CT影像组学预测T1期周围型非小细胞肺癌气腔播散

Prediction of CT-Based Radiomics in T1 Peripheral Non-Small Cell Lung Cancer via Spread Though Air Spaces

目的探究基于胸部CT的影像组学对T1期周围型肺癌气腔播散的预测价值。资料与方法回顾性收集2020年1月—2021年12月山西白求恩医院经手术病理确诊T1期非小细胞肺癌173例,根据有无气腔播散分为阳性组49例与阴性组124例,比较两组肺癌患者的一般资料及病灶CT形态学特征差异,并对两组病灶进行影像组学分析,以7∶3分为训练组122例与验证组51例,以肺癌原发区(病灶主体)、外周浸润区(沿病灶边缘向外扩大5mm环形区域)及肿瘤边缘区(沿病灶边缘向内回缩5 mm环形区域)为感兴趣区提取影像组学特征,建立肿瘤原发区模型、外周浸润区模型及肿瘤边缘区模型3个影像组学模型,并联合CT形态学特征,建立3个联合模型,评价各模型效能,筛选最优模型。结果肺癌气腔播散阳性组比阴性组中病灶分叶征更多见(χ^(2)=9.946,P=0.002)。基于3个感兴趣区建立的影像组学模型训练组曲线下面积(AUC)分别为0.899、0.825、0.840,验证组AUC分别为0.876、0.811、0.832,AUC最高的模型为肿瘤原发区影像组学模型(P=0.043、P<0.001、P=0.017),加入分叶征建立的联合模型AUC分别为0.917、0.835、0.851,验证组AUC分别为0.912、0.832、0.845,其中肿瘤原发区联合模型AUC最高(P<0.001、P=0.017、P=0.049)。结论利用影像组学探究肺癌气腔播散具有一定可行性,分叶征可作为肺癌气腔播散的危险预测因素。

Purpose To investigate the predictive value of chest CT-based radiomics for spread through air spaces in stage T1 peripheral type lung cancer.Materials and Methods A total of 173 patients with surgically pathologically confirmed stage T1 non-small cell lung cancer were retrospectively collected and divided into positive group(49 cases)and negative group(124 cases)according to the presence or absence of spread through air spaces.All lesions were randomly divided into training set(122 cases)and validation set(51 cases)according to the ratio of 7∶3.The primary area of lung cancer(the main body of the lesion),the peripheral infiltrative area(a 5-mm annular area expanding outward along the edge of the lesion)and the tumor margin area(a 5-mm annular area retracting inward along the edge of the lesion)were used as areas of interest to extract imaging histological features.Three imaging histological models were established for the primary area of lung cancer,the peripheral infiltrative area and the tumor margin area,and combined with the morphological features of CT to establish three combined models.The efficacy of each model was evaluated and the optimal model was selected.Results The lobulation signs of positive group was significantly more than that of negative group(χ^(2)=9.946,P=0.002).The area under the curve(AUC)of the imaging histological model based on the three regions of interest were 0.899,0.825,0.840 for the training group and 0.876,0.811 and 0.832 for the validation group,respectively.The model with the highest AUC was the primary tumor imaging model(P=0.043,P<0.001,P=0.017),the AUC of the combined model established by adding the lobar sign were 0.917,0.835 and 0.851,respectively.The AUC of the three regions of interest in the validation group were 0.912,0.832,and 0.845 and the highest AUC was found in the primary tumor area(P<0.001,P=0.017,P=0.049).Conclusion It is feasible to study lung cancer with airway metastasis via CT-based radiomics,taken lobulation signs as the risk predictive factor.