Date Approved


Graduate Degree Type


Degree Name

Medical Dosimetry (M.S.)

Degree Program

Diagnostic & Treatment Sciences

First Advisor

Destiny Jacobs

Academic Year





The past 20 years have seen a continuous rise in the cancer patient population, resulting in staffing shortages, longer wait times for patients, and potentially lower-quality care. Artificial intelligence has been incorporated into the healthcare system and has provided ease with understaffing. Using AI technology to delineate structures can reduce planning time and variation between planners. The objective of this research is to assess the level of consistency in the rectum boundary delineation performed by different dosimetrists working at a single institution, as compared to the performance of the AutoContour software. This analysis will help determine the efficacy of AutoContour and its potential role in enhancing the accuracy and efficiency of rectum contouring.


This study compares the measurement of rectum volume using two methods: manual segmentation by experts and AI technology. Five participating dosimetrists at the same institution created rectum contours on ten male pelvis CT scans. The AutoContour software executed the contour generation process five times per CT scan, creating distinct datasets for each contour. The planner group compared the AI datasets launched successively through the AutoContour software.


In the planner group, the average dice score was 0.78, indicating a moderate level of success. However, in the group that utilized AI, the mean dice score was a perfect 1.0, signifying a highly accurate outcome.


When comparing manually contoured structures to AI-driven contours, AI revealed superior consistency and agreement across all launches compared to manual contours. This quantitative analysis discusses the importance of contour continuity and implements an opportunity for further qualitative assessments of AI contouring within a single institution.