Table 1.
Patients’ characteristics.
Table 2.
Overview of Ten DIR Algorithms.
Fig 1.
Time trend of the relative volume change for the rPG (a), lPG (b), rSMG (c) and lSMG (d) over the course of the treatment.
Table 3.
The mean, standard deviation (SD), 95% confidence interval (95% CI) and interquartile range (IQR) of the slopes of relative volume changes for each structure.
Fig 2.
Boxplots of DSC (a), PE (b) and HD (c) by different DIR algorithms for all structures in all treatment stages. The boxes run from the 25th to 75th percentile; the two ends of the whiskers represent the 5% and 95% percentiles of the data, the horizontal line and the square in the box represent the median and mean values, respectively. The diamonds represent outliers. The letters above each box indicate whether a statistically significant difference exists between any two DIRs. No common letter between any two algorithms indicates that the two DIRs are significantly different. RR is an abbreviation for rigid registration.
Fig 3.
Boxplots of DSC (a), PE (b) and HD (c) by different DIR algorithms for the bony and soft-tissue groups in all treatment stages. The meanings of the symbols in this figure are the same as in Fig 2.
Fig 4.
Boxplots of comparisons of performance on different ROIs for representative DIR algorithms, including the HSLK (optical flow-based), OD/DISC (demons-based), LS (level-set-based), BSpline (spline-based) and rigid registration (RR).
For each subgraph, no significant difference in pairwise comparisons is marked with “~”, otherwise a statistically significant difference exists. The meanings of each box in this figure are the same as in Fig 2.
Fig 5.
Comparisons across ROIs and stages for different registration algorithms.
In each subgraph, the horizontal and vertical axes represent stages ranging from 1 to 6 and the metric values, respectively. Eleven algorithms are marked by different symbols and colors.