Fig. 2

Feature dimensionality reduction analysis. A Features were ranked according to their mRMR (maximum correlation and minimum redundancy) scores. The top 20 features were selected using the mRMR algorithm. B Selection of the tuning parameter (Lambda) in the LASSO model using tenfold cross-validation. Binomial deviances from the LASSO regression cross-validation model were plotted as a function of log (Lambda). The dotted vertical line at the right was drawn at the optimal value based on the minimum criteria and the 1-standard error rule (the 1-SE criteria). An optimal Lambda value of 0.067 with log (Lambda) = − 1.174 and 6 non-zero coefficients were selected. C Profiles of the LASSO coefficients for the 6 texture features. The vertical line was drawn at a value selected from the log (λ) sequence using tenfold cross-validation. Six features of non-zero coefficients are shown. D Selected radiomic features and corresponding coefficients