基于高分辨CT特征及AI定量参数预测Ⅰ期浸润性肺腺癌病理侵袭性的价值

The value of predicting the pathological aggressiveness of stage I invasive lung adenocarcinoma based on high�resolution CT features and AI quantitative parameters

目的探讨高分辨CT特征及人工智能(AI)定量参数预测Ⅰ期浸润性肺腺癌的病理侵袭性的价值。方法选取95例术后经病理证实为Ⅰ期浸润性肺腺癌患者的影像学资料,将符合脉管侵犯、脏层胸膜侵犯、区域淋巴结转移、肺泡腔内播散中的一种或多种病理征象定义为侵袭组(46例),反之为非侵袭组(49例);使用单因素及二元Logis‐tic回归分析对CT征象及AI定量参数进行分析。结果结节类型、毛刺征、胸膜牵拉征、结节大小、体积、质量、实性占比、最小CT值、平均CT值、中位数、标准差、偏度、峰度等组间比较均差异具有统计学意义(P<0.05);最小CT值、平均CT值、体积实性占比、结节类型及胸膜牵拉征为侵袭性肺腺癌的独立危险因素,且平均CT值诊断效能最高;侵袭组主要表现为实性结节,胸膜牵拉征,平均CT值高于非侵袭组,差异有统计学意义(P<0.05)。结论CT特征及AI定量参数可预测肺腺癌的病理侵袭性,为患者术后治疗方案提供帮助。

Objective To explore the pathological invasiveness of stage I invasive lung adenocarcinoma predicted by high-resolution CT features and AI quantitative parameters.Methods The imaging data of 95 patients with stage I invasive lung ad‐enocarcinoma were selected,and one or more pathological signs that met the criteria of vascular invasion,visceral pleural inva‐sion,regional lymph node metastasis,and pulmonary alveolar dissemination were defined as the invasive group,and vice versa as the non-invasive group.The univariate and binary logistic regression analysis was used to analyze CT signs and AI quantitative parameters.Results There was statistical significance(P<0.05)in intergroup comparisons of nodule type,spiciness sign,pleural traction sign,nodule size,volume,mass,solid proportion,minimum CT value,average CT value,median,standard de‐viation,skewness,kurtosis,etc.The minimum CT value,average CT value,proportion of solid volume,nodule type,and pleu‐ral traction sign were independent risk factors for invasive lung adenocarcinoma,and the average CT value had the highest diag‐nostic efficacy.The invasive group mainly showed solid nodules and pleural traction sign,with an average CT value higher than that of the non-invasive group.Conclusion CT features and AI quantitative parameters can predict the pathological invasive‐ness of lung adenocarcinoma,providing assistance for clinical postoperative treatment plans.