A Fusion-Based Approach for Breast Ultrasound Image Classification Using Multiple-ROI Texture and Morphological Analyses

Table 4

Objective performance metrics obtained using the (a) conventional classification approach using texture features, (b) conventional classification approach using morphological features, (c) conventional classification approach using both texture and morphological features, (d) proposed classification approach using multiple-ROI texture analysis, and (e) proposed classification approach using the fused multiple-ROI texture analysis and morphological analysis.

	(a)	(b)	(c)	(d)	(e)
Accuracy	85.5%	87.3%	90.9%	95.5%	98.2%
Specificity	84.4%	89.1%	90.6%	93.8%	98.4%
Sensitivity	87.0%	84.8%	91.3%	97.8%	97.8%
PPV	80.0%	84.8%	87.5%	91.8%	97.8%
NPV	90.0%	89.1%	93.6%	98.4%	98.4%
MCC	70.7%	73.9%	81.5%	90.9%	96.3%