Vision-Based Gait Analysis for the Detection of Drug-Induced Parkinsonism (DIP)

This study introduces a vision-based gait analysis system developed to detect and quantitatively evaluate drug-induced parkinsonism (DIP) in psychiatric patients. By utilizing a hybrid model combining Convolutional Neural Networks (CNNs) and Support Vector Machines (SVMs), the system addresses the limitations of traditional gait analysis methods, which often rely on wearable sensors and subjective assessments. The research was conducted at Jianan Psychiatric Center from January 2020 to December 2022 with subjects diagnosed with schizophrenia. Utilizing high-resolution cameras and AI-based algorithms, the study accurately captured and analyzed gait features, demonstrating the effectiveness of the model in distinguishing between healthy individuals and those affected by DIP. Results from the analysis showed that the CNN-SVM model significantly outperformed standalone CNN or SVM models, achieving an accuracy of 85.1%, with sensitivity and specificity exceeding 87.8% and 90.2%, respectively. These findings not only highlight the potential of integrating AIoT with vision-based technologies for early detection and management of DIP but also suggest broader applications for AIoT in enhancing psychiatric care. The study emphasizes the importance of early detection and accurate classification of DIP, which can substantially improve treatment outcomes and patient quality of life, marking a significant advancement in the use of technology to support clinical diagnostics in psychiatric settings.