Insights into Imaging

Table 3 Deep learning model diagnostic performance in the two-way and seven-way classification on the test dataset

From: Saliency-based 3D convolutional neural network for categorising common focal liver lesions on multisequence MRI

	Two-way classification	Seven-way classification
	Two-way classification	Cyst	FNH	Haemangioma	Abscess	HCC	ICC	Metastasis
ACC (95% CI)	0.903 (0.834, 0.945)	0.991 (0.951, 1.000)	0.965 (0.913, 0.986)	0.956 (0.901, 0.981)	0.885 (0.813, 0.932)	0.991 (0.952, 1.000)	0.938 (0.878, 0.970)	0.805 (0.723, 0.868)
Sensitivity (95% CI)	0.930 (0.814, 0.976)	1.000 (0.806, 1.000)	0.909 (0.623, 0.995)	0.955 (0.782, 0.998)	1.000 (0.845, 1.000)	1.000 (0.796, 1.000)	0.909 (0.623, 0.995)	0.941 (0.730, 0.997)
Specificity (95% CI)	0.886 (0.790, 0.941)	0.990 (0.944, 0.999)	0.971 (0.917, 0.990)	0.956 (0.892, 0.983)	0.859 (0.773, 0.916)	0.990 (0.944, 0.999)	0.941 (0.878, 0.973)	0.781 (0.689, 0.852)
PPV (95% CI)	0.833 (0.704, 0.913)	0.941 (0.730, 0.997)	0.769 (0.497, 0.918)	0.840 (0.653, 0.936)	0.618 (0.450, 0.761)	0.938 (0.717, 0.997)	0.625 (0.386, 0.815)	0.432 (0.287, 0.591)
NPV (95% CI)	0.954 (0.873, 0.984)	1.000 (0.962, 1.000)	0.990 (0.946, 0.999)	0.989 (0.938, 0.999)	1.000 (0.954, 1.000)	1.000 (0.962, 1.000)	0.990 (0.944, 0.999)	0.987 (0.929, 0.999)
PLR (95% CI)	8.140 (4.218, 15.706)	97.000 (13.802, 681.695)	30.909 (9.981, 95.722)	21.716 (8.294, 56.860)	7.077 (4.076, 11.712)	98.000 (13.943, 688.794)	15.455 (60,955, 340,342)	4.303 (2.895, 6.395)
NLR 95% CI)	0.079 (0.026, 0.235)	0.000 (0.000, –)	0.094 (0.014, 0.607)	0.048 (0.007, 0.323)	0.000 (0.000, –)	0.000 (0.000, –)	0.097 (0.015, 0.626)	0.075 (0.011, 0.506)

ACC accuracy, PLR positive likelihood ratio, NLR negative likelihood ratio, NPV negative predictive value, PPV positive predictive value

Back to article page