Figure 3.

Feature selection by LASSO. (A) To differentiate STEMI ECG from control ECG, LASSO regression was used for variable screening. The results showed that 40 variables were retained when the error was the smallest; that is, the place corresponding to the dotted line on the left. To avoid overfitting and simplicity of the model, no more than one standard error was selected compared with the minimum error, and 14 variables were retained, which corresponded to the place on the dotted line on the right. (B) LASSO coefficient profiles of the 180 ECG features. A coefficient profile plot was produced against the ${log}_{\left(\lambda \right)}$ sequence. A vertical line was drawn at the selected optimizing value $(\lambda )$, which resulted in 14 nonzero coefficients. (C) To discriminate LAD and RCA/LCX, the results showed that when the error was minimal, approximately four variables were reserved; that is, the place corresponding to the dotted line on the left. (D) The LASSO screening results were further stepwise screened based on AIC, and four variables were retained.