Enhanced Early Parkinson's Disease Detection Using Resnet-101 Based on MRI Images

Early diagnosis of this specified disease is believed to be determing in arresting the progression and allowing patients a chance at disease modifying treatment. One of the primary opportunities for such diagnosis is provided by the premotor stage when slight symptoms may herald the beginning of the disease. Traditional approaches have used the SVM algorithm in predicting the onset of PD; however, they are quite limited in their capabilities concerning the processing of image data. This is a huge disadvantage because image data, such as MRI scans or other medical images, play a very significant role in coming up with accurate diagnoses. To attempt to overcome this limitation, we use the ResNet-101 deep learning algorithm. ResNet-101 is very powerful because it is a CNN, which is an excellent network for image analysis because its deep architecture has 101 layers. This depth allows it to identify detailed patterns and features from medical images, significantly improving prediction accuracy as compared to SVM. We will make it possible for early diagnosis of PD by image data, which also means diagnosis at a premotor stage with much higher precision using ResNet-101. This solution increases the accuracy in predictions but also creates a strong tool for the early diagnosis by the clinicians so that interventions may be provided at a timely and potentially life-altering point. The proposed results shows better and outperformed well than the existing approaches. This research improves prediction accuracy 96.0%, Sensitivity 97.5%, Specificity 96.8%, Precision 94.5%, F1-Score 92.7%, ROC-AUC 98.0%.