Computer-Aided Assessment of Tumor Grade for Breast Cancer in Ultrasound Images

Table 2

Performance of the proposed CAD system when different feature sets were used.

	Feature type
	Selected	All	Morphological	Ellipsoid fitting	Textural	Posterior acoustic
Accuracy	85.14%	77.03%	66.89%	70.95%	66.22%	78.38%
Sensitivity	79.31%	62.07%	37.93%	41.38%	72.41%	48.28%
Specificity	86.55%	80.67%	73.95%	78.15%	64.71%	85.71%
PPV	58.97%	43.90%	26.19%	31.58%	33.33%	45.16%
NPV	94.50%	89.72%	83.02%	84.55%	90.59%	87.18%
Az	0.7940	0.6953	0.4490	0.5575	0.7068	0.6647

Accuracy = (TP + TN)/(TP + TN + FP + FN); sensitivity = TP/(TP + FN); specificity = TN/(TN + FP); PPV = TP/(TP + FP); NPV = TN/(TN + FN), where TP is true positive (the number of high-grade tumors classified correctly); FN is false negative (the number of high-grade tumors classified incorrectly); FP is false positive (the number of low-grade tumors classified incorrectly); TN is true negative (the number of low-grade tumors classified correctly).