Table 4.

The calculation formulas of performance metrics.

Metric	*Formula
TPR	$\frac{TP}{TP+FN}$
TNR	$\frac{\text{TN}}{\text{TN}+\text{FP}}$
Accuracy	$\frac{\text{TP}+\text{TN}}{\text{TP}+\text{FP}+\text{TN}+\text{FN}}$
Precision	$\frac{\text{TP}}{\text{TP}+\text{FP}}$
AUC	${\int }_{\text{x}=\text{0}}^{\text{1}}TPR(FP{R}^{-1}(x))dx$, where x1 is the score for a positive instance and x0 is the score for a negative instance.
Kappa	$\text{Kappa}=\frac{{\text{P}}_o-P_e}{1-P_e}$, ${\text{P}}_e=\frac{\text{P}(TP+FP)+N(TN+FN)}{{(T+N)}^2}$ where Po = Accuracy,
MAE	$\frac{\text{1}}{n}{\displaystyle \sum _{i=1}^n}|p(i)-a(i)|$, where p(i) is the prediction case, and a(i) is real case, n is the total cases.

^*TP is true positive, it means that the outcome from a prediction is lung adenocarcinoma (Adc) and the actual value is also Adc. FN is false negative, it means that the prediction outcome is another lung cancer histological type(Oth) while the actual value is Adc. TN is true negative, it means that both the prediction outcome and the actual value are Oth. FP is false positive, it means that the outcome from a prediction is Adc while the actual value is Oth. P is condition positive, N is condition negative, and MAE is the mean absolute errors. TPR is true positive rate, it measures the proportion of actual patients with Adc that are correctly identified. A negative result in a test with high TPR is useful for ruling in disease, it signifies a high probability of the presence of Oth. TNR is true negative rate, it measures the proportion of actual patients with Oth that are correctly identified. A test with 100% TNR will recognize all patients with Oth by testing negative, and a positive test result would definitively rule out the presence of Oth in a patient.