肾嫌色细胞癌与不典型透明细胞癌的CT纹理分析研究

目的研究影像组学方法在肾嫌色细胞癌和强化方式不典型的透明细胞癌二者中的应用。方法搜集行肾动脉CTA扫描的108例肾细胞癌患者的临床资料及影像学图像。应用影像组学中的Lasso回归统计方法和机器学习中的随机森林算法提取病例的CTA图像特征并使计算机学习,通过20次重复试验得到平均诊断准确率。患者的临床特征处理采用SPSS 20.0软件,计量资料用t检验,计数资料用χ^2检验,P<0.05为差异具有统计学意义。结果108例肾细胞癌中,透明细胞癌57例,嫌色细胞癌51例。两组病例临床特征中的性别和吸烟史差异具有统计学意义(P<0.05),透明细胞癌更多见于吸烟的男性患者。放射科医师对两组病例诊断的平均准确性为(45.42±3.32)%,低于Lasso回归(76.5±12.26)%和随机森林算法(78.5±6.3)%。在两组病例中,随机森林算法给出的总准确性、对嫌色细胞癌诊断的特异性要高于Lasso回归,Lasso回归对透明细胞癌诊断的敏感性高于随机森林算法。结论影像组学方法可以对肾嫌色细胞癌及透明细胞癌做出有效的鉴别诊断,且诊断能力高于放射科医师的能力。影像组学作为一种新兴的研究方法,有望为医学发展带来重要变革。

CT-Based Radiomic Analysis of Chromophobe Cell Renal Carcinoma and Clear Cell Renal Carcinoma with Low-Contrast Enhancement

Objective To study the application of radiomics in chromophobe cell renal carcinomas(chRCCs)and clear cell carcinomas with low-contrast enhancement(Ice-ccRCCs).Methods Clinical features and images of 108 patients undergoing renal cell carcinoma CTA scan in the Department of Radiology from January 2014 to December 2017 were analyzed.Lasso regression and random forest were used to analyze the imaging features.The average diagnostic accuracy was obtained through 20 repeated tests.The measurement data were analyzed by t-test,and the enumeration data were analyzed by chi-square test using Statistical Product and Service Solutions(SPSS)version 20.0.P<0.05 was considered statistically significant.Results Of all the 108 patients,57 patients were treated for lce-ccRCCs and 51 patients were treated for chRCCs.Our study showed two indicators(gender and smoking)were statistically significantly different between the two groups(P<0.05).The accuracy of the diagnoses of the two experienced radiologists together was significantly lower than the accuracies of the Lasso regression(76.5%±12.26%)and random forest model(78.5%±6.3%).In the two groups of cases,the total accuracy and the specificity of the random forest for the diagnosis of chRCCs was higher than Lasso regression,while the sensitivity of Lasso regression for the diagnosis of lce-ccRCCs was higher than that of the random forest.Conclusion This study demonstrated that radiomic features could potentially provide great diagnostic value for classifying lce-ccRCCs and chRCCs than radiologists.As a new research method,radiomics is expected to bring important changes to medical development.