乳房腫瘤在磁振造影擴散係數上之特性分析:擴散峰度影像與內體素不相干運動之比較

Abstract

背景 乳癌乃本國女性新診斷癌症之首，過去文獻指出乳房磁振造影對於協助診斷乳房病灶擁有較乳房攝影更高的敏感度，然而其特異性與對於乳房病灶之診斷效度仍有進步空間。 目的 本研究目的希望透過分析多項新穎磁振造影參數以探討並提升乳房病灶辨別之準確性。本研究透過探討磁振造影之 1)擴散係數影像 (Diffusion-weighted Imaging, DWI)、2) 擴散峰度影像 (Diffusion Kurtosis Imaging, DKI ) 、 3) 內體素不相干運動(Intravoxel Incoherent motion, IVIM) 、 4) 動態對比增強磁振造影(Dynamic Contrast-enhanced MRI; DCE-MRI)等四項技術所衍生之參數，與比較各參數之差異性以尋求辨別乳房病灶之最佳影像工具。 研究方法 本研究為前瞻性研究，研究對象為臨床乳房超音波發現可疑乳房腫瘤且須切片之病患，於病患接受切片前執行磁振造影，單一次造影過程中收集DWI、DKI、IVIM及DCE-MRI等影像，並計算其定量參數。各參數在良惡腫瘤之間的統計差異性使用Student t-test檢定，而各參數辨別乳房病灶良惡性之效度則以接收者操作特徵曲線(Receiver Operating Characteristic curves, ROC curves)作為評量基準。 結果與討論 本研究總共分析來自於57名病人之61個乳房病灶(22個良性病灶與39個惡性腫瘤)。於各類別參數中，DCE-MRI的kep, ve、DWI之ADC、DKI之D與K、以及IVIM之D*等指標，均顯示在良惡性乳房腫瘤間有顯著差異（p<0.05）。其中從DKI所得之參數Ｄ（平均擴散係數）為最準確的診斷指標(AUC=0.957)，同時特異性高達100%。DCE-MRI之Ktrans與IVIM之f則在良惡性乳房腫瘤之間沒有顯著差異。 結論 本研究採用一合併新穎擴散磁振成像技術與動態對比增強磁振造影之模型作為分析乳房腫瘤之工具，並比較各種定量參數之診斷效度。結果發現由擴散峰度影像與內體素不相干運動所計算出之定量參數診斷效度優於動態對比增強造影之定量參數。

Abstract Background Breast cancer is the most commonly diagnosed cancer among women in Taiwan. Breast magnetic resonance imaging (MRI) is a sensitive imaging technique to assess breast cancer but its effectiveness still remains to be improved. Objectives Our study aim was to assess MRI diagnostic accuracy for breast lesions by comparing quantitative parameters derived from 1) diffusion-weighted imaging (DWI), 2) diffusion kurtosis imaging (DKI), 3) intravoxel incoherent motion (IVIM) and 4) dynamic contrast-enhanced MRI (DCE-MRI) and to explore an optimal model for breast tumor differentiation. Methods This was a prospective study performed on patients with suspicious breast lesions found on breast sonography prior to biopsy. All MR experiments were conducted on a 3-T MRI scanner. Quantitative parameters from DWI (apparent diffusion coefficient, ADC; fractional anisotropy, FA), DKI (mean diffusivity, D; mean kurtosis, K), IVIM (pseudo-diffusion coefficient, D*; perfusion fraction, f) and DCE-MRI (Ktrans, kep, ve and vp) were derived and compared between malignant and benign lesions. These parameters in benign and malignant lesions were compared by GEE. (Generalized estimating equations). Area under the receiver-operating characteristic (ROC) curve (AUC) analysis was used to evaluate diagnostic accuracy among parameters. Results and Discussion Total 61 suspicious breast lesions from 57 patients were evaluated in this prospective study. The quantitative parameters, including kep and ve from DCE-MRI, ADC, D, DSTD, DMINIMUM, K, KSTD and KMINIMUM from DKI between benign and malignant lesions were significantly different (p<0.05). DMINIMUM derived from DKI demonstrated the largest AUC (0.93) and had the highest specificity (95.45%). The Ktrans from DCE-MRI and those parameters derived from IVIM were shown to be not significantly different between benign and malignant lesions. Conclusion In this study, we investigated the effectiveness of differentiating malignant breast tumors from benign ones by using a DKI-IVIM approach. The quantitative parameters derived from DKI and IVIM have been proved to be more accurate than perfusion parameters derived from DCE-MRI and conventional ADC. Key Words: MRI, breast tumors, breast cancer, DCE, DKI, IVIM, diffusion, kurtosis, intravoxel incoherent motion