Table 1.

Segmentation Performance Analysis by Area Normalized Error

	Hough transforma	Gabor waveletsa	AP algorithmb	MST algorithmb	Proposed method
FPm	0.0198	0.0058	0.0371	0.0255	0.0064
FNm	0.2519	0.0577	0.0595	0.1168	0.0558
FP < 0.05 and FN < 0.05	10	45	50	40	43
min(FP, FN) < 0.05 and 0.05 < max(FP, FN) < 0.10	0	0	18	20	19
min(FP, FN) < 0.05 and max(FP, FN) > 0.10	0	0	11	18	22
0.05 < FP < 0.10 and 0.05 < FN < 0.10	8	22	0	0	0
0.05 < min(FP, FN) < 0.10 and max(FP, FN) > 0.10	0	0	0	1	0
FP > 0.10 and FN > 0.10	66	17	5	3	0

The first two rows report the mean FN and FP values. The next six rows report the distribution of FP and FN values. For all methods except the MST algorithm, the values were reported for 84 mammograms

^aValues for the Hough transform and Gabor wavelets were previously reported by Ferrari et al.¹³

^bValues for the adaptive pyramid and minimum spanning tree algorithm were previously reported by Fei ma et al.¹⁷ For the MST algorithm, the values were reported for 82 mammograms as the method did not identify the pectoral muscle in two mammograms