Discussion
The study findings suggest that remote measurement systems are valid and cost-effective options to estimate functional movement and posture in people with axSpA. Specifically, the validity results suggest there is moderate to strong correlation and agreement between a majority of functional movements and postural alignment compared to a standard clinician assessment. The strongest correlational relationships were shown in lumbar forward flexion, lumbar side flexion, shoulder flexion, hip internal rotation, tragus-to-wall, and thoracic kyphosis, particularly in people with axSpA. The only test from either a functional movement or posture that showed no correlation between methods was found in the lumbar lordosis, which was consistent between the axSpA and non-back pain groups.
The two groups demonstrated expected clinical presentation differences, including higher BASDI and BASFI scores and more restricted range of motion and hyperkyphosis in the axSpA group. The limited range of motion among the axSpA group in all functional movements tested demonstrates the broader use of this technology in clinical groups that fall outside the normal range of motion. In the end, the results did indicate varied correlative relationships between the axSpA and non-back paing groups in several functional movement and postural tests, notably shoulder flexion, lumbar side flexion, tragus-to-wall and kyphosis. In both shoulder flexion and lumbar side flexion, the axSpA group had smaller ranges of motion compared to the non-back pain group and stronger correlation (r = .787-.906) between the CV-aided system and clinical measurement compared to the non-back pain group (r = .468-.655). One reason for this discrepancy could be due to altered anatomical landmark visibility or increased trunk compensation in higher ranges of motion as were seen in the non-back pain group. Posture measurement demonstrated similar incongruence; there was a stronger correlation in the axSpA group, who presented with more kyphotic and forward flexed posture compared to the non-back pain group. This discrepancy could stem from less accurate and reliable measurement of smaller kyphosis curvature, which is one limitation of the tragus-to-wall test which has a floor effect.
The tests that did not demonstrate strong correlation were hip abduction, cervical rotation and lumbar lordosis posture. Hip abduction was adapted into a standing test to provide a more practical testing position for video recording compared to the BASMI hip mobility test, where the patient is lying on the ground and abducting both hips to their maximum range. Although more practical to perform, standing hip abduction has challenges that include both the participant performing it correctly and the landmarks needed for automation. Participants often compensate during standing hip abduction by either elevating their ipsilateral hip or externally rotating their hip. If the clinician does not correct the compensatory movements, it could cause an overestimation of the range. Similarly, the compensatory movements can cause an overestimation or inaccurate landmark identification by the CV algorithm. Cervical rotation in a seated position with a tape measure was also chosen for its practicality. This test was taken from the EDASMI since the supine cervical rotation test in the BASMI similar camera positioning challenges as the hip mobility test. The difficulty with frontal plane measurement of a rotational movement was demonstrated in the lack of a strong correlation between the CV-aided system and clinician measurement, in both groups. Lastly, the lumbar lordosis postural alignment measured by surface topography using the Kinect sensor showed agreement, but no correlation and no significant difference between groups. This could be on account of the documented difficulty of measuring lumbar lordosis with surface measurement tools, and the practicality of clothing interference in some participants during the testing.
An important aspect of this study was the feasibility of the CV system in a home setting because of the potential for many benefits of remote testing. The first barrier for the participants was uploading the videos, which was less successful in the axSpA group (n=8 missing) than the non-back pain group (n=1 missing). Developing a user-friendly interface for uploading videos would lower the barrier for home use. Two other aspects of feasibility at home were the ability of participants to successfully record the correct movement and the quality of the videos for automated CV analysis. More than 70% of the recorded videos were useable. The reasons for non-usable data were incorrect use of the calibration grid, camera movement and incompatible data format from one participant’s smartphone. These issues could be addressed by improving instructions and calibration method. Pragmatic use of this technology at home would be a key to helping people, with and without back pain, track and maintain functional movement, range of motion and posture with the option of remote clinician support. Not only does this remote system widen accessibility to specialists who may not be local, it is a cost-efficient method and has many social and environmental benefits. It can benefit both patients and the health system in terms of time and opportunity. Furthermore, it can have a positive environmental impact by reducing the carbon footprint associated with each face-to-face visit. For the appropriate patient, it could result in a cost savings of £64 per assessment. While these results look specifically at people with axSpA, it can reasonably be generalised to similar long-term musculoskeletal conditions.
Limitations
The limitations of this study include the relatively small sample size and the cross-sectional method. While it was not possible for simultaneous measurement from the clinician and video, the repetitions were performed within the same session under the same conditions. Further studies would benefit from repeated testing to measure the sensitivity to change of these remote technologies.
Conclusions
Using a clinically-validated computer vision system in rehabilitation has the potential to reduce inequality in the health system and make it more cost and time effective for both patients and the health system. In addition to demonstrating validity in most of the functional and postural measures, results show that using this CV-aided system in a home environment is a safe and feasible method which can widen accessibility and affordability.