A qualitative and a quantitative analysis of an auto-segmentation module for prostate cancer.

PURPOSE: This work describes the clinical validation of an automatic segmentation algorithm in CT-based radiotherapy planning for prostate cancer patients. MATERIAL AND METHODS: The validated auto-segmentation algorithm (Smart Segmentation, version 1.0.05) is a rule-based algorithm using anatomical reference points and organ-specific segmentation methods, developed by Varian Medical Systems (Varian Medical Systems iLab, Baden, Switzerland). For the qualitative analysis, 39 prostate patients are analysed by six clinicians. Clinicians are asked to rate the auto-segmented organs (prostate, bladder, rectum and femoral heads) and to indicate the number of slices to correct. For the quantitative analysis, seven radiation oncologists are asked to contour seven prostate patients. The individual clinician contour variations are compared to the automatic contours by means of surface and volume statistics, calculating the relative volume errors and both the volume and slice-by-slice degree of support, a statistical metric developed for the purposes of this validation. RESULTS: The mean time needed for the automatic module to contour the four structures is about one minute on a standard computer. The qualitative evaluation using a score with four levels ("not acceptable", "acceptable", "good" and "excellent") shows that the mean score for the automatically contoured prostate is "good"; the bladder scores between "excellent" and "good"; the rectum scores between "acceptable" and "not acceptable". Using the concept of surface and volume degree of support, the degree of support given to the automatic module is comparable to the relative agreement among the clinicians for prostate and bladder. The slice-by-slice analysis of the surface degree of support pinpointed the areas of disagreement among the clinicians as well as between the clinicians and the automatic module. CONCLUSION: The efficiency and the limits of the automatic module are investigated with both a qualitative and a quantitative analysis. In general, with efficient correction tools at hand, the use of this auto-segmentation module will lead to a time gain for the prostate and the bladder; with the present version of the algorithm, modelling of the rectum still needs improvement. For the quantitative validation, the concept of relative volume error and degree of support proved very useful.

Accuracy of patient-specific organ dose estimates obtained using an automated image segmentation algorithm.

The overall goal of this work is to develop a rapid, accurate, and automated software tool to estimate patient-specific organ doses from computed tomography (CT) scans using simulations to generate dose maps combined with automated segmentation algorithms. This work quantified the accuracy of organ dose estimates obtained by an automated segmentation algorithm. We hypothesized that the autosegmentation algorithm is sufficiently accurate to provide organ dose estimates, since small errors delineating organ boundaries will have minimal effect when computing mean organ dose. A leave-one-out validation study of the automated algorithm was performed with 20 head-neck CT scans expertly segmented into nine regions. Mean organ doses of the automatically and expertly segmented regions were computed from Monte Carlo-generated dose maps and compared. The automated segmentation algorithm estimated the mean organ dose to be within 10% of the expert segmentation for regions other than the spinal canal, with the median error for each organ region below 2%. In the spinal canal region, the median error was [Formula: see text], with a maximum absolute error of 28% for the single-atlas approach and 11% for the multiatlas approach. The results demonstrate that the automated segmentation algorithm can provide accurate organ dose estimates despite some segmentation errors.