Innovative Deep Learning Approach for Parkinson's Disease Prediction: Leveraging Convolutional Neural Networks for Early Detection

INTRODUCTION: Parkinson's disease (PD) is a progressive neurodegenerative disorder affecting movement control, highlighting the importance of timely detection and intervention to improve patient quality of life. However, accurate diagnosis remains challenging due to its similarity with other neurological conditions, leading to a 25% rate of inaccurate manual diagnoses. Convolutional Neural Networks (CNNs) offer a promising solution for medical image classification and analysis, capable of learning complex patterns in images. In this study, we introduce an innovative automated diagnostic model using CNN that gives an appropriate output about if the person is diagnosed with PD or not. OBJECTIVES: The study aims to develop an automated diagnostic model using CNNs to accurately diagnose PD. By leveraging the Parkinson Progression Markers Initiative (PPMI) dataset, which provides benchmarked MRI images of PD and healthy controls, the model seeks to differentiate between PD and non-PD cases. METHODS: A Convolutional Neural Network (CNN) is a deep learning algorithm that is suitable for medical image classification and analysis as they are able to learn complex patterns in images and identify the hidden patterns and trend of data. We have used VGG16 and ResNet50 pretrained CNN models to achieve high accuracy and prediction. RESULTS: These models collectively achieved an outstanding accuracy rate of 97%. To validate our model performance, we test our model by applying various algorithms and activation functions such as EfficientNetB0, EfficientNetB1 and softmax, sigmoid, and ReLu respectively. CONCLUSION: This research introduces an innovative framework for the early detection of Parkinson’s disease using convolutional neural networks. Our system demonstrates remarkable capability to identify subtle patterns indicative of PD in its early stages.