以與病理診斷關聯之指標，建立口腔癌患者以[F-18]FDG正子電腦斷層掃描診斷淋巴結轉移之診斷標準

目的： [氟-18] 去氧葡萄糖 ([F-18]Fluoro-deoxyglucose, FDG)正子放射電腦斷層掃描(Positron Emission Tomography/Computed Tomography，PET/CT）影像應用於口腔癌分期時，一直缺少具共識的判斷淋巴結轉移與否的診斷標準，與FDG非為腫瘤特異性有直接相關。本研究旨在藉由回溯性的收集口腔癌患者術前FDG PET/CT影像上有關參數，與病理報告比對，找出提供最高之診斷正確性的指標組合與其取捨值(或切點cutoff)，建立一診斷淋巴結轉移之客觀標準以供臨床實務上依循。材料與方法： 回溯性分析151位口腔鱗狀細胞癌患者之術前FDG PET/CT影像，紀錄原發腫瘤大小與SUVmax，頸部淋巴結的短軸長(NSAL)、SUVmax (NSUV)與其滯留指數(retention index, SUVRI) [1]，與PET/CT的正式報告，以及手術的病理報告比對。利用混淆矩陣求得核醫專科醫師PET/CT報告的正確性。利用獨立樣本t檢定找出在淋巴結轉移陽性與陰性兩組之間有顯著差距之參數；利用接收者操作特徵曲線（receiver operating characteristic curve, ROC curve）下最大面積(area under curve, AUC) 找出各參數之最佳敏感度及特異性與其對應之取捨值(cutoff)。其次依單一參數取捨值之相近概數，利用混淆矩陣，求得各兩參數組合之最佳敏感度及特異性。以SPSS 18版進行上述資料分析。結果： 核醫專科醫師以目視經驗法判讀FDG PET/CT報告病人別的敏感度為69.0%，特異性為61.1%，準確度為62.9%；區域別的敏感度為64.5%，特異性為58.3%，準確度為59.7%。162個淋巴結區域的上述參數被記錄。原發腫瘤大小與SUVmax、頸部淋巴結的短軸長、與SUVmax在淋巴結轉移陽性與陰性兩組之間有顯著差距，但SUVRI沒有。單一參數中NSAL>6mm診斷淋巴結是否轉移效力最高，其次是NSUV>5。參數組合中以NSUV> 5 且NSAL> 6mm診斷正確性最高，準確度為78.5%。NSUV>4 或NSAL> 6mm則有最高的敏感度。結論： 口腔癌患者FDG PET影像上頸部淋巴結，單一淋巴區域者若SUVmax > 4或短軸長 > 6mm；多區域者若SUVmax > 5且短軸長> 6mm時，臨床懷疑轉移陽性。

Purpose: When [F-18]Fluorodeoxyglucose (FDG) positron emission tomography (PET/CT) imaging is applied to stage oral cancer, there has been a lack of consensus to differentiate possible lymph node metastasis. The purpose of this study is to retrospectively collect relevant parameters on pre-operative FDG PET/CT images of oral cancer patients, compare them with pathology reports, and find the combination of indicators with cutoffs that provide the highest diagnostic accuracy. The resultant objective criteria could be referred to in clinical practice.Material and method： 151 patients of oral squamous cell carcinoma who were diagnosed during 2018.01 to 2021.01 were enrolled. Total FDG PET/CT images were retrospectively reviewed to collect sizes & SUVmax of primary tumors, as well as short axis lengths and SUVmax & retention index of SUVmax of visible lymph nodes. With correlating with formal interpretation reports by two board-certified nuclear medicine physicians and pathological reports, the diagnostic accuracy of interpretation reports were generated by confusion matrix. The parameters which differed significantly between pN(+) and pN(-) were identified by independent t test. The cutoff of each single parameter that generated the highest accuracy was identified by ROC curve with the largest AUC. Two parameters with approximated number were used to generate the diagnostic accuracy. SPSS v18 was used for all the above-mentioned statistical analysis.Result: The sensitivity and specificity of visual interpretation by 2 nuclear medicine physicians was 69.0% and 61.1 respectively for patient-based analysis, 64.5% and 58.3% respectively for level-based analysis. Lymph nodes of total 162 levels were recorded and analyzed. Except SUVRI, the difference of sizes and SUVmax of primary tumor as well as short axis length and SUVmax of lymph nodes were statistically significant between pN(+) and pN(-) groups. For single parameter, NSAL>6mm had the highest diagnostic efficacy for lymph node metastasis, followed by NSUV>5. For combinations, NSUV>5 and NSAL>6 mm seemed to have the highest diagnostic efficacy, while NSUV>4 or NSAL>6 mm had the highest sensitivity. In conclusion, NSUV>4 or NSAL>6 mm for a single level lymph node, while NSUV>5 & NSAL>6 mm for lymph nodes in multiple levels may be submitted for metastasis positive on FDG PET/CT scan.