Utilizing Keras Model for Dynamic Drawing Analysis in Predicting Parkinson’s Disease through Spiral and Wave Patterns

Identifying concrete cues associated with specific health issues is critical to developing effective clinical decision support systems. For patients with Parkinson’s disease (PD), the severity of the disease has been found to be directly reflected in impaired handwriting. In addition, individuals with PD exhibit decreased movement and pen pressure displayed during writing or drawing. Accurate detection of these symptoms at the onset of disease improves clinical diagnosis. This paper proposes a framework for analyzing spiral and wave patterns in PD patients and healthy subjects. The system uses the Keras model to estimate drawing patterns and uses ensemble classification based on group results to predict the probability of spiral and wave sketches. This model is trained on patient data, the model has 93.4 % accuracy overall, 94% average recall, 93.5% average accuracy, obtaining an f1 score of 93.94%.