P2.05-040 Interobserver Variability in the Definition of the Primary Lung Cancer and Lymph Nodes on Different 4DCT Reconstructions: Topic: RT Techniques

Abstract

Delineation variability is a major uncertainty in radiotherapy for lung cancer. As respiratory motion is an important part of this uncertainty, respiratory correlated computed tomography (4DCT) imaging is widely used. Several image reconstruction techniques are available to generate 3D data for delineation, such as the Maximum Intensity Projection (MIP) and the mid-ventilation (MidV) technique. The latter selects data at the time weighted mean tumor position. Both techniques are prone to motion artefacts. The new Mid-position (MidP) technique averages CT data after motion compensation to the mean position reducing such artefacts. The aim of this study is to evaluate interobserver variation for tumor delineation for these three image reconstruction techniques. 4DCTS of 10 patients were reconstructed using MIP, MidV and MidP methods. Seven specialised radiation oncologists delineated the primary tumor (GTVp) and lymph nodes (GTVln) on each reconstruction with a minimum of 4 weeks interval between delineations, using a provided protocol. The interobserver variation in the delineation of the GTVs was evaluated by calculating delineated volumes, conformity index (CI), and local SD between delineated contours (SDlocal) for GTVp and GTVln. The differences in delineation variability are small (Table 1), with the only significant differences in overall volume (Friedman test): the MidP volume is slightly smaller than the MidV volume indicating a larger confidence in delineation. Counter-intuitively the observer variation was higher on the Midp images for the GTVln which seems to be related to increased reliance on the PET-CT images when delineation on the lower quality MIP images.Table 1Interobserver variability in the delineation of the GTVp and GTVlnImageMean Volume (cc)Mean CIMean SDlocal (cm)ALLGTVpGTVlnMIP81.910.5680.3630.3300.477MidV66.170.5760.3270.3140.425MidP62.230.6020.3370.2610.543MIP>MidV (p=0.000) MIP>MidP (p=0.000) MidP<MidV (p=0.115)p=0.354p=0.655p=0.648p=0.834p Open table in a new tab Although not statistically significant, the MidP images had the highest CI, lowest volume and the lowest SD for the GTVp but not for the GTVln. Overall the MidP had the smallest interobserver variation. Adherence to delineation protocols for lymph nodes must be improved to benefit from the better image quality of MidP.