Computer Vision-Based Approach for Quantifying Occupational Therapists’ Qualitative Evaluations of Postural Control

link

Abstract

This study aimed to develop a computer vision-based approach for quantifying occupational therapists’ qualitative evaluations of postural control. The approach was developed using a deep learning model (OpenPose, pose estimation) that was used to predict the child body joints. Then compare the predicted joints with the therapist’s evaluation of the child’s postural control, in video frames and TQCE (Therapists’ Qualitative Clinical Evaluation) scores.

This research contributes to the occupational therapy field first by demonstrating the possibility of more accurate and detailed quantitative evaluations, and second by simplifying conventional evaluation tasks with the elimination of special equipment and reduced time investments, imposing less burden on clients.

Here are two important score used in this research: Static Postural Balance (SPB) score and Antigravity score (AG).

Materials and Methods

Three types of indices were calculated:

1. A conventional quantitative index measuring duration time (DT)
2. therapists’ qualitative clinical evaluation (TQCE) using the 7-point Likert scale
3. CV-based quantitative indices based on image analysis algorithms for static postural balance(SPB) and antigravity(AG) scores.