Table 2.
Results of the VTT with three radiologists (A, B, C) for the different ICM reduction levels, models, and the slice/volume wise evaluation. Each model configuration is evaluated based on the real/fake accuracy and pathological consistency for single slices and complete volumes
| Volume | Slice | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ICM reduction | Models | Real/fake accuracy | Pathological consistency | Real/fake accuracy | Pathological consistency | ||||||||
| A | B | C | A | B | C | A | B | C | A | B | C | ||
| −80% | 2D | 100% | 59% | 100% | 23% | 55% | 32% | 88% | 90% | 76% | 85% | 80% | 90% |
| 2.5D | 100% | 100% | 100% | 27% | 45% | 45% | 90% | 88% | 82% | 92% | 81% | 90% | |
| −50% | 2D | 95% | 91% | 95% | 100% | 100% | 100% | 69% | 80% | 62% | 99% | 98% | 100% |
| 2.5D | 73% | 77% | 91% | 100% | 100% | 91% | 55% | 77% | 64% | 100% | 100% | 100% | |