Table 1.

The characteristics of the mammography CAD data sets used in this work: N_V is the number of views per case, $N_N^C$ is the number of normal cases, $N_A^C$ is the number of abnormal cases, N_N is the number of normal views, N_A is the number of abnormal views, and N_L is the relevant denominator for calculation of LLF (see text). N′ and U′ are the observed numbers of NLs and LLs, respectively, and R is the number of discrete ratings. The quantity α is the x-coordinate of the operating point at which the CAD algorithm is to be used in the clinic. C and V denote the case-based and view-based methods, respectively.

Data set	Task	NV	Cases	Views	NL	N′	U′	R	α	Analysis Method
			$N_N^C/N_A^C$	NN / NA
CAD_A	Mass	4	200 / 250	1314 / 486	250	5014	234	20	0.48	C
CAD_B	Mass	4	200 / 250	1314 / 486	250	4447	233	19	0.68	C
CAD_C	Mass	4*	0 / 111	195 / 217	111	1139	95	17	0.59	C
CAD_D	Mass	4*	0 / 195	342 / 380	195	2029	189	18	0.46	C
CAD_E	Micro	2	71 / 96	142 / 192	218	2048	202	18	1.0	V
CAD_F	Mass	2	0 / 58	0 / 116	116	465	101	12	2.0	V
CAD_G	Micro	2	71 / 96	142 / 192	96	2048	96	12	1.0	C

^*a few cases had only 2 views