Automated Segmentation of Mass Regions in DBT Images Using a Dilated DCNN Approach

Table 2

Comparisons of typical studies in mass regions detection of the DBT images.

Method	Classifier	DBT image data set size	Sensitivity	Accuracy	AUC
Chan et al. [17]	LDA	100	80%	/	/
van Schie et al. [18]	NN	752	80%	/	/
Palma et al. [20]	SVM	101	90%	/	/
Kim et al. [28]	SVM	160	/	/	0.847
Fotin et al. [29]	DCNN	344	89%	86.4%	/
Samala et al. [30]	DCNN	324	80%	/	0.80
Reiser et al. [31]	LDA	36	90%	/	/
Proposed	Dilated DCNN	97	85.6%	86.3%	0.852