2.4. Statistical Analysis

Statistical analysis was carried out with the Trace4Research platform.

The accuracy of the DL-based macular detector was calculated in terms of DICE coefficient. Diagnostic performances of the ML-based radiomics model were obtained in terms of sensitivity, specificity, and accuracy.

The distribution of the most relevant features in the two classes was presented both numerically, by their medians with 95% CI, and graphically, by the use of violin and box plots for intuitive visualisation and interpretation. For each of such features, the statistical significance in discriminating the two classes was assessed by a non-parametric univariate Wilcoxon rank-sum test (Mann–Whitney U test) [29] with Bonferroni correction [30] and significance levels set at 0.05 and 0.005.