GTV spatial conformity between different delineation methods by 18FDG PET/CT and pathology in esophageal cancer

To find optimal threshold of length and GTV delineation for esophageal cancer using 18FDG PET/CT.

Sixteen patients with esophageal carcinoma underwent surgery. For each patient, six GTVs were defined. GTVCT was based on CT data alone. GTV20%, GTV40%, GTV2.5 and GTV40%M were generated by PET/CT, using SUVbgd + 20%(SUVmax(slice)--SUVbgd), SUVbgd + 40%(SUVmax(slice)--SUVbgd), 2.5 and 40%SUVmax(total) as thresholds. GTVpath was derived from pathology. Lengths of GTVs were recorded as LCT, L20%, L40%, L2.5, L40%M and Lpath, respectively. The former five GTVs/lengths were compared with GTVpath/Lpath by means of a conformity index CI/CI', which is the square of intersection of two GTVs/lengths divided by their product.

Mean LCT, L20%, L40%, L2.5, L40%M and Lpath were 6.30 +/- 2.69, 5.55 +/- 2.48, 6.80 +/- 2.92, 6.65 +/- 2.66, 4.88 +/- 1.99 and 5.90 +/- 2.38 cm. Mean , , , and were 0.68 +/- 0.16, 0.84 +/- 0.17, 0.76 +/- 0.14, 0.78 +/- 0.15 and 0.80 +/- 0.11. and was significantly superior to (P < 0.05). Mean GTVCT, GTV20%, GTV40%, GTV2.5, GTV40%M and GTVpath were 29.16 +/- 18.56, 18.75 +/- 12.37, 12.52 +/- 8.08, 22.69 +/- 14.84, 9.18 +/- 5.96 and 28.16 +/- 17.02 cm3. Mean CIs increased significantly from CI40%&path(0.27 +/- 0.09) and CI'40% M&path (0.28 +/- 0.08) < CI'20% & path (0.52 +/- 0.16) and CI'2.5&path (0.52 +/- 0.20) < CICT&path(0.77 +/- 0.17).

The SUVbgd + 20% (SUVmax(slice)--SUVbgd) method optimally estimated gross tumor length, but only reached an unsatisfactory CI for GTV. Due to possible motion factor enveloped in PET images and lack of histopathologic transverse reference, the information from both PET and CT should be referred to complementarily when delineating GTV.