摘要
目的探讨双能量CT(DECT)碘图相关定量参数联合CT征象建模鉴别肺癌病理亚型的应用价值。方法2015年12月至2016年10月前瞻性连续纳入疑似肺癌患者162例,于手术前或活检前行双能量CT动脉期扫描,110例患者明确肺癌病理组织分型。评估110例肺癌的CT形态学特征,并测量病变碘图相关定量参数(碘浓度、碘比值)。比较不同肺癌亚型之间CT征象及碘图相关定量参数的差异。对定性资料采用χ2检验分析,组间两两比较采用Kruskal-Wallis秩和检验。对满足正态分布的定量参数采用单因素方差分析,对不满足正态分布者或等级资料采用Kruskal-Wallis秩和检验分析。以CT征象与CT征象联合碘图相关定量参数分别构建多项式logistic回归模型,评价其诊断肺癌亚型的准确性。结果110例肺癌中,腺癌48例,鳞癌36例,小细胞肺癌26例。3种肺癌亚型的CT征象评估中,病变直径、分布、毛刺征、胸膜牵拉征、血管受累、淋巴结融合征、纵隔侵犯及病变强化特征差异有统计学意义(P〈0.05)。鳞癌[(5.73±3.67)cm]和小细胞肺癌[(6.08±4.39)cm]直径均大于腺癌[(3.75±2.80)cm],差异有统计学意义(H=13.806,P〈0.05);分布上,腺癌多为周围型(31例),鳞癌(26例)和小细胞肺癌(21例)多为中央型;相对于小细胞肺癌(13例),腺癌多见毛刺征(44例);相对于鳞癌(10例)和小细胞肺癌(5例),腺癌多见胸膜牵拉征(36例);相对于腺癌(15例),小细胞肺癌多伴血管受累(19例);相对于腺癌(3例)和鳞癌(4例),小细胞肺癌多见淋巴结融合征(15例);相对于腺癌(7例),小细胞肺癌多侵犯纵隔(13例);相对于鳞癌(6例),小细胞肺癌多强化均匀(10例);余各组间CT征象差异无统计学意义。碘浓度及碘比值在不同肺癌亚型中差异有统计学意义(H=16.817,20.338,P〈0.001)。腺癌及鳞癌碘浓度分别为(1.50±0.80)、(1.40±0.40)mg/ml,均高于小细胞肺癌[(1.20±0.40)mg/ml];腺癌及鳞癌碘比值分别为(16.10±7.02)%、(15.05±4.62)%,均高于小细胞肺癌[(11.55±3.15)%],腺癌与鳞癌碘图相关定量参数差异无统计学意义。CT征象模型总体诊断准确率为69.1%,CT征象联合碘图定量参数诊断准确率为80.9%。结论双源CT双能量碘图定量参数在肺腺癌、鳞癌与小细胞肺癌中有明显差异,其联合CT征象可提高诊断肺癌亚型的准确性。
ObjectiveTo investigate the clinical usefulness of quantitative dual-source dual-energy CT (DECT) iodine enhancement metrics combined with morphological CT features in distinguishing different lung cancer subtypes. MethodsOne hundred and sixty-two consecutive patients suspected with lung cancer were prospectively enrolled and underwent DECT in arterial phase prior to biopsy or surgery. Tumor histological subtypes were determined in 110 patients. Two radiologists interpreted CT morphologic features of 110 lesions in a consensual manner. In addition, two radiologists independently contoured lesions and placed regions of interest in descending aorta or subclavian artery on the same section for normalization, from which automated computer measurements were generated: iodine density and iodine ratio (the ratio of iodine density of lesion to that of artery on the same section). DECT metrics and morphological CT features were compared among different lung cancer subtypes. Chi-square was used to compare qualitative parameters. One way ANOVA was used to compare quantitative parameters satisfying normal distribution, while those parameters not satisfying normal distribution or ranked data were compared by Kruskal-Wallis rank sum test. Multinomial logistic regression models were used to differentiate the histological subtypes of lung cancer: adenocarcinoma, squamous cell carcinoma (SCC), small cell lung cancer (SCLC). ResultsThere were 48 cases of adenocarcinomas, 36 cases of SCC and 26 cases of SCLC. In analysis of CT features, tumor diameter, distribution, spiculation, pleural retraction, vascular involvement, confluent mediastinal lymphadenopathy, encasement of mediastinal structures and enhancement heterogeneity showed statistical difference (all P〈0.05). The diameter of SCC[(5.73±3.67)cm] and SCLC [(6.08±4.39)cm] were larger than adenocarcinoma [(3.75±2.80 cm)] (H=13.806, P〈0.05). Adenocarcinomas were mostly located in the periphery (31 cases), while SCC (26 cases) and SCLC (21 cases) were mainly centrally located. Spiculation was mostly found in adenocarcinoma (44 cases) rather than SCLC (13 cases). Pleural retraction was mostly observed in adenocarcinoma (36 cases) rather than SCC (10 cases) and SCLC (5 cases). Vascular involvement was mostly found in SCLC (19 cases) rather than adenocarcinoma (15 cases). Confluent mediastinal lymphadenopathy was more frequently found in SCLC (15 cases) compared with adenocarcinoma (3 cases) and SCC (4 cases). Encasement of mediastinal structures was mostly found in SCLC (13 cases) rather than adenocarcinoma (7 cases). Homogeneous enhancement was more frequently found in SCLC (10 cases) than SCC (6 cases). No significant differences were observed in other CT features between any other two groups. Iodine density and iodine ratio were statistically different among these three subtypes lung cancer (H=16.817, 20.338, P〈0.001). Iodine density of adenocarcinoma and SCC was (1.50±0.80) and (1.40±0.40) mg/ml, respectively, higher than the (1.20±0.40) mg/ml for SCLC (P〈0.01). Iodine ratio of adenocarcinoma and SCC was (16.10±7.02)% and (15.05±4.62)%, respectively, higher than the (11.55±3.15)% for SCLC (P〈0.01). No significant difference was observed between adenocarcinoma and SCC. Accuracy of the model based on CT features was 69.1%, accuracy of the model based on CT features combined with DECT parameters was 80.9%. ConclusionsQuantitative DECT metrics are different among adenocarcinoma, SCC and SCLC, when combined with morphological CT features, higher diagnostic accuracy can be achieved.
作者
徐晓莉
隋昕
钟巍
徐燕
王子兴
宋兰
黄耀
王晓
金征宇
宋伟
Xu Xiaoli;Sui Xin;Zhong Wei;Xu Yan;Wang Zixing;Song Lan;Huang Yao;Wang Xiao;Jin Zhengyu;Song Wei(Department of Radiology,Peking Union Medical College Hospital,Chinese Academy of Medical Sciences and Peking Union Medical College,Beijing 100730,China)
出处
《中华放射学杂志》
CAS
CSCD
北大核心
2018年第11期823-828,共6页
Chinese Journal of Radiology
基金
中央级公益性科研院所基本科研项目(2017PT32004)
北京协和医学院青年科研基金(2017310049)