Table IV.
The average number of FPs at selected sensitivities for the single-view and two-view fusion detection methods. The average number of FPs is compared for the entire test data set, as well as the malignant and benign subsets. For computing the sensitivity in mammogram-based scoring, the same cluster seen on two views was counted independently. Since there were 218 microcalcification cluster locations marked by the radiologist on the test set, the denominator for sensitivity in mammogram-based analysis was 218. In case-based analysis, a TP was defined as a positive case for which a cluster was correctly detected on one or both views. Since we had 96 mammogram pairs in our data positive test set, denominator for sensitivity in case-based analysis was 96.
Scoring method | Data subset | Average number of FPs per image |
|||
---|---|---|---|---|---|
80% sensitivity |
70% sensitivity |
||||
Single-view | Fusion | Single-view | Fusion | ||
Mammogram-based | Entire test set | 0.53 | 0.46 | 0.30 | 0.18 |
Malignant | 0.35 | 0.16 | 0.24 | 0.04 | |
Benign | 0.63 | 0.56 | 0.37 | 0.30 | |
Case-based | Entire test set | 0.26 | 0.30 | 0.17 | 0.04 |
Malignant | 0.22 | 0.04 | 0.13 | 0.03 | |
Benign | 0.35 | 0.39 | 0.24 | 0.25 |