Effect of Using Multiple Imaging Modalities on GTV Delineation of Sarcomas and Chordomas: a Reproducibility Study

Introduction: Accurate delineation of the gross target volume (GTV) is critical for radiation treatment planning of soft tissue sarcomas and chordomas. Inter- and Intra-physician delineation variability is a major challenge. The aim of this work is to improve the accuracy and reproducibility of GTV by using multi-modality versus CT. We compared GTV based on CT alone vs CT+MR vs CT+MR+PET. We also tested intra-observer variability in repeated GTV delineation.

Methods: Three GTV contours were repeatedly delineated by one physician on eight patients, who underwent CT, MR and PET. Images from different modalities were registered using the MIMVista software. First, GTV contours were drawn on CT for all patients. 3D contouring was performed slice by slice using the MIMVista software. Multiple trials were performed for each patient in a randomized order to avoid recall bias. Subsequently, contours were drawn on CT using fused CT+MR in a random order and finally using CT+MR+PET. For each modality, intra-observer variability was measured by calculating the 3D Dice score between pairs of trials, resulting in three Dice scores per modality and patient. The Dice score measures the similarity between masks and is defined by D(X, Y) = 2 | intersection(X, Y) | / (|X| + |Y|) where X and Y are two segmentation masks and |.| calculates then number of pixels in a binary mask. Higher scores indicate larger agreement between masks.

Results: An example of contours obtained for three trials for one of the patients is shown on Fig. 1. The Dice scores are reported in Table 1. The difference in average Dice score between CT and CT+MR+PET is statistically significant (t-test p-value 0.02, Wilcoxon p-value 0.05). Additionally, remote sites for one patient were detected only when using PET. The full results of this ongoing study will be presented at SNMMI. Comparison with another reader will also be done to test for inter-reader variability. This full dataset will be processed to test whether our hypothesis on the reduction in variability when using multiple modalities for GTV stands.

Conclusions: This work studies the effect of using multiple modalities (CT, MR and PET) when performing GTV delineation. Preliminary results suggest that using CT along with MR and PET significantly reduces the intra-reader variability. Additionally, PET can help the detection of remote sites. These observations demonstrates the added value of PET & MR for treatment planning of soft-tissue sarcomas and chordomas. Research support T32EB013180, R01CA165221, R21EB021710 and P41EB022544. $$graphic_C34D3669-D672-4EA5-B30B-19595FED9259$$

• Download figure
• Open in new tab
• Download powerpoint