Room: AAPM ePoster Library
Purpose: Objective assessment of deformable image registration (DIR) accuracy often relies on the identification of anatomical landmarks in image pairs, a manual process known to be extremely time-expensive. The goal of this study is to propose a method to automatically detect vessel bifurcations in images and assess their use for the computation of target registration errors (TRE).
Methods: Two datasets were retrospectively analyzed. The first dataset included 10 pairs of inhale/exhale phases from lung 4DCTs, with 300 corresponding landmarks available for each case (DIRLAB). The second dataset included 10 pairs of pre/post-radiotherapy liver contrast-enhanced CTs, each with 5 manually picked vessel bifurcation correspondences. For all images, the vasculature was autosegmented by computing and thresholding a vesselness image. Images of the vasculature centerline were computed and bifurcations were detected based on centerline voxel neighbors’ count. The vasculature segmentations were independently registered using a Demons algorithm between representations of their surface with distance maps. Detected bifurcations were considered as corresponding when distant by less than 4mm after vasculature DIR. For each reference landmark pair, the closest detected correspondence was selected. All pairs of images were registered considering rigid registration, Anaconda in RayStation and a Demons algorithm. Each time, the mean TRE was computed using the reference and automatically selected landmarks pairs.
Results: For the lung dataset, the mean±sd (min;max) TRE were 5.3±3.4(1.2;15)mm and 4.9±2.9(1.5;12.7)mm when considering the reference or detected landmarks, respectively. The squared correlation coefficient between the two distributions was R2=0.95. For the liver dataset, the TRE were 7.3±3.0 (2.8;16.0)mm and 6.8±3.1 (1.4;15.8)mm with R2=0.89.
Conclusion: For lungs or liver CT scans DIR, a strong correlation was obtained between TRE calculated using manually picked or landmarks automatically detected with the proposed method. This tool should be particularly useful in studies requiring to efficiently assess the reliability of a large DIR number.
Funding Support, Disclosures, and Conflict of Interest: Research was supported by RaySearch Laboratories AB and UT MD Anderson, the Helen Black Image-Guided Fund, NIH (1R01CA221971-01A1, U54CA210181.01, U54CA143837, U01CA196403), the Image Guided Cancer Therapy Research Program from UT MD Anderson, the Cancer Center Support Grant (CA016672) to MD Anderson. Dr Brock has a licensing agreement with RaySearch Laboratories.
Not Applicable / None Entered.