Parkinson’s disease (PD) is a neurodegenerative condition that affects millions of individuals worldwide. One of its main symptoms is motor impairment, which includes tremors, rigidity, and bradykinesia. Accurate and timely diagnosis of Parkinson’s disease (PD), with corresponding stage is crucial for designing effective treatment strategies, monitoring its progression, and delivering personalized care. In this paper, we address the challenging problem of simultaneous estimation of the four different stages of PD, which are often found to be overlapped. We introduce MM-PDSnet, a hybrid ML-DL network to achieve the above goal. Our solution strategy takes advantage of the complementary information offered by multimodal data, namely, T1-weighted MRI and clinical assessment. Clinical assessment is based on MDS-UPDRS features and the H&Y scale. We apply various ML models on clinical data and select the best-performing classifier. We investigate the impact of different loss functions on the performance of our DL model on the MRI data. Our findings demonstrate that the focal loss function outperforms other loss functions in terms of accuracy. Importantly, we carry out a multivariate analysis using Hotelling’s \(T^2\) method to highlight the fact that statistically significant differences do exist in MRI scans across various stages of PD. We have evaluated our proposed method on the publicly available PPMI dataset. The experimental results show that our approach surpasses several state-of-the-art methods for PD stage estimation, achieving an accuracy of 98.15%.