Fig. 3.

Screening of characterization factors using Lasso regression analysis. (a) The plot displays vertical lines representing the lambda values selected through 10-fold cross-validation. The optimal lambda resulted in 9 nonzero coefficients, highlighting the importance of these variables in the predictive model. (b) The distribution of coefficients for the 16 texture features is depicted across the log(λ) series in the Lasso regression model. Vertical dashed lines indicate the minimum mean square error (λ = 0.004), corresponding to the lambda that minimizes cross-validation error, and the standard error of the minimum distance (λ = 0.022), which identifies a reduced set of 9 variables. This streamlined approach enhances model interpretability while minimizing the risk of overfitting, offering key insights into the regularization and parameter selection process