目的：建立一种多参数乳腺MRI检查与诊断方式，与美国放射学院乳腺影像报告和数据系统（BI-RADS）分类对应，改进乳腺疾病的处理建议。资料与方法回顾分析278例乳腺疾病患者301个经病理证实的病灶，使用1 mm×1 mm×1 mm空间分辨率、120 s时间分辨率的动态增强扫描（DCE）序列和b=1000 s/mm2的扩散加权成像（DWI）序列进行扫描，将DCE显示早期病灶形态学恶性征象、时间-信号强度曲线（TIC）II型或III型、小于良恶性表观扩散系数（ADC）阈值的3个诊断因素各计1分，肿块和非肿块样强化病灶区别对待，积分≥2分诊断为BI-RADS 5类，积分=1分诊断为BI-RADS 4类，积分<1分诊断为BI-RADS 3类，其他特异性良性发现诊断为BI-RADS 2类，DCE和DWI无异常发现评价为BI-RADS 1类，并与病理学的良性（B）-高危（HR）-恶性（M）病灶分级进行对照，评价其对病灶处理的建议。结果以HR作为恶性时（M+HR），得到的ROC曲线下面积为0.860；以HR作为良性时（B+HR），得到的ROC曲线下面积为0.876，两者很接近。经过ROC曲线优化，在病理上将HR作为良性、在MRI上将BI-RADS 5类作为恶性，获得敏感度为85.3%，特异度为86.8%，准确度为85.1%，高于其他组合。如果将病理上HR病灶的处理原则定义为局部切除或短期随访，则BI-RADS 5类对M+HR病灶（可切除病灶）阳性预测值为93.2%；BI-RADS 4类病灶对M+HR病灶的阳性预测值为46.9%，必须活检以决定局部切除或短期随访；BI-RADS 3类及以下对B+HR病灶的阳性预测值（随访观察）为90.4%。结论本研究建立了一个简单的诊断模型，动态增强显示的形态学特征、动态时间-信号强度曲线和DWI-ADC值取相同的权重进行BI-RADS分类，可以很好地预测乳腺病灶良性、高危和恶性特征，对指导乳腺疾病的处理方式有实用价值。%PurposeTo investigate a multi-parametric protocol for breast MRI examination and lesions assessment correlated to the American College of Radiology (ACR) breast imaging reporting and data system (BI-RADS) categorization, and to improve the management of the breast lesions.Materials and Methods 301 pathologically confirmed lesions on 278 patients were retrospectively included. The scan protocol used a dynamic contrast enhancement sequence (DCE) of 1 mm×1 mm×1 mm spatial resolution, 120 temporal resolution and a diffusion weighted imaging (DWI) of b=1000 s/mm2. The malignant morphological features on the early-enhanced images, type II or III time intensity curve and the apparent diffusion coefficient (ADC) value less than benign/malignant threshold was equally weighted. Each was given 1 point when present malignant features and treated different on mass and non-mass-like enhancement lesions. When the sum of score was ≥2 points, the lesion was categorized as BI-RADS 5. When the sum of score was 1 point, the lesion was categorized as BI-RADS 4. When the sum of score was <1 point, the lesion was categorized as BI-RADS 3. The other specific benign findings were categorized as BI-RADS 2. No abnormality on DWI, DCE, T2WI and T1WI was categorized as BI-RADS 1. The final categories were correlated to the pathological grades as benign (B), high risk (HR) and malignant (M).Results When grouped HR as malignant (M+HR), the area under curve (AUC) of the ROC was 0.860. When grouped HR as benign (B+HR), the AUC of the ROC was 0.876, and the optimized sensitivity, specificity and accuracy was 85.3%, 86.8% and 85.1%, respectively, which were better than the other grouping. If the management of HR lesions could be lumptoectomy or short-term follow-up, the positive predictive value (PPV) of BI-RADS 5 for excisable lesions (M+HR) was 93.2%, the PPV of BI-RADS 4 for excisable lesions (M+HR) was 46.9% and the biopsy was essential. The PPV of BI-RADS 3 and below for follow-up lesions (B+HR) was 90.4%.Conclusion A simple diagnosis algorithm was established, which equally weighted the DCE morphological feature, DCE-TIC and DWI-ADC. The diagnosis protocol was well consistent with BI-RADS categorization and could predict the benign, high risk and malignant lesions in pathology as well as the proper management.
展开▼
