Table 1. Median (IQR) performance metrics and time to complete for all object detection models for kidney and liver segmentation.
| Cohort | Segmentation | Metric | Model | p | ||||
|---|---|---|---|---|---|---|---|---|
| macBGRemoval | remBGisnet | remBGu2net | detectron2 | yoloV8 | ||||
| Internal validation n=246 |
Whole kidney | IoU | 0.3 (0.21–0.41) |
0.43 (0.29–0.82) |
0.54 (0.35–0.82) | 0.93 (0.9–0.95) |
0.94 (0.91–0.96) | <0.0001 |
| DSC | 0.46 (0.35–0.59) |
0.6 (0.45–0.9) |
0.7 (0.52–0.9) |
0.96 (0.95–0.97) |
0.97 (0.95–0.98) |
<0.0001 | ||
| AUROC | 0.73 (0.67–0.83) |
0.91 (0.81–0.97) | 0.92 (0.81–0.98) |
0.98 (0.96–0.98) |
0.99 (0.98–0.99) |
<0.0001 | ||
| Time (sec) | 169 | 383 | 314 | 271 | 76 | – | ||
| Time per image | 0.69 | 1.56 | 1.28 | 1.10 | 0.31 | – | ||
| Clear view | IoU | – | – | – | 0.92 (0.87–0.94) |
0.92 (0.88–0.95) |
0.001 | |
| DSC | – | – | – | 0.96 (0.93–0.97) |
0.96 (0.93–0.97) |
0.001 | ||
| AUROC | – | – | – | 0.97 (0.95–0.98) |
0.98 (0.97–0.99) |
<0.001 | ||
| Time (sec) | – | – | – | 179 | 66 | – | ||
| Time per image | – | – | – | 0.73 | 0.27 | – | ||
| External validation n=203 |
Whole kidney | IoU | 0.49 (0.33–0.8) |
0.59 (0.38–0.79) |
0.5 (0.35–0.76) |
0.93 (0.89–0.94) |
0.94 (0.91–0.96) | <0.0001 |
| DSC | 0.65 (0.5–0.89) |
0.74 (0.55–0.88) | 0.67 (0.51–0.86) |
0.96 (0.94–0.97) |
0.97 (0.95–0.98) |
<0.0001 | ||
| AUROC | 0.86 (0.78–0.96) |
0.89 (0.82–0.95) |
0.86 (0.8–0.94) |
0.97 (0.96–0.98) |
0.98 (0.97–0.99) |
<0.0001 | ||
| Time (sec) | 53 | 139 | 96 | 165 | 30 | – | ||
| Time per image | 0.26 | 0.69 | 0.47 | 0.81 | 0.15 | – | ||
| Clear view | IoU | – | – | – | 0.91 (0.87–0.93) | 0.9 (0.85–0.93) |
0.261 | |
| DSC | – | – | – | 0.95 (0.93–0.97) |
0.95 (0.92–0.96) |
0.271 | ||
| AUROC | – | – | – | 0.98 (0.97–0.99) |
0.99 (0.97–0.99) |
<0.001 | ||
| Time (sec) | – | – | – | 165 | 30 | – | ||
| Time per image | – | – | – | 0.81 | 0.15 | – | ||
| Internal validation n=120 |
Whole liver | IoU | 0.86 (0.66–0.95) |
0.81 (0.62–0.95) | 0.89 (0.72–0.95) |
0.97 (0.95–0.98) |
0.97 (0.95–0.98) |
<0.0001 |
| DSC | 0.93 (0.79–0.97) |
0.9 (0.76–0.97) | 0.94 (0.84–0.98) | 0.98 (0.97–0.99) |
0.99 (0.97–0.99) |
<0.0001 | ||
| AUROC | 0.87 (0.78–0.93) |
0.88 (0.79–0.95) |
0.9 (0.8–0.95) |
0.97 (0.96–0.98) |
0.97 (0.96–0.98) |
<0.0001 | ||
| Time (sec) | 20 | 80 | 45 | 92 | 16 | – | ||
| Time per image | 0.16 | 0.67 | 0.38 | 0.77 | 0.13 | – | ||
| Clear view | IoU | – | – | – | 0.89 (0.83–0.92) |
0.89 (0.83–0.93) |
0.360 | |
| DSC | – | – | – | 0.94 (0.91–0.96) | 0.94 (0.9–0.96) |
0.327 | ||
| AUROC | – | – | – | 0.95 (0.92–0.96) |
0.95 (0.93–0.96) |
0.920 | ||
| Time (sec) | – | – | – | 91 | 15 | – | ||
| Time per image | – | – | – | 0.76 | 0.13 | – | ||
| External validation n=208 |
Whole liver | IoU | 0.43 (0.35–0.52) |
0.56 (0.41–0.72) |
0.59 (0.44–0.71) |
0.92 (0.87–0.95) |
0.91 (0.82–0.95) | <0.0001 |
| DSC | 0.61 (0.52–0.69) | 0.72 (0.58–0.84) |
0.74 (0.61–0.83) |
0.96 (0.93–0.97) |
0.95 (0.9–0.98) |
<0.0001 | ||
| AUROC | 0.72 (0.65–0.8) |
0.83 (0.76–0.9) |
0.84 (0.78–0.91) | 0.97 (0.95–0.98) |
0.97 (0.94–0.98) |
<0.0001 | ||
| Time (sec) | 178 | 368 | 330 | 248 | 79 | – | ||
| Time per image | 0.86 | 1.77 | 1.59 | 1.19 | 0.38 | – | ||
| Clear view | IoU | – | – | – | 0.7 (0.52–0.86) |
0.64 (0.43–0.81) | <0.001 | |
| DSC | – | – | – | 0.82 (0.68–0.93) |
0.78 (0.6–0.89) |
<0.001 | ||
| AUROC | – | – | – | 0.96 (0.92–0.98) |
0.94 (0.91–0.97) | <0.001 | ||
| Time (sec) | – | – | – | 310 | 70 | – | ||
| Time per image | – | – | – | 1.49 | 0.34 | – | ||
IoU, intersection over union; DSC, dice coefficient; AUROC, area under the receiver operating characteristic curve. For comparisons between 2 groups, the Wilcoxon signed-rank test was used. For comparisons among more than 2 groups, the Friedman test was applied.