基于全病灶的动态增强MRI强度直方图鉴别肺部炎性结节与肺癌

Differentiating pulmonary inflammatory nodules from lung cancer based on whole-focus dynamic enhanced MRI intensity histogram

目的应用动态增强磁共振成像(dynamic contrast-enhancement magnetic resonance imaging,DCE-MRI)强度直方图分析,对肺部直径0.8~3.0 cm的炎性结节与肺癌进行全病灶分析并评估其对二者的鉴别诊断价值。材料与方法回顾性分析2019年7月至2022年6月共123例经手术或穿刺活检病理以及临床影像随访证实的肺部炎性结节及肺癌患者的DCE-MRI资料,其中肺部炎性结节63例、肺癌60例。使用FireVoxel软件在病灶峰值增强期图像及其剪影图像上逐层手动勾画全病灶感兴趣区(region of interest,ROI),得到3D ROI的信号强度直方图参数,包括最小值、最大值、平均值、中位数、标准差、偏度、峰度、熵值、变异系数(coefficient of variation,CoV)、第10百分位数(P10)、第25百分位数(P25)、第50百分位数(P50)、第75百分位数(P75)、第90百分位数(P90)等,进行组间比较,利用受试者工作特征(receiver operating characteristic,ROC)曲线确定强度直方图参数对于二者的诊断效能。应用logistic回归分析模型得到联合变量(joint variable,JV),利用ROC曲线来确定其诊断效能。结果峰值增强期图像强度直方图参数中,肺癌的最小值、平均值、中位数、P10及P25均高于炎性结节,而CoV、偏度低于炎性结节,且差异均具有统计学意义(P<0.05)。其中以最小值155.5为阈值的AUC[0.668,95%置信区间(confidence interval,CI):0.573~0.764]最大,诊断效能最佳,敏感度与特异度分别为35.0%和93.7%。峰值增强期剪影图像强度直方图参数中,肺癌的最小值、P10及P25均高于炎性结节,而CoV、熵值低于炎性结节,且差异均具有统计学意义(P<0.05)。其中以CoV 0.275为阈值的AUC(0.775,95%CI:0.692~0.858)最大,诊断效能最佳,敏感度与特异度分别为88.9%和58.3%。分析时间-信号强度曲线(time-signal intensity curve,TIC)所衍生的半定量参数包括达峰时间(time to peak,TTP)、对比增强比及曲线斜率,肺癌的TTP比炎性结节短,而曲线斜率大于炎性结节,且差异均具有统计学意义(P<0.05)。TTP以204.2 s为阈值的AUC(0.737,95%CI:0.647~0.828)最大,敏感度与特异度分别为58.7%和85.0%;曲线斜率以1.76为阈值的AUC(0.732,95%CI:0.641~0.822)最大,敏感度与特异度分别为61.7%和82.5%。联合剪影图像的直方图参数与半定量参数得到JV以0.43为阈值的AUC(0.885,95%CI:0.823~0.947)最大,敏感度与特异度分别为88.3%和79.4%。结论基于全病灶的DCE-MRI强度直方图可以为肺部炎性结节与肺癌的鉴别诊断提供信息,且剪影图像的诊断效能更高,联合应用直方图参数与TIC衍生的半定量参数可进一步提高对二者的鉴别能力,为鉴别诊断提供可靠的客观依据。

Objective:Dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI)intensity histogram analysis was used to analyze the whole focus of pulmonary inflammatory nodules and lung cancer which diameter 0.8-3.0 cm and to evaluate their value in differential diagnosis.Materials and Methods:The DCE-MRI data of 123 patients with pulmonary inflammatory nodules and lung cancer confirmed by operation or puncture biopsy and clinical imaging follow-up from July 2019 to June 2022 were analyzed retrospectively,including 63 cases of pulmonary inflammatory nodules and 60 cases of lung cancer.Using FireVoxel software,the region of interest(ROI)of the whole lesion was manually delineated layer by layer on the peak enhancement image and its silhouette image,and the signal intensity histogram parameters of 3D ROI were obtained.Including minimum,maximum,average,median,standard deviation,skewness,kurtosis,entropy,coefficient of variation(CoV),10th percentile(P10),25th percentile(P25),50th percentile(P50),75th percentile(P75),90th percentile(P90),etc.The diagnostic ability of intensity histogram parameters for both groups was determined by using thereceiver operating characteristic(ROC)curve.The joint variable(JV)was obtained by logistic regression analysis model,and the diagnostic ability was determined by ROC curve.Results:In the histogram parameters of peak enhancement,the minimum,mean,median,P10 and P25 of lung cancer were higher than those of inflammatory nodules,while the CoV and skewness of lung cancer were lower than those of inflammatory nodules,and the difference was statistically significant.Among them,the AUC[0.668,95%confidence interval(CI):0.573-0.764]with the minimum value of 155.5 as the threshold is the largest,the diagnostic efficiency is the best,and the sensitivity and specificity are 35.0%and 93.7%.In the histogram parameters of peak enhancement,the minimum,P10 and P25 of lung cancer were higher than those of inflammatory nodules,while the values of CoV and entropy were lower than those of inflammatory nodules,and the difference was statistically significant.Among them,AUC(0.775,95%CI:0.692-0.858)with CoV 0.275 as the threshold was the largest,and the diagnostic efficiency was the best,with a sensitivity and specificity of 88.9%and 58.3%.The semi-quantitative parameters derived from time-signal intensity curve(TIC)included time to peak(TTP),contrast enhancement ratio and curve slope.The TTP of lung cancer was shorter than that of inflammatory nodules,but the slope of curve was larger than that of inflammatory nodules,and the difference was statistically significant.The AUC(0.737,95%CI:0.647-0.828)of TTP with a threshold of 204.2 seconds was the largest,with a sensitivity and specificity of 58.7%and 85.0%,and AUC(0.732,95%CI:0.641-0.822)with a slope of 1.76 was the largest,with a sensitivity and specificity of 61.7%and 82.5%.The combination of histogram parameters and semi-quantitative parameters of the silhouette showed that AUC(0.885,95%CI:0.823-0.947)with a threshold of JV of 0.43 was the highest,with a sensitivity and specificity of 88.3%and 79.4%.Conclusions:DCE-MRI intensity histogram based on whole focus can provide information for differential diagnosis of lung cancer and inflammatory nodules,and the diagnostic efficiency of silhouette images is higher.The combined application of histogram parameters and semi-quantitative parameters derived from TIC can further improve the ability to distinguish between malignant nodules and inflammatory nodules,and provide reliable objective basis for differential diagnosis.