(Sunday, 7/14/2019) 4:00 PM - 5:00 PM
Room: 302
Purpose: Accurate delineation of target and organs at risk contours is critical in treatment planning. We developed a software tool using a new quantitative metric termed the localized signed surface distance (LSSD) to give real time contouring feedback to a trainee based on local spatial information of the contour when comparing to expert contours. The purpose of this study is to reduce contouring variability using this contouring tool.
Methods: Three trainees each contoured the heart and left ventricle (LV) for three thoracic cancer patients. One trainee contoured the LV on six patients. The initial contour together with eight expert physician-drawn contours were used to generate reference contours by using the simultaneous truth and performance level estimation (STAPLE) algorithm. The reference contour was compared to the trainee contour and a LSSD map was generated for eight equally spaced sectors on each axial slice. A positive LSSD indicates the user contour is outside of the reference contour within that sector, a negative indicates the user contour is inside the reference contour. Interactively, the trainee can view the sector and LSSD value to edit their contour to better coincide with the blind reference contour. Statistics were computed on the initial and final LSSD maps for each contoured case as a metric of deviation from the expert contours.
Results: A reduction in standard deviation of the LSSD map was seen in every contoured case. For the trainee who contoured the LV on six patients, the observed reduction in standard deviation was 8.7mm, 5.3mm, 3.7mm, 1.3mm, 2.4mm, and 5.6mm between the initial and final contours for each case. The reduction in standard deviation indicates that using real-time feedback reduces contour variability.
Conclusion: A contour training tool with real-time feedback was developed and used to reduce contouring variability for both the heart and the LV.
Funding Support, Disclosures, and Conflict of Interest: This project was supported in part by the Department of Radiation Physics Internal Grant and The University of Texas MD Anderson Cancer Center Institutional Research Grant (IRG) Program.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
