Planning- and repeat-CT scan in sagittal and coronal view of one example case (volume ratio = 3.1) with contours of bladder and GTV, together with extracted 3D bladder, GTV, and lipiodol surfaces. In the sagittal view of both CT scans, one visible lipiodol spot is indicated by the arrow. The white spots on the 3D bladder surface are a result of lighting that was applied for ease of visualization. This figure serves as a reference for Figs. 6 and 7 .
Geometric accuracy (SDE, SCE, and ICE) vs bladder weight (w) for bladder and GTV structures, separated by volume ratio category (low and high), for λ = 10−8, λ = 10−6, λ = 10−4. The symbols indicate the mean and the bars indicate the 95% confidence interval of the mean values.
Anatomical accuracy (RDE) versus bladder weight (w) for lipidol (circles) and GTV (squares) for high (solid line and closed symbols) and low (dashed line and open symbols) volume ratio. The symbols indicate the mean and the bars indicate the 95% confidence interval of the mean values. Original average distance between lipiodol spots (after rigid bone match) was 2.5 and 6.3 mm for low and high volume ratio, respectively, and for GTV, 2.9 and 4.0 mm for low and high volume ratio, respectively.
Box plot of lipiodol (a) and GTV (b) RDE values for four registration groups, for all cases (volume ratio 1.01–3.23). Horizontal lines indicate statistical significant difference. Open circle are outliers, i.e., values that are lower than the 25th percentile or higher than the 75th percentile; closed circles are extreme values, i.e., values that are more than 3 times the interquartile range.
Lipiodol RDE values versus the relative volume ratio for three registration groups. The lines indicate the best linear fit. Different patients are indicated with different symbols. All correlations are significant with p < .0001. Regression equations (VR = volume ratio): Bone match: RDE = −1.5 +3.1•VR; R = 0.34, S-TPS-RPM: RDE = −0.94 + 2.3•VR, w-S-TPS-RPM w = 0.8: RDE = 0.32 + 0.99•VR. For each volume ratio, two or three data points per group are visible, which represent different lipiodol spots.
Visual representation of registration results of one example case, with transformations applied to the GTV and lipiodol spots. 3D bladder representation is the same as in Fig. 1 . Planning bladder (dark surface) is registered to the bladder in repeat-CT (light surface) and the resulting transformation is applied to sampled surface points (solid circles) of the planning GTV [(a)–(e)] and to sampled surface points of lipiodol spots [(f)–(j)]. The transformed GTV and lipiodol points are indicated by the darker solid circles. w refers to w bladder and w GTV = 1 − w. (a) and (f) Original position of the GTV and lipiodol spots after bone match. (b) and (g) S-TPS-RPM registration with λ = 10−4; (c) and (h) w-S-TPS-RPM registration with λ = 10−8 and w = 0.8; (d) and (i) w-S-TPS-RPM registration with λ = 10−8 and w = 0.5; (e) and (j) w-S-TPS-RPM registration with λ = 10−8 and w = 0.2. The corresponding RDE values are also listed per case, together with the original distance between the spots (i.e., after bone match).
Visual representation of the transformation of an example registration case by using a system of curves. 3D bladder representation is the same as in Fig. 1 . (a) Planning bladder with curves indicated by black dots. Points on the curves that are located in the area of the GTV are indicated by grey-filled circles. (b) Deformed curves as a result of transformation by the S-TPS-RPM registration with λ = 10−4. (c)–(e) Deformed curves as a result of transformation by the w-S-TPS-RPM registration with λ = 10−8, for w = 0.8, w = 0.5, and w = 0.2, respectively. w refers to w bladder and w GTV = 1 − w.
Lipidol RDE values and statistics for different registration groups.
GTV RDE values and statistics for different registration groups.
Article metrics loading...
Full text loading...