Figure 4.
Violin plots of pixel area differences (annotation/AI prediction–GT, in mm2) for each method (ReticularNet and ophthalmologists 1–4). The width of each violin reflects the full distribution of per-image differences, the black diamond denotes the mean difference, and the vertical error bar shows the 95% confidence interval. The horizontal dashed line at zero indicates perfect agreement between predicted/annotated and GT areas. AI = artificial intelligence; GT = ground truth.
