| Purposes:To prospectively evaluate whether the accuracy of target colume delineation could be increased by integrating 18FDG PET/CT into radiotherapy planning ofeasophageal carcinoma.Materials/Methods:From May 2007 to Jan 2008,16 patients with esophageal squamous cell carcinoma underwent routine examinations including contrast-enhanced chest CT,esophagogram and endoscopy,as well as 18FDG PET/CT before operations.GTVcTwas delineated on CT images blinded to PET results.Four GTVs were determined by PET/CT,using SUVbgd+20%(SUVmax(slice)-SUVbgd), SUVbgd+40%(SUVmax(slice)-SUVbgd),2.5 and 40%SUVmax(total) as the thresholds,and called GTV20,GTVP40,GTV2.5 and GTV40% respectively(SUVbgd,average SUV of the background;SUVmax(slice),maximum SUV of every slice of the lesion;SUVmax(total), maximum SUV of the total lesion).Then all patients underwent radical surgery and three-field lymph node(LN) dissection,with titanic clips marked at both ends of the tumor beds,and CT scan again two weeks later.LN groups were recorded according to the LN map for esophageal cancer from the Dpt.of thoracic surgery.GTVsurg was contoured on fusion images of pre- and post-surgical CT,according to ranges of markers on post-surgical CT longitudinally and extents of lesions on pre-surgical CT transversely.The lengths of 6 GTVs were recorded as LCT,L20,L40,L2.5,L40%and Lsurg respectively.Impact on staging:To evaluate the diagnositic value for T,N(to draw a ROC curve of diagnosing LN group metastasis with different SUV thresholds and to seek for the optimal threshold[SUVthre]) and M.Impact on GTV delineation for primary tumor:The former 5 lengths or corresponding GTVs were compared with Lsurg or GTVsurg by means of a conformity index(CI,square of the intersection of the two lengths or GTVs divided by their product),e.g.CI'CT&surg=L2CT&surg/(LCT*Lsurg) and CICT&surg=GTV2CT&surg/(GTVCT* GTVsurg).Impact on GTVN delibeation for lymph node:Three LN GTVs were defined, one based on CT alone(GTVNcT),one based on PET/CT with SUVthre(GTVNPET) and another based on pathologic results(GTVNsurg,delineated on presurgical CT). GTVNCT and GTVNPET were compared with GTVNsurg by means of a conformity index(CI,the intersection of the two GTVNs divided by their union),e.g.CICT&surg= GTVNCT&surg/(GTVNCT+ GTVNsurg- GTVNCT&surg).Stata7.0 software was used.McNemar chi-square test with exact method was used to compare paired categorical datas,and paired t-test for paired measurement datas.Randomized block designed datas were analyzed by two-way analysis of variance(ANOVA),further compared with each other if the difference was significant, with typeâ… error revised by Bonferronni method.If the condition of parametric test was not satified,nonparametric one would be alternative.All tests are two-sided,and a value of P＜0.05 was considered significant.Results:Of 16 patients,2 had pT1 tumors,6 pT2,5 pT3 and 3 pT4,10 pN0 and 6 pN1,10 pM0,0 pM1a and 6 pM1b.A total of 144 LN groups(452 LNs) were dissected and 21groups(33 LNs) proved to be malignant at pathologic examination.Impact on staging:â‘ PET/CT detected all primary tumors,and maximum SUVs showed no significant correlation to the lengths,X-ray types,differentiation degrees or pTs.â‘¡The area under curve(AUC) of the ROC curve was 0.9017 and SUVthre was 2.36.For depiction of LN group metastasis,the sensitivity,specificity and accuracy of PET/CT with 2.36 were 76.19%,95.93%and 93.06%,whereas those of CT were 33.33%,94.31%and 85.42%(P values:0.0117,0.7539 and 0.0266 respectively).No Mla was observed in our study.The sensitivity,specificity and accuracy of PET/CT for N1 and Mlb staging were 100.00%,90.00%,93.75%and 50.00%,80.00%,68.75%,and those for diagnosing N1 LN groups were 92.31%, 95.38%,94.87%.â‘¢In our study,no more distant visceral metastasis was detected by PET/CT than by conventional examinations,except for Mlb.Impact on GTV delineation for primary tumor:The mean LCT,L20,L40,L2.5,L40% and Lsurg were 6.30±2.69,5.55±2.48,6.80±2.92,6.65±2.66,4.88±1.99 and 5.90±2.38cm respectively.The mean CI'CT&surg,CI'20&surg,CI'40&surg,CI'2.5&surg and CI'40 %&srug were 0.68±0.16,0.84±0.17,0.76±0.14,0.78±0.15 and 0.80±0.11 respectively. CI'20&surg and CI'40&surg was significantly superior to CI'CT&surg(P＜0.05) and no difference was observed between each other of the else.The mean GTVCT,GTV20, GTV40,GTV2.5,GTV40% and GTVsurg were 29.16±18.56,18.75±12.37,12.52±8.08, 22.69±14.84,9.18±5.96 and 28.16±17.02cm3 respectively.The mean CIs of GTVs increased significantly from CI40&surg(0.27±0.09) and CI40%surg(0.28±0.08)＜CI20&surg(0.52±0.16) and CI2.5&surg(0.52±0.20)＜CICT&surg(0.77±0.17).The median percentage volumes of GTV20,GTV40,GTV2.5 not included in GTVsurg were 8.8%, 18.6%and 16.3%respectively and almost 0 for GTV40%.Impact on GTVN delibeation for lymph node:The mean GTVNcT,GTVNPET and GTVNsurg were 1.52±2.38,2.82±4.51 and 2.68±4.16cm3,respectively.The mean CICT&surg and CIPET&surg were 0.31 and 0.65(P value= 0.0352).Conclusions:1) Despite of the high qualitative value for primary tumor,it is still difficult for PET/CT to differentiate T1-T4.High qualitative and quantitive values for N1 and poor sensitivity for Mlb were observed.In our study,no more distant visceral metastasis was detected by PET/CT than by conventional examinations, except for M1b.2) Considered both longitudinally and transversely,GTV20 may be most conformed to GTVsurg,especially when T/B≥9.Because gold standard of transverse information of primary lesions was not available,it is recommended that GTV20 and the transverse contour of CT should be referred to complementarily when delineating GTV for esophageal cancer 3) The optimal SUV threshold of diagnosing LN group metastasis in esophageal cancer may be 2.36,at which the sensitivity and accuracy for PET/CT turned out to be superior to those for CT.Comparing to LN GTV based on CT,the one based on PET/CT was much better conformed with the real range of malignant LNs.4) But further large-sample research with information of treatment failure should be needed. |
PDF Full Text Request