Does registration of PET and planning CT images decrease interobserver and intraobserver variation in delineating tumor volumes for non-small-cell lung cancer?

Abstract

To compare tumor volume delineation using registered positron emission tomography (PET)/CT vs. side-by-side image sets. A total of 19 patients with non-small-cell lung cancer had 18-fluorine-deoxyglucose (FDG)-PET scans registered with planning CT scans. The disease was Stage I-II in 26%, IIIA in 42%, and IIIB in 32%. Two radiation oncologists contoured 9 tumor volumes using registered images (registered) and 10 using separate FDG-PET images as a guide (nonregistered). A third physician, who had done the treatment planning for these patients a median of 40 months before using registered images, repeated all contours: 10 on registered images (registered/registered) and 9 without registration (registered/nonregistered). Each pair of volumes (A and B) was compared. Quantitative comparison used the concordance index, (A intersection B)/(A union or logical sum B). For qualitative analysis, pairs of volumes were projected onto digitally reconstructed radiographs. The differences were graded as insignificant, minor, moderate, or major. The median interobserver percentage of concordance among nonregistered pairs was 61% vs. 70% in the registered group (p <0.05). On qualitative analysis, in the nonregistered group, the differences were insignificant in 5, minor in 3, and moderate in 2 of 10. The differences in the registered group were insignificant in 7 and minor in 2 of 9. The median intraobserver percentage of concordance in the registered/nonregistered group was 58% vs. 71% in the registered/registered group (p = 0.10). On qualitative analysis, the intraobserver differences in the registered/nonregistered group were insignificant in 2, minor in 2, moderate in 0, and major in 5 of 9. In the registered/registered group, the differences were insignificant in 2, minor in 6, moderate in 2, and major in 0 of 10. Registration of FDG-PET and planning CT images results in greater consistency in tumor volume delineation.