基于胸部增强CT影像组学模型用于胸腺瘤分类的研究

蒋佻宴, 贾田颖, 张琴

Contrast-enhanced computed tomography-based radiomics models for the risk categorization of thymoma

JIANG Tiaoyan, JIA Tianying, ZHANG Qin

图2 使用LASSO回归算法进行特征选择以及基线 logistic 回归模型的性能

Fig. 2 Feature selection using the LASSO regression algorithm and the performance of the baseline logistic regression model A: The LASSO coefficients of the 581 features; B: Mean square error with respect to log(λ). The average binominal deviance values for each model at a given λ were indicated by the dashed red curve; C: Mean accuracy of the baseline logistic model at a given number of features. Two regional maxima were found at 4 and 14 features.