Table 1.

Average prediction performance of different pathomic models for classifying TILs from noTILs in the training set with 5-fold CV. Bold indicates best performing models. Abbreviations: AUCROC = Area Under the Receiver Operating Characteristic curve; SMOTE = synthetic minority oversampling technique; LDA = Linear Discriminant Analysis; KNN = K-Nearest Neighbour; DT = Decision Tree; RF = Random Forest; MLP = Multi-layer Perceptron.

Models	Resample	AUCROC	Accuracy	Sensitivity (Recall)	Specificity	Precision	F1-score
RF	None	0.8576 ± 0.0029	0.8306 ± 0.0024	0.9398 ± 0.0025	0.4386 ± 0.0036	0.8572 ± 0.0009	0.8966 ± 0.0015
	SMOTE	0.8541 ± 0.0027	0.8162 ± 0.0034	0.8744 ± 0.0033	0.6077 ± 0.0083	0.8888 ± 0.0022	0.8815 ± 0.0023
	Down	0.8572 ± 0.003	0.7617 ± 0.0025	0.7486 ± 0.0029	0.8086 ± 0.0073	0.9335 ± 0.0023	0.8309 ± 0.0019
DT	None	0.7383 ± 0.0046	0.8085 ± 0.0018	0.9413 ± 0.0133	0.3322 ± 0.0556	0.8351 ± 0.0096	0.8849 ± 0.0006
	SMOTE	0.7712 ± 0.0165	0.7162 ± 0.01	0.7013 ± 0.0216	0.7699 ± 0.032	0.9164 ± 0.008	0.7943 ± 0.0106
	Down	0.7702 ± 0.021	0.7012 ± 0.0113	0.6763 ± 0.0203	0.7905 ± 0.027	0.9207 ± 0.0075	0.7796 ± 0.0115
LDA	None	0.8365 ± 0.0019	0.813 ± 0.0016	0.9467 ± 0.003	0.3332 ± 0.0069	0.8359 ± 0.0011	0.8878 ± 0.0011
	SMOTE	0.8395 ± 0.0024	0.725 ± 0.003	0.6949 ± 0.005	0.833 ± 0.0069	0.9372 ± 0.0021	0.798 ± 0.0029
	Down	0.8392 ± 0.0024	0.7237 ± 0.0026	0.693 ± 0.0037	0.8335 ± 0.0065	0.9372 ± 0.0022	0.7968 ± 0.0023
KNN	None	0.8232 ± 0.0036	0.817 ± 0.0022	0.9176 ± 0.002	0.4563 ± 0.0067	0.8582 ± 0.0015	0.8869 ± 0.0014
	SMOTE	0.8151 ± 0.0025	0.7159 ± 0.0039	0.6905 ± 0.0048	0.8069 ± 0.0079	0.9277 ± 0.0027	0.7917 ± 0.0033
	Down	0.8335 ± 0.0032	0.7385 ± 0.0056	0.7208 ± 0.0061	0.8017 ± 0.0044	0.9288 ± 0.002	0.8117 ± 0.0046
MLP	None	0.8524 ± 0.0035	0.8253 ± 0.0034	0.9279 ± 0.0093	0.4569 ± 0.0313	0.8598 ± 0.0059	0.8925 ± 0.0024
	SMOTE	0.8518 ± 0.0024	0.7628 ± 0.0254	0.7568 ± 0.0461	0.7844 ± 0.0492	0.9271 ± 0.0117	0.8324 ± 0.0235
	Down	0.8534 ± 0.0029	0.7559 ± 0.0196	0.743 ± 0.0344	0.8023 ± 0.0341	0.9313 ± 0.0085	0.826 ± 0.0186