To determine the optimal method of using 18 F-fluorodeoxyglucose positron emission tomography (FDG-PET) to estimate gross tumor length in esophageal carcinoma.Thirty-six patients with esophageal squamous cell carcinoma treated with radical surgery were enrolled. Gross tumor volumes (GTVs) were delineated using three different methods: visual interpretation, standardized uptake value (SUV) 2.5, and 40% of maximum standard uptake value (SUV max ) on FDG-PET imaging. The length of tumors on PET scan were measured and recorded as Length vis , Length 2.5 , and Length 40 , respectively, and compared with the length of gross tumor in the resected specimen (Length gross ). All PET data were reviewed again postoperatively, and the GTV was delineated using various percentages of SUV max . The optimal-threshold SUV was generated when the length of PET matched the Length gross .The mean (±SD) Length gross was 5.48 ± 1.98 cm. The mean Length vis , Length 2.5 , and Length 40 were 5.18 ± 1.93 cm, 5.49 ± 1.79 cm, and 4.34 ± 1.54 cm, respectively. The mean Length vis (p = 0.123) and Length 2.5 (p = 0.957) were not significantly different from Length gross , and Length 2.5 seems more approximate to Length gross. The mean Length 40 was significantly shorter than Length gross (p < 0.001). The mean optimal threshold was 23.81% ± 11.29% for all tumors, and it was 19.78% ± 8.59%, 30.92% ± 12.28% for tumors ≥5 cm, and <5 cm, respectively (p = 0.009). The correlation coefficients of the optimal threshold were −0.802 and −0.561 with SUV max and Length gross , respectively.The optimal PET method to estimate the length of gross tumor varies with tumor length and SUV max ; an SUV cutoff of 2.5 provided the closest estimation in this study.