Noninvasive Prediction of TERT Promoter Mutations in High-Grade Glioma by Radiomics Analysis Based on Multiparameter MRI

Figure 3

Radiomics feature selection using LASSO. (a) Selection of the optimal value of lambda (λ). Tuning log(λ) selection in the LASSO model used to perform 5-fold cross-validation via the minimum criteria. (b) The LASSO coefficient profiles included 1230 features. The vertical line was drawn at the selected log(λ) above, and 6 features with nonzero coefficients were finally identified.

(a)

(b)