Table 4 ICCs for assessment of task difficulty: the ICC between expert readers, the ICC between expert readers and deep learning model, and the ICC mentioned by previous literature [24]
Two objects | Post-nCRT ICC |
|---|---|
Two human radiologists (T2WI) | 0.739 (95% CI: 0.515, 0.865) |
DL model and human radiologists (T2WI) | 0.669 (95%CI: 0.612, 0.719) |
Two human radiologists [24] (DWI) | 0.750 (95%CI: 0.630, 0.830) |
Automated segmentation using the software [24] (DWI) | 0.530–0.660 |
Semiautomated segmentation using the software [24] (DWI) | 0.610–0.750 |