Deep Learning Techniques for Identification of Parkinson’s Disease based on MRI Data: Novel Diagnostic Framework

Background: The neurodegenerative disease known as Parkinson's Disease (PD) causes both motor and non-motor symptoms due to structural brain abnormalities and dopamine depletion. Conventional clinical approaches provide significant challenges to achieving accurate early diagnosis. MRI-based analysis supported by deep learning offers a promising diagnostic tool. Methodology: This study presents a hybrid deep learning model that combines MobileNet and ResNet50 to classify PD using brain MRI images. Preprocessing involved image resizing, Discrete Wavelet Transform (DWT) for noise reduction, CLAHE for contrast enhancement, normalization, one-hot encoding, and data augmentation. Hyperparameters were optimized across learning rates, batch sizes, and epochs. Results: Experimental results show the hybrid model achieved 97% accuracy, surpassing MobileNet (83.23%) and ResNet50 (70.05%), with low training (0.1108) and validation (0.1216) losses, indicating excellent convergence. Conclusion: The results demonstrate the hybrid model’s effectiveness in PD classification, providing a scalable and reliable AI-based framework for early detection and clinical decision support.