Published in:
01-07-2014 | Original article
Reliability and accuracy assessment of radiation therapy oncology group-endorsed guidelines for brachial plexus contouring
Authors:
Joris Van de Velde, Msc, Tom Vercauteren, Msc, Werner De Gersem, Ir, PhD, Johan Wouters, Ir, Katrien Vandecasteele, MD, PhD, Philippe Vuye, Msc, Frank Vanpachtenbeke, Msc, Katharina D’Herde, MD, PhD, Ingrid Kerckaert, MD, PhD, Wilfried De Neve, MD, PhD, Tom Van Hoof, Msc, PhD
Published in:
Strahlentherapie und Onkologie
|
Issue 7/2014
Login to get access
Abstract
Purpose
The goal of this work was to validate the Radiation Therapy Oncology Group (RTOG)-endorsed guidelines for brachial plexus (BP) contouring by determining the intra- and interobserver agreement. Accuracy of the delineation process was determined using anatomically validated imaging datasets as a gold standard.
Materials and methods
Five observers delineated the right BP on three cadaver computed tomography (CT) datasets. To assess intraobserver variation, every observer repeated each delineation three times with a time interval of 2 weeks. The BP contours were divided into four regions for detailed analysis. Inter- and intraobserver variation was verified using the Computerized Environment for Radiation Research (CERR) software. Accuracy was measured using anatomically validated fused CT–magnetic resonance imaging (MRI) datasets by measuring the BP inclusion of the delineations.
Results
The overall kappa (κ) values were rather low (mean interobserver overall κ: 0.29, mean intraobserver overall κ: 0.45), indicating poor inter- and intraobserver reliability. In general, the κ coefficient decreased gradually from the medial to lateral BP regions. The total agreement volume (TAV) was much smaller than the union volume (UV) for all delineations, resulting in a low Jaccard index (JI; interobserver agreement 0–0.124; intraobserver agreement 0.004–0.636). The overall accuracy was poor, with an average total BP inclusion of 38 %. Inclusions were insufficient for the most lateral regions (region 3: 21.5 %; region 4: 12.6 %).
Conclusion
The inter- and intraobserver reliability of the RTOG-endorsed BP contouring guidelines was poor. BP inclusion worsened from the medial to lateral regions. Accuracy assessment of the contours showed an average BP inclusion of 38 %. For the first time, this was assessed using the original anatomically validated BP volume. The RTOG-endorsed BP guidelines have insufficient accuracy and reliability, especially for the lateral head-and-neck regions.