Fig. 1: SCAI-gait framework.
From: 3D pose estimation for scalable remote gait kinematics assessment
incorporates a scalable remote assessment of the patient’s gait to identify atypical gait patterns for medical expert to monitor rehabilitation and preventive strategies to improve the patient’s gait. Extracting 3D pose with neural networks for joint keypoints which are then used to calculate 3D joint camera frame angles. These camera frame angles are pre-processed using K-Means Clustering approach followed by spatiotemporal gait feature extraction qualitatively through K-Means Classifier and quantitatively through SHapley Additive exPlanations (SHAP) values extracted from the Multi-Layered Perceptron (MLP) Classifier. The final outcome is an automated report to the medical expert who can review it to provide a clinical assessment as well as rehabilitation strategies for the patient.
