The effect of glenoid rotational malalignment on best-fit circles based on AI-generated mathematically true glenoid en face views

The majority of studies that measure glenoid bone loss in the context of shoulder instability are based on a sagittal image, termed the en face view, with the aid of best-fit circles. However, the en face view has never been standardized. We created a mathematical en face view of the glenoid and demonstrated the effect that rotation of the glenoid has on best-fit circle size and position. We used a custom deep learning platform, to create a mathematically "perfect" en face view using computed tomographic (CT) scans of 40 patients with anterior shoulder instability. We retrospectively measured the effect of rotating the glenoid around the vertical and horizontal axes, and all combined positions, on the measured bone loss and on best-fit circle diameter and position. Rotation of the glenoid en face view by as little as 5° may have a statistically significant effect on bone loss, on the diameter of the best-fit circle, and on the position of the best-fit circle. We described a reproducible and mathematically "perfect" en face view, not previously described with manual or semiautomated segmentation methods. We demonstrated how small degrees of rotation from that position may affect the best-fit circle and subsequently the calculated size of the glenoid defect and the glenoid track determination. This underpins the necessity of an accurate establishment of the en face view, before positioning the best-fit circle. Using a programmatically true en face view has the potential to increase the precision of measurement of glenoid bone loss.