以全腫瘤體積表觀擴散係數直方圖參數預測頭頸癌患者之誘導化療效果

研究目的：頭頸部鱗狀上皮細胞癌(Head and Neck Squamous Cell Carcinoma, HNSCC)是世界上最普遍的癌症之一，其危險因數包括酒精，煙草和檳榔等。儘管誘導化療對HNSCC患者的作用及效果仍處於研究階段，但在一項綜合分析研究中表明，適當時機對患者進行化療可以提高患者的存活率。近年來，已有研究證明通過使用擴散加權影像(Diffusion Weighted Imaging, DWI)得到的表觀擴散係數值(Apparent Diffusion Coefficient, ADC)可以反應出腫瘤組織內部的微結構狀態，並可作為一項潛在生物標誌物評估HNSCC患者的治療反應。然而先前一些研究表明，可能由於常規目標區域（ROI）方法用於表徵腫瘤異質性的能力有限，治療後使用平均ADC值區分有治療反應者與無反應者的效果甚微。除此之外，利用常規回訊平面造影(Echo Planar Imaging, EPI)取得的影像品質通常較差，特別是在空氣-組織交界面處產生磁化率假影而造成的幾何扭曲。因此，本研究的目的是利用讀出方向分段採樣回訊平面造影(Readout-Segmented EPI)評估全腫瘤體積表觀擴散係數的直方圖參數用於預測頭頸部鱗狀上皮細胞癌患者誘導化療效果的能力。材料與方法：本研究總共納入19位晚期頭頸部鱗狀上皮細胞癌患者，醫師依據固體腫瘤反應評估標準(RECIST criteria)評估患者治療效果，其中15位有治療效果歸為有效組，4位無治療效果歸為無效組。所有患者分別於誘導化療的第一個療程前後於3.0T的磁振造影掃描儀接受檢查，所有的影像在經過精準的影像對位之後通過DWI上所得的訊號強度計算出相應的ADC map，參照T2加權影像 (T2 Weighted Image, T2WI)對手動圈選的全腫瘤體積進行直方圖分析，直方圖參數包括ADC平均值(。並計算出各直方圖參數對應的誘導化療前後變化率和腫瘤體積變化率。最後通過無母數統計中的曼-惠特尼U檢定(Mann–Whitney U test)比較兩個群組的直方圖參數變化率，並用接受者操作特徵曲線(receiver operating characteristic, ROC)取得的最佳截斷值。結果：對兩組患者直方圖參數變化率進行曼-惠特尼U檢定的結果表明：有效組的ΔADCmin顯著高於無效組，分別為1.78±2.98和-0.73±0.36，p < 0.01；有效組的ΔADC25值亦顯著高於無效組，分別為0.19±0.23和-0.09±0.15，p < 0.05。接受者操作特徵曲線的結果表明ΔADCmin和ΔADC25值在預測患者最終治療反應上皆具有良好表現，當ΔADCmin取-0.183時，可得到最佳鑒別閾值，靈敏度為80.0%，特異度為100%，曲線下面積值(Area Under Curve, AUC)為0.933；當ΔADC25取0.026時，可得到最佳鑒別閾值，靈敏度和特異度為73.3%和100%，AUC值為0.850。討論：本研究初步證明了使用ADC直方圖分析預測HNSCC患者對誘導化療反應的可行性。在直方圖參數中，ΔADCmin和ΔADC25值可以在區分有效組和無效組方面提供更高的診斷效能，這可能是由於晚期腫瘤細胞中異常增生最活躍的部分細胞密度更高，表現為水分子擴散受限更為顯著，因此更能體現腫瘤細胞內真實增殖及分化狀況。另外，兩組之間ΔADCmean和ΔVtumor值無顯著差異也說明了常規使用ADCmean值進行比較可能存在局限性。 綜上，本研究表明了表觀擴散係數的全腫瘤體積直方圖分析的重要作用。ΔADCmin和ΔADC25值有希望成為早期有效預測HNSCC患者對誘導化療反應的潛在生物標誌物，有助於醫師確定進一步的治療策略。

Introduction: Head and neck squamous cell carcinoma (HNSCC), one of the most common cancers worldwide, could be associated with increasing consumption of alcohol, tobacco or betel quid. While the significance of induction chemotherapy to patients with HNSCC remains under investigation, a meta-analysis study indicated that adding chemotherapy to the patients at an appropriate timing could improve their survival. Recently, quantified apparent diffusion coefficient (ADC) value derived using diffusion weighted imaging (DWI) protocol has been demonstrated to reveal information on the tumoral microstructure, as a potential biomarker for response assessment in HNSCC patients. However, some previous studies indicated poor differentiation of responders from non-responders after the treatment using mean ADC value, probably due to limited capability of characterizing tumor heterogeneity using a conventional region of interest (ROI) approach. In addition, poor image quality, such as geometric distortion at air-tissue interfaces, was frequently obtained in the conventional EPI images. As a result, the purpose of this study is to assess the ability of whole-volume histogram analysis of water diffusion to predict response to induction chemotherapy in patients with HNSCC using REadout Segmentation Of Long Variable Echo trains (RESOLVE) DWI.Methods: A total of 19 patients with advanced head and neck squamous cell carcinoma were enrolled in this study. MRI examinations were performed on the patients before and after the first cycle of induction chemotherapy at a 3.0T MR scanner (Magnetom Skyra, Siemens Healthcare, Erlangen, Germany) with a 20-channel head and neck coil. In addition to other conventional MR imaging protocols, a multishot readout-segmented EPI was acquired with: TR/TE = 5800/63 ms , FOV = 230x230mm², matrix size = 160x160 (zero-filled to 320x320), slice thickness = 5 mm, number of slices = 32 ,iPAT = 2 , bandwidth = 919Hz/Px , b value = 0, 800s/mm².All images in our study have been accurately registered, meanwhile, ADC maps were calculated by measuring signal intensities in corresponding DWI. Manual ROIs covering the whole tumor were performed, excluding necrotic and cystic portions, on DWI by a radiologist with references of T2-weighted images. Afterwards, the mean, median, 25th percentile (ADC25), 75th percentile (ADC75), minimum, maximum, c within the whole volume were derived using histogram analysis. And delta (Δ) change ratios of the above indices were generated by calculating the difference of the two examinations and then divided by the prior [e.g. ΔADC= (posttreatment ADC−pretreatment ADC)/pretreatment ADC]. Moreover, patients were stratified as responders (n = 15) and non-responders (n = 4) according to the clinical outcome. Finally, a nonparametric Mann–Whitney U test was used to compare the differences of the extracted histogram features between these two groups, and a receiver operating characteristic (ROC) curve analysis was performed to obtain the optimal cut-off value for the response assessment.Results: Figure 1 and 2 are two representative cases of responders and non-responders, including the T2-weighted image (A), DWI image (B) ADC map (C), and histogram (D), respectively Results of the Mann–Whitney U test of the histogram features between the two groups are shown in Figure 3. ΔADCmin values of the responders were significantly higher than non-responders (1.78±2.98 versus -0.73±0.36; p < 0.01); ΔADC25 values of the responders were significantly higher than non-responders (0.19±0.23 versus -0.09±0.15; p < 0.05). Furthermore, results of ROC analysis are shown in Figure 4, and found that an optimal cutoff value for ΔADCmin of -0.183 effectively differentiated responders and non-responders with 80.0% sensitivity and 100% specificity (AUC=0.933), an optimal cutoff value for ΔADC25 of 0.026 with 73.3% sensitivity and 100% specificity (AUC=0.850). Discussion: Our preliminary study demonstrated the feasibility of using ADC histogram analysis to predict response to induction chemotherapy in patients with HNSCC. Among the extracted features, ΔADCmin and ΔADC25 values could provide better opportunities in differentiating responders from non-responders after induction chemotherapy, probably because of an elevated water diffusion resulting from lower cellularity in the responsive tumorous regions. In addition, no significant differences were found in the delta change ratios of mean ADC and tumor volume between two groups, which emphasizes the possible limitation of using conventional comparisons of the mean ADC values. In conclusion, this study indicates the important role of whole volume histogram analysis of water diffusion. The ΔADCmin and ΔADC25 values, are potential biomarkers for predicting early response to induction chemotherapy in patients with HNSCC, which could facilitate in further determination of treatment strategy.