Table 4 Comparison of inter-reader agreement between the board-certified radiologist (reader 1) and the intensively trained radiology resident (reader 2) and the respective agreement of these readers with the predicted segmentations of the nnU-Net on the preHIFU data in the hold-out test set (n = 11)
Dice score | Relative volume difference | |
|---|---|---|
Reader 1 vs. Reader 2 | 0.92 ± 0.02 | 3.69% ± 3.54% |
Reader 1 vs. nnU-Net | 0.94 ± 0.02 | 3.63% ± 2.91% |
Reader 2 vs. nnU-Net | 0.91 ± 0.03 | 4.49% ± 5.26% |