磁共振表观扩散系数成像在鉴别肺癌和肺良性病变中的应用

Application of magnetic resonance apparent diffusion coefficient imaging in differentiating lung cancer from benign lung lesions

目的:评估基于表观扩散系数的磁共振扩散加权成像不同模型在鉴别肺癌和肺良性病变中的应用效果。方法:这项回顾性单中心研究选择2020年1月―2021年6月49例患者为研究对象,包括肺癌组(n=32)和良性病变组(n=17)。所有患者在手术或活检前均接受了扩散加权成像(DWI)检查。ADC直方图参数,包括平均值、百分位值(第10位和第90位)、峰态和偏度,由两名放射科医生独立计算。比较肺癌患者和良性病变患者的直方图参数。构建接收者操作特征曲线以评估这些参数的诊断性能。结果:肺癌组中的ADC_(平均值)、ADC_(10th)、D_(平均值)和D_(10th)显著低于良性病变组,并且ADC_(偏度)和D_(偏度)显著高于良性病变组。D_(10th)和ADC_(10th)诊断良性病变能力最好。D_(10th)的曲线下面积(AUC)为0.851,在0.453×10^(-3)mm/s的最佳临界值下的准确率为79.6%;ADC_(10th)的AUC为0.814,在0.504×10^(-3)mm/s最佳临界值下的准确率为83.4%。结论:使用DWI单指数和双指数模型得出的ADC参数进行全病灶直方图分析可以区分肺癌与良性病灶。

Objective To evaluate the application effect of different MRI diffusion-weighted imaging models based on the apparent diffusion coefficient in distinguishing lung cancer from benign lung lesions.Methods This retrospective single-center study selected 49 patients from January 2020 to June 2021 as the study subjects,including the lung cancer group(n=32)and the benign disease group(n=17).All patients underwent diffusion-weighted imaging(DWI)before surgery or biopsy.ADC histogram parameters,including mean,percentile values(10th and 90th),kurtosis,and skewness,were calculated independently by two radiologists.The histogram parameters were compared between patients with lung cancer and benign lesions.ROC curves were constructed to evaluate the diagnostic performance.Results The ADC_(mean),ADC_(10th),D_(mean)and D_(10th)in the lung cancer group were significantly lower than the benign lesion group,and the ADC_(skewness)and D_(skewness)were significantly higher than the benign lesion group.D_(10th)and ADC_(10th)had the best ability to diagnose benign lesions.The area under the curve(AUC)of D_(10th)was 0.851,and the accuracy was 79.6%at the best critical value of 0.453×10^(-3)mm/s;the AUC of ADC_(10th)was 0.814,and the accuracy was 83.4%at the best critical value of 0.504×10^(-3)mm/s.Conclusions Whole-lesion histogram analysis of ADC parameters derived from the DWI single-index and double-index models can distinguish lung cancer from benign lesions.