A Medical Low-Back Pain Physical Rehabilitation Dataset for Human Body Movement Analysis

TL;DR

This work tackles the need for clinically relevant datasets to support automatic coaching of physical rehabilitation for low-back pain. It introduces the Keraal dataset, comprising multi-modal data (3D Kinect skeletons, RGB video, 2D skeletons, and medical annotations) collected from clinical patients during a rehabilitation program, and provides initial benchmarks using Gaussian Mixture Models and LSTM networks. The authors formalize four rehabilitation-analysis challenges—motion assessment, error recognition, and both spatial and temporal localization of errors—and annotate data at frame level with body-part and timing information. The dataset supports development of intelligent tutoring systems and telemedicine for rehab, and the baseline results highlight the need for more advanced methods to robustly detect and localize errors in real clinical settings.

Abstract

While automatic monitoring and coaching of exercises are showing encouraging results in non-medical applications, they still have limitations such as errors and limited use contexts. To allow the development and assessment of physical rehabilitation by an intelligent tutoring system, we identify in this article four challenges to address and propose a medical dataset of clinical patients carrying out low back-pain rehabilitation exercises. The dataset includes 3D Kinect skeleton positions and orientations, RGB videos, 2D skeleton data, and medical annotations to assess the correctness, and error classification and localisation of body part and timespan. Along this dataset, we perform a complete research path, from data collection to processing, and finally a small benchmark. We evaluated on the dataset two baseline movement recognition algorithms, pertaining to two different approaches: the probabilistic approach with a Gaussian Mixture Model (GMM), and the deep learning approach with a Long-Short Term Memory (LSTM). This dataset is valuable because it includes rehabilitation relevant motions in a clinical setting with patients in their rehabilitation program, using a cost-effective, portable, and convenient sensor, and because it shows the potential for improvement on these challenges.

Figures (9)

• Figure 1: The three rehabilitation exercises in our dataset
• Figure 2: Each column illustrates 3 errors for each exercise.
• Figure 3: LSTM models for assessment and classification
• Figure 4: Confusion matrix of error detection using SVM and LSTM autoencoder.
• Figure 5: GMM baseline : true positives and true negatives evolution with respect to different thresholds employed to evaluate a movement as correct or incorrect.