Objective:

To evaluate the accuracy and sensitivity of computer-aided detection (CAD) for breast calcified lesions

and the correlation between the accuracy and sensitivity of CAD and the characteristics of calcified lesions.

Methods:

Full-field digital mammography and CAD system of GE Company were used to evaluate 45 mammographically calcified lesions proven by pathology. The impacts of calcification type

distribution and number of pure calcified lesions on the accuracy and sensitivity of CAD were evaluated.

Results:

The total accuracy of CAD for calcified lesions was 73.3% (33/45)

the sensitivity was 94.

1% (32/34)

and the positive predictive value was 76.2% (32/42). The accuracy (42%

8/19) and sensitivity (88%

7/8) of CAD for intermediate calcification were all lower than those for typical malignant calcification (both 96%

25/26). The accuracy of CAD was significantly correlated with classification and distribution characteristics of calcification (

P

0.001)

but not correlated with calcification number (

P

0.05). The sensitivity of CAD was only significant correlated with calcification number (

P

0.05).

Conclusion:

The accuracy and sensitivity of CAD for pure calcified lesions of breast in full-field digital mammography are high

and the same as the positive predictive value. It can be used sufficiently in clinic and breast cancer screening.