An Efficient Neural Network for the Diagnosis of Parkinson’s Disease Using Dynamic Handwriting Analysis

Dynamic handwriting analysis of patients has become an important auxiliary method for the early diagnosis and treatment of Parkinson’s disease. In this paper, a novel and efficient hybrid neural network model is proposed to evaluate the potential of handwriting sequence information in identifying Parkinson’s disease, in which the hybrid model combines the respective advantages of long short-term memory and one-dimensional convolutional neural networks. Extensive experiments on the public dataset confirm that the proposed hybrid model can effectively extract discriminative sequence information for efficient diagnosis of Parkinson’s disease, where the hybrid neural network has only \(0.084\) MB and above \(90\%\) diagnostic accuracy, making it highly reliable and robust for future clinical applications.