Range of Target | PTVGS
| RFB
| RFB-Aug
| RAIP
| RMIP
|
---|
Motion (cm) | (cc) | | | | |
---|
1 | 48.4 | 0.88 (0.740) | 0.99 (0.913) | 0.89 (0.886) | 1.02 (0.913) |
2 | 58.4 | 0.73 (0.633) | 0.97 (0.905) | 0.87 (0.860) | 1.02 (0.910) |
3 | 71.1 | 0.52 (0.515) | 0.95 (0.884) | 0.86 (0.864) | 1.02 (0.894) |
4 | 84.7 | 0.41 (0.408) | 0.93 (0.877) | 0.85 (0.850) | 1.02 (0.879) |
- In theory, a maximal DSC = 1.0 would correspond to R = 1.00. In this case, both volumes are not only numerically the same but also occupy the same position in physical space. However, a perfect R value can have DSC = 0.0 which would indicate the two volumes are the same but have no spatial overlap. It is interesting to note in this phantom study the motion is restricted to only one dimension, for which the centroid positions are similar with the exception of the FB technique where there is a relative superior/inferior (Δz) shift with respect to other target delineation strategies. This is true because FB is only capturing a 3D snapshot of the target at a random location along the superior/inferior axis. Therefore, the fundamental difference between FB-Aug, AIP and MIP delineation strategies is the general shape when compared to ten phase. For R < 1 the physical volume was generally underestimated (ie; FB, AIP and FB-Aug), whereas when R > 1 the volume was overestimated (ie; MIP)