Table 5.
Performance of KBP on prostate IMRT/VMAT (studies with 10 or more test cases)
| Articles | Method type | Sample size | Validation target | Validation metrics | Rectum | Bladder | Target |
|---|---|---|---|---|---|---|---|
| Chanyavanich et al56 | Case/voxel dose | 10 | Re‐planned vs clinical | Percent difference: mean |
D20 1.8 D30 ‐2.5 D50 ‐13.9 |
D20 ‐5.9 D30 ‐12.2 D50 ‐24.9 |
D98 ‐0.03 D95 0.62 D1 2.5 |
| Appenzoller et al.8 | Model/DVH | 20 | Predicted vs clinical |
Sum of residuals: mean Restricted sum of residuals: mean |
0.003 0.02 |
−0.008 0.013 |
|
| Yuan et al.9 | Model/DVH | 24 | Predicted vs clinical | Error bound of V99,85, 50% | 71% of cases within 6% of error bound | 71% of cases within 6% of error bound | |
| Good et al.57 | Case/voxel | 55 | Re‐planned vs clinical | Percent difference: mean |
V75 ‐1.15* V65 ‐4.10* V40 ‐11.97* |
V75 ‐0.48 V65 ‐1.18* V40 ‐2.18 |
HI ‐2.8* D1 ‐2.5* |
| Nwankwo et al.59 | Case/voxel | 33 | Predicted vs clinical | Mean voxel dose difference (magnitude) | 0.23 – 8.22 | 0.26 – 12.19 | |
| Nwankwo et al.60 | Case/voxel | 30 | Re‐planned vs clinical | Mean difference |
D10 3.0* D30 5.6* D50 2.4 D70 ‐0.3 D90 ‐0.7 |
D10 0.1 D30 ‐3.0* D50 ‐2.7 D70 0.0 D90 1.0 |
D05, D95, UI = |
| Sheng et al.63 | Atlas/voxel | 20 | Re‐planned vs clinical | Mean difference |
gEUD ++* V65 = V100 + +* |
gEUD ++* V65 + +* V100 + +* |
CI ++* HI = |
| Yang et al.20 | Model/DVH | 10 | Re‐planned vs clinical | Percent difference |
Dmax ++0.14% D10 cc ++2.11% D17 + +2.72% D40 + +0.27% |
Dmax –0.46% D10 cc –<1.54% D25 + +0.69% D40 + +0.81% |
D98 = <2.31% Dmax –0.06% |
| Boutilier et al.21 | Model/DVH | 100 | Predicted vs clinical | Absolute difference |
D30 ~10 D50 ~7 |
D30 ~7 D50 ~3 |
|
| Hussein et al.25 | RapidPlan | 10 | Re‐planned vs clinical | Mean difference |
V30 ‐0.8 V50 ‐3.1* V70 ‐0.4 D1 cm ‐0.3* |
V50 ‐3.5 V75 ‐0.2 D1 cm 0.0 |
PTV High D98 0.1 D2 0.7* PTV Inter D98 ‐0.2 D2 0.3* PTV Low CI ‐0.1* D98 0.8* D2 ‐1.2* |
| Cagni et al.27 | RapidPlan | 20 | Re‐planned vs clinical | Percent differences |
Dmean ‐1.66* V20 ‐6.32* V50 ‐1.03 V60 0.54 V65 3.71* V70 0.55 NTCP 3.02* |
Dmean 0.52 V20 0.93 V60 0.61 V70 0.15 NTCP 3.72* |
D98 0.63* D50 0.44 D2 1.09* HI 2.03* CI 5.02 |
| Masi et al.34 | RapidPlan | 10 | Re‐planned vs clinical | Mean difference |
Dmean ‐3.6* Max to 0.1 cc 0.2* V70 ‐0.8 V65 ‐1.5 V50 ‐4.0* |
Dmean ‐3.9* V75 ‐1.8* V70 ‐0.6 V65 ‐2.9* |
D95 ‐0.1 Dmax 0.3 HI5% ‐0.01* HI1% ‐0.01 |
| Schubert et al.35 | RapidPlan | 60 | Re‐planned vs clinical | Mean difference |
Dmean 0.9* D1% ‐0.4* V40 0.0 V45 ‐2.4* V50 1.6 |
Dmean 0.6* D1% 0.1 V40 ‐2.5* V45 ‐1.1 V50 1.9* |
Dmean 0.0 D2% 0.2 D98 0.0 HI 0.0 CI 0.0 |
| Wall et al.55 | Case/DVH indices | 31 | Re‐planned vs clinical | Mean difference | Dmean ‐9.41 | Dmean ‐7.81 |
V98 = V100 = Dmean = Dmax = |
| Zhang et al.42 | Model/DVH | 111 | Predicted vs clinical | Weighted root mean square error of DVH | ~3% | ~3% |
KBP, knowledge‐based planning; IMRT, intensity‐modulated radiation therapy; VMAT, volumetric arc therapy.
The difference direction is “KBP ‐ Clinical”. Thus, negative values mean KBP value is smaller. Where no value is provided, ++ indicates better metrics, – indicates worse metrics, = indicates similar metrics. The sign * means the metric is statistically significant with a P‐value < 0.05. The sign ~ indicates the value is estimated from a graph.