Validation of Catheter Segmentation for MR-Guided Gynecologic Cancer Brachytherapy
Barber, Lauren Ashley
MetadataShow full item record
CitationBarber, Lauren Ashley. 2017. Validation of Catheter Segmentation for MR-Guided Gynecologic Cancer Brachytherapy. Doctoral dissertation, Harvard Medical School.
AbstractPurpose: To present a catheter segmentation method for use with 3T MR-guided brachytherapy and validate its results on a phantom and in clinical cases. We report a novel image-processing method for catheter segmentation that extends the distal catheter tip, interactively provided by the physician, to its proximal end, using catheter geometry, appearance, and behavior in MRI.
Methods: Comparisons were performed between results of iGyne and expert human segmentations on phantom and patient MRI. In phantom experiments, the maximum disagreement between automatic and manual MR segmentation, as computed using the Hausdorff distance (HD), was 1.5 mm, which is the same order as the MR image spatial resolution. The disagreement between automatic segmentation of MR images and “the ground truth” (which is manual segmentation of CT images) was 3.5mm. Based on the failure modes demonstrated by these initial studies, a second algorithm (NeedleFinder), guided by a catheter-specific mechanical model, performed segmentation using image features with a final quality control step to remove outliers or conflicting catheter trajectories.
Results: The initial segmentation method (iGyne) was applied to 10 interstitial brachytherapy patients, which included a total of 101 catheters. Compared with manual segmentations, the automatic method correctly segmented 93 out of 101 catheters, at an average rate of 0.3 seconds per catheter (correct defined as HD < 2mm, which is the catheter diameter). The mean Hausdorff error for the improved, mechanically-based algorithm (NeedleFinder) on a 54 patient, 760 catheter reference database was 1.49; 51 of the outliers deviated more than two catheter widths (3.4 mm) from the gold standard, corresponding to catheter identification accuracy of 93%. In a multi-user simulation experiment for evaluating RMS precision by simulating varying manually provided superior tip positions, 3σ maximum error was 2.44 mm.
Conclusions: These results suggest that the speed, precision, and accuracy of the proposed catheter segmentation method allow for technical and clinical feasibility. Future directions will continue to compare patient MRI and CT to determine target doses and dosages to organs-at-risk. If, according to published guidelines, the MR-based dose and complication rate are significant improvements over those of CT, then a case will be made for performing MR-based treatments instead of CT.
Citable link to this pagehttp://nrs.harvard.edu/urn-3:HUL.InstRepos:41973450