. 2019 Apr 23;91(10):6530–6540. doi: 10.1021/acs.analchem.8b05598

Table 1. Tissue Classification Performance (Avg ± Std Dev) of the PLS-DA Models, Using the Optimal Number of Pixels Equal to 900, and the Optimal Number of PLS-DA Components Equal to 7^a.

Cross-Validation
accuracy	0.9158 ± 0.0262
	breast	colorectal	ovarian
sensitivity	0.9030 ± 0.0385	0.9471 ± 0.0195	0.9840 ± 0.0308
specificity	0.9731 ± 0.0255	0.8605 ± 0.0589	0.9536 ± 0.0254

External Test
accuracy	0.8059 ± 0.0346
	breast	colorectal	ovarian
sensitivity	0.6775 ± 0.0481	-	0.9771 ± 0.0482
specificity	0.9911 ± 0.0310	-	0.6775 ± 0.0481

The cross-validation results represent the performances of the models on the 10-fold cross-validation, and the external test results represent the performances of the prediction of the external test MS images using the optimal parameters identified by the cross-validation.

Table 1. Tissue Classification Performance (Avg ± Std Dev) of the PLS-DA Models, Using the Optimal Number of Pixels Equal to 900, and the Optimal Number of PLS-DA Components Equal to 7a.

Table 1. Tissue Classification Performance (Avg ± Std Dev) of the PLS-DA Models, Using the Optimal Number of Pixels Equal to 900, and the Optimal Number of PLS-DA Components Equal to 7^a.