Enhancing Spiral and Wave Drawing Analysis for Parkinson's Disease Diagnosis Using VGG19 with Attention Mechanism

Parkinson's disease (PD) is a brain disorder that mostly affects older people, causing problems with movement and other aspects of daily life. Doctors typically diagnose PD by looking at a person's medical history, and symptoms, and doing neurological and physical exams. Detecting PD early is important for getting timely medical help. In the field of PD research handwriting problems and repeated spiral and wave drawing assignments are early signs of PD. Recently, there's been a focus on using computer-assisted methods in Artificial Intelligence to identify Parkinson's disease. The latest investigation targets PD swiftly identified by analyzing the spiral and wave drawings of Parkinson's and normal people using a novel classification model based on a Convolution based pre-trained model (VGG19) with attention mechanism. Attention mechanisms in neural networks aim to help the model focus on important parts of an image while ignoring irrelevant areas. Adding attention layers to the VGG19 baseline model enhances its ability to spot key features, aiding better decision-making in classification. The proposed method is trained and validated on an Augmented spiral and wave drawing PD dataset. Adding an attention layer with VGG19 effectively improved the detection accuracy of 97.5% outperforming the baseline VGG 19 model.