Fig. 3
From: Evaluation of a deep learning software for automated measurements on full-leg standing radiographs

Bland–Altman plots showing the differences between AI and ground truth predictions against their means. Plots are displayed for hip–knee–ankle angle (A), pelvic obliquity (B), leg length measured from the top of the femoral head (C), leg length from the center of the femoral head (D), femoral length measured from the top of the femoral head (E), femoral length measured from the center of the femoral head (F), and tibial length (G). The red line depicts the scenario in which AI estimates would perfectly align with the ground truth, indicating no differences between the two. The light red interval around the black dotted line corresponds to the 95% confidence interval for the mean difference. The black dotted lines at the extremities of the plot represent the upper and lower limits of agreement with their respective 95% confidence intervals in light green. For all measurements but pelvic obliquity, a mixed effects Bland–Altman analysis was used to account for dependencies within the dataset