Table 3.

Evaluative studies of the computer vision-based analysis system for scoliosis screening

Author	Patients (n)	Computer vision	Comparison/evaluation	Results	Year
Yıldırım et al. [17]	42	Hand-held 3D scanner with tablet	Conventional ultrasound scoliosis diagnosis system/point-to-point matching, correlational analysis	Root mean square, correlation (rmax = 0.92, rmin = 0.47 in standing posture)	2021
Lai et al. [37]	19	3D ultrasound imaging system	Commercial 3D ultrasound imaging system/absolute dataset difference	Absolute difference between the two data sets (2.9° ± 1.8°)	2021
Zhang et al. [34]	367	Built-in smartphone camera	Plain X-ray images/deep learning-based vertebral landmark detection and difference analysis	Average L2 error (2.8 pixels), Recall (0.99)	2021
Cho et al. [36]	629	U-net segmentation, binary mask	Plain X-ray images/Cobb angle measurement difference	Matching score (0.821), mean absolute error of 8.055° for Cobb angle	2020
Mishra et al. [33]	22	3D vision with surface topography	Plain X-ray images/topographical differences	Standard deviation (± 3.4)	2020