Premium
SU‐E‐T‐654: Quantifying Plan Quality Can Effectively Distinguish Between Competing Equivocal IMRT Prostate Plans
Author(s) -
Price A,
Lo J,
Das S
Publication year - 2015
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4925017
Subject(s) - spearman's rank correlation coefficient , medicine , correlation , rank correlation , medical physics , statistics , mathematics , geometry
Purpose: The purpose of this study was to create a prostate IMRT plan quality index (PQI) that may be used to quantitatively compare competing plans using a methodology that mimics physician preference. This methodology allows planners to choose between plans with equivocal characteristics, prior to physician scrutiny. Methods: An observer study was conducted to gather data from 3 radiation oncology physicians who ranked a set of 20 patients (each with 5 plans). The rankings were used to optimize a PQI that combined weighted portions of the rectum, bladder, and planning target volume DVHs, such that the relative PQI mimicked physician rankings as best as possible. Once optimized, a test study assessed the PQI by comparison to physician rankings in a new set of 25 patients (each with 4 plans). The physician group in the test study included 6 physicians, 5 of whom were not included in the modeling study. PQI scores were evaluated against the physicians’ rank list using Spearman rank correlation. Results: The optimized plan quality index combined the following DVH features: high dose regions above 40Gy/60Gy (rectum/bladder), organ weightings, and PTV shoulder coverage. Mean correlation of the PQI vs. physicians’ rankings in the modeling study was 0.507 (range: 0.345–0.706). By comparison, the mean correlation between physicians was 0.301 (range: 0.242–0.334). The mean correlation of the PQI vs. physician rankings in test study was 0.726 (range: 0.416–0.936), indicating robustness of the PQI by virtue of producing similar results to the modeling study. Intra‐physician correlation was 0.564 (range: 0.398–0.689). Conclusion: The correlation coefficients of the PQI vs. physicians were similar to the correlation coefficients of the physicians with each other, implying that the PQI developed in this work shows promise in reflecting physician clinical preference when selecting between competing, dosimetrically equivocal plans.