TABLE 2.

Performance of models using 5 × 10-fold Cross-validation—Mean (standard error).

	Accuracy	Precision	Sensitivity	Specificity	True positive	True negative	False positive	False negative
L1 regularized—Logistic regression	72.5 (3.8)	74.5 (4.1)	78.0 (4.9)	67.0 (5.4)	13.0 (0.6)	11.2 (0.3)	5.8 (0.3)	4.0 (0.6)
CNN	59.41 (1.93)	59.43 (1.90)	58.82 (4.07)	60.0 (3.07)	10.0 (0.69)	10.2 (0.52)	6.8 (0.52)	7.0 (0.69)
Pre-trained CNN	68.82 (1.05)	69.63 (1.90)	68.24 (4.27)	69.41 (3.86)	11.6 (0.72)	11.8 (0.65)	5.2 (0.65)	5.4 (0.72)

Bold indicates the best performing model.