From the Department of Radiology (D.B., P.S., J.P.R., P.K., K.Y., M.F., H.P.S.), Division of Medical Image Computing (S.K., M.G., N.G., K.H.M.H.), Division of Statistics (M.W.), and Department of Medical Physics (T.A.K., F.D.), German Cancer Research Center (DKFZ), Im Neuenheimer Feld 280, Heidelberg 69120, Germany; German Cancer Consortium (DKTK), Heidelberg, Germany (D.B., H.P.S., K.H.M.H.); and Departments of Urology (J.P.R., B.H., M.H., B.A.H.) and Neuroradiology (P.K.), University of Heidelberg Medical Center, Heidelberg, Germany.
Radiology. 2018 Oct;289(1):128-137. doi: 10.1148/radiol.2018173064. Epub 2018 Jul 31.
Purpose To compare biparametric contrast-free radiomic machine learning (RML), mean apparent diffusion coefficient (ADC), and radiologist assessment for characterization of prostate lesions detected during prospective MRI interpretation. Materials and Methods This single-institution study included 316 men (mean age ± standard deviation, 64.0 years ± 7.8) with an indication for MRI-transrectal US fusion biopsy between May 2015 and September 2016 (training cohort, 183 patients; test cohort, 133 patients). Lesions identified by prospective clinical readings were manually segmented for mean ADC and radiomics analysis. Global and zone-specific random forest RML and mean ADC models for classification of clinically significant prostate cancer (Gleason grade group ≥ 2) were developed on the training set and the fixed models tested on an independent test set. Clinical readings, mean ADC, and radiomics were compared by using the McNemar test and receiver operating characteristic (ROC) analysis. Results In the test set, radiologist interpretation had a per-lesion sensitivity of 88% (53 of 60) and specificity of 50% (79 of 159). Quantitative measurement of the mean ADC (cut-off 732 mm/sec) significantly reduced false-positive (FP) lesions from 80 to 60 (specificity 62% [99 of 159]) and false-negative (FN) lesions from seven to six (sensitivity 90% [54 of 60]) (P = .048). Radiologist interpretation had a per-patient sensitivity of 89% (40 of 45) and specificity of 43% (38 of 88). Quantitative measurement of the mean ADC reduced the number of patients with FP lesions from 50 to 43 (specificity 51% [45 of 88]) and the number of patients with FN lesions from five to three (sensitivity 93% [42 of 45]) (P = .496). Comparison of the area under the ROC curve (AUC) for the mean ADC (AUC = 0.84; AUC ≤ 0.87) vs the RML (AUC = 0.88, P = .176; AUC ≤ 0.89, P ≥ .493) showed no significantly different performance. Conclusion Quantitative measurement of the mean apparent diffusion coefficient (ADC) improved differentiation of benign versus malignant prostate lesions, compared with clinical assessment. Radiomic machine learning had comparable but not better performance than mean ADC assessment. © RSNA, 2018 Online supplemental material is available for this article.
目的 比较双参数对比剂-free 放射组学机器学习（RML）、平均表观扩散系数（ADC）和放射科医师评估在预测性 MRI 解读中检测到的前列腺病变特征描述中的作用。
材料与方法 本单中心研究纳入了 2015 年 5 月至 2016 年 9 月期间因 MRI-经直肠超声融合活检而有指征的 316 名男性（平均年龄±标准差，64.0 岁±7.8 岁）（训练队列 183 例患者；测试队列 133 例患者）。通过前瞻性临床阅读确定的病变通过手动分割进行平均 ADC 和放射组学分析。在训练集上开发用于分类临床显著前列腺癌（Gleason 分级组≥2）的全局和区域特定随机森林 RML 和平均 ADC 模型，并在独立测试集上测试固定模型。使用 McNemar 检验和受试者工作特征（ROC）分析比较临床阅读、平均 ADC 和放射组学。
结果 在测试集中，放射科医师的解读在每例病变的敏感性为 88%（60 例中的 53 例），特异性为 50%（159 例中的 79 例）。平均 ADC（截断值 732 mm/sec）的定量测量显著降低了假阳性（FP）病变的数量，从 80 例降至 60 例（特异性 62%[159 例中的 99 例]），并减少了假阴性（FN）病变的数量，从 7 例降至 6 例（敏感性 90%[60 例中的 54 例]）（P=.048）。放射科医师的解读在每例患者的敏感性为 89%（45 例中的 40 例），特异性为 43%（88 例中的 38 例）。平均 ADC 的定量测量减少了 FP 病变患者的数量，从 50 例降至 43 例（特异性 51%[88 例中的 45 例]），FN 病变患者的数量从 5 例降至 3 例（敏感性 93%[45 例中的 42 例]）（P=.496）。比较平均 ADC（AUC=0.84；AUC≤0.87）和 RML（AUC=0.88，P=.176；AUC≤0.89，P≥.493）的 ROC 曲线下面积（AUC）发现，定量测量平均 ADC 对良性与恶性前列腺病变的鉴别性能有所改善，优于临床评估。放射组学机器学习的表现与平均 ADC 评估相当，但没有更好。
结论 与临床评估相比，平均 ADC（表观扩散系数）的定量测量提高了良性与恶性前列腺病变的区分能力。放射组学机器学习的表现与平均 ADC 评估相当，但没有更好。
