Enhancing Parkinson’s Disease Classification Using LSTM Networks and Attention Mechanisms

Parkinson's disease (PD) is a neurological condition that leads to both non-motor and motor symptoms, including tremors, bradykinesia, altered posture, and balance.These symptoms progressively worsen over time, emphasizing the importance of early identification in slowing the degeneration of dopamine-producing neurons and improving patients' quality of life.Despite extensive research, no therapeutic method has yet been discovered to completely eradicate PD after it has begun.The research community has been utilizing various machine-learning approaches that use custom characteristics to diagnose problems.This paper proposes a 3D convolutional recurrent neural network (CRNN) for diagnosing PD.In this study, 3T T1-magnetic resonance imaging (MRI) images were obtained for 936 individuals, including 118 control subjects, 399 prodromal subjects, and 419 PD patients.The proposed CRNN architecture integrates convolutional layers, Long Short-Term Memory (LSTM) units, and self-attention mechanisms to effectively capture spatial and temporal features from MRI slices.Achieving remarkable performance with 93.58% accuracy, 90.38% sensitivity, and 91.67% precision, the model demonstrates superior diagnostic capabilities, outperforming existing methods in PD classification.