Table 3 Performance evaluation of depth myometrial invasion prediction model for multiclass classification. The metrics are reported in percentage terms. The classification performances of the Radiomic-based model were evaluated in terms of mean and confidence interval 95% (CI95%).
Subjective assessment | Radiomic-based model | Subjective assessment + Radiomic-based model | ||
|---|---|---|---|---|
Overall accuracy | 74.02 | 84.93 (82.72–87.15)* | 82.82* | |
Per-class recall | Absent or focal infiltration | 93.33 | 76.67 (69.84–83.48)* | 93.33 |
Infiltration less than 50% | 56.25 | 85.63 (83.01–88.24)* | 71.88* | |
Infiltration greater than 50% | 83.33 | 88.33 (84.79–91.88)* | 90.00 | |
Per-class precision | Absent or focal infiltration | 70.00 | 78.22 (71.57–84.87)* | 82.35 |
Infiltration less than 50% | 78.26 | 86.57 (83.09–90.06)* | 85.19 | |
Infiltration greater than 50% | 73.53 | 87.34 (85.42–89.26)* | 78.79 | |
Per-class F1-score | Absent or focal infiltration | 80.00 | 76.86 (71.73–82.02)* | 87.50 |
Infiltration less than 50% | 65.45 | 86.06 (83.22–88.91)* | 77.97 | |
Infiltration greater than 50% | 78.12 | 87.69 (86.07–89.31)* | 84.40 | |
Macro F1-score | 74.52 | 84.48 (82.10–86.86)* | 83.29 |
