Texture-based classification of different single liver lesion based on SPAIR T2W MRI images

BMC Medical Imaging

Table 3 Summary of classification results obtained by 10 fold cross-validation on three classification groups by BP-ANN, KNN, SVM and logistic regression model

Models	HH VS.HM						HM VS. HCC						HH VS. HCC
	Acc%	Sens %	Spec %	Mcc	RSD %	AUC %	Acc %	Sens %	Spec %	Mcc	RSD %	AUC %	Acc %	Sens %	Spec %	Mcc	RSD %	AUC %
BP-ANN	88	89	86	0.86	5.5	89	90	97	92	0.79	5.8	91	90	96	85	0.80	6.4	91
KNN	85	86	85	0.75	3.8	86	87	87	88	0.76	4.6	88	85	57	90	0.77	4.3	83
SVM	83	81	86	0.75	6.1	84	83	70	93	0.77	6.0	84	88	42	95	0.83	5.2	86
Logistic regression	79	81	78	0.83	2.6	80	87	87	88	0.73	1.9	85	88	86	90	0.78	3.8	89
Radiologist	87	81	86	0.76	-	-	90	96	85	0.78	-	-	92	100	85	0.70	-	-

Abbreviations: BP-ANN back-propagation artificial neural network, KNN K-nearest neighbor, SVM Support vector machine, HH hepatic hemangioma, HM hepatic metastases, HCC hepatocellular carcinoma. Area Under the ROC Curve, Accuracy, Sensitivity, Specificity and relative standard deviation are denoted as AUC, Acc, Sens, Spec and RSD respectively

ISSN: 1471-2342