An Interpretable Deep Learning Optimized Wearable Daily Monitoring System for Parkinson's Disease Patients

Abstract

Walking monitoring in the daily life of patients with Parkinson's disease (PD) is of great significance for tracking the progress of the disease. This study aims to implement an accurate, objective, and long-term walking monitoring algorithm optimized by the interpretable deep learning architecture, and explore the most representative spatiotemporal motor features. Five inertial measurement units attached to the wrist, ankle, and waist are used to collect motion data from 100 subjects during a 10-meter walking test. The raw data of each sensor are subjected to the continuous wavelet transform before training in the constructed deep learning architecture based on a 6-channel convolutional neural network (CNN). The results show that the sensor located at the waist has the best classification performance with an accuracy of 98.01% ± 0.85% and the area under the receiver operating characteristic curve (AUC) of 0.9981 ± 0.0017 under ten-fold cross-validation. The gradient-weighted class activation mapping shows that the feature points with greater contribution to PD were concentrated in the lower frequency band (0.5~3Hz) compared with healthy controls. The visual maps of 3D CNN show that only three out of six time series have a greater contribution, which is used as a basis to further optimize the model input, and the raw data processing cost is reduced by 50% while the classification performance (AUC = 0.9929 ± 0.0019) is still maintained. As far as we know, this is the first study that considers visual interpretation-based optimization of an intelligent classification model in the objective diagnosis of PD.