Automating the Clinical Assessment of Independent Wheelchair Sitting Pivot Transfer Techniques

Top Spinal Cord Inj Rehabil, 2020 · DOI: 10.46292/sci20-00050 · Published: July 1, 2020

Assistive Technology Bioinformatics Rehabilitation

Simple Explanation

The study aims to use a low-cost camera (Kinect v2) and machine learning to objectively assess the quality of wheelchair transfers. Machine learning classifiers were developed using data from 91 wheelchair users to discern between proper and improper transfer techniques based on the Transfer Assessment Instrument (TAI). The system aims to automate TAI scoring by combining Kinect data with clinician input, potentially improving reliability and reducing therapist burden.

Study Duration

Not specified

Participants

91 full-time wheelchair users

Evidence Level

Not specified

Key Findings

1
The machine learning models achieved high accuracy in predicting TAI item scores, with AUC values ranging from 0.79 to 0.94 and precisions over 0.87.
2
The models effectively discerned between proper and improper transfer techniques, showing promise for objective assessment using a low-cost camera.
3
Specific biomechanical features, such as joint angles and trunk displacements, were identified as important predictors of transfer quality.

Research Summary

This study investigated the use of a low-cost Kinect v2 sensor and machine learning (ML) algorithms to automate the assessment of independent wheelchair sitting pivot transfer techniques. The results demonstrated that the ML models could predict Transfer Assessment Instrument (TAI) item scores with accuracies ranging from 71% to 92%, showing promise for objective assessment of transfer techniques. The study also identified clinically relevant biomechanical features that contribute to predicting TAI item outcomes, enhancing the interpretability of the models.

Practical Implications

Objective Assessment

Provides an objective method to evaluate transfer techniques, reducing subjectivity in clinical assessments.

Reduced Therapist Burden

Automating TAI scoring can reduce therapist workload and improve assessment reliability.

Early Deficit Identification

Identifying transfer deficits early and more effectively may help reduce the prevalence of secondary injuries among wheelchair users.

Study Limitations

1
The data labeling was based on the TAI, which is typically scored after visual observation of the movement patterns and is subject to rater subjectivity and interpretation.
2
The data splitting process could have resulted in an overfitting of the models thus reducing the generalizability of our model results and the ability to make accurate predictions on new data.
3
Only a subset of the TAI items could be modelled in this study.

Your Feedback

Was this summary helpful?

Back to Assistive Technology