Tools for consensus analysis of experts' contours for radiotherapy structure definitions

Area of Science:

• Medical imaging and radiation oncology
• Computational anatomy and image analysis
• Biostatistics and data analysis

Background:

• Accurate delineation of organs and tumors is crucial for effective radiation therapy planning.
• Manual contouring by multiple experts can lead to inter-observer variability.
• Developing tools to standardize contouring is essential for improving treatment consistency.

Purpose of the Study:

• To evaluate a novel software tool designed for estimating and analyzing consensus contours.
• To assess the tool's ability to standardize contouring from multiple expert radiation oncologists.
• To quantify the agreement and improvements in contouring before and after expert consensus.

Main Methods:

• Development of statistical methods and a graphical user interface for contour analysis.
• Utilized three breast cancer CT scans from the RTOG Breast Cancer Atlas Project.
• Employed qualitative and quantitative evaluations, including Dice-similarity and Dice-Jaccard coefficients, to compare contours.

Main Results:

• High agreement levels observed for structures like the left breast, lumpectomy, and heart.
• Significant improvements in contouring accuracy for the chest wall and breasts post-consensus.
• Dice similarity coefficients of 0.80 or greater achieved for key structures including heart, breasts, and chest wall.

Conclusions:

• The developed consensus software tool, incorporating STAPLE estimates, effectively generates contours comparable to those created by expert physicians.
• The tool aids in reducing inter-observer variability and enhancing the reliability of contouring in radiation oncology.
• This technology has the potential to improve the consistency and quality of radiation treatment planning.