter model, a cerebellar white matter model, and a combined gray and white matter model, to classify PD patients and HC. The results showed that the cerebellar gray matter model achieved an area under the receiver operating characteristic (ROC) curve (AUC) of 0.931 in the training set, with a sensitivity of 60.8% and specificity of 97.1%, while in the testing set, it obtained an AUC of 0.874, with a sensitivity of 86.1% and specificity of 62.6%. The white matter-based model demonstrated an AUC of 0.846 (sensitivity, 59.8%
specificity, 87.3%) in the training set and an AUC of 0.868 (sensitivity, 81.0%
specificity, 75.8%) in the testing set. Notably, the combined gray and white matter model exhibited superior performance, achieving an AUC of 0.936 (sensitivity, 65.7%
specificity, 96.1%) in the training set and an AUC of 0.881 (sensitivity, 82.3%
specificity, 63.7%) in the testing set. Key radiomic features contributing to PD classification included Gray-level Dependence Matrix, Gray-level Co-occurrence Matrix and First-Order from gray matter, as well as Gray-level Size Zone Matrix from white matter, highlighting significant radiomics changes in the cerebellum associated with PD. In conclusion, this study demonstrates that MRI radiomics of cerebellar gray and white matter can effectively differentiate PD patients from HC, supporting the cerebellum's pivotal role in PD pathology and its potential as an imaging biomarker for PD.