When used as decision support, interactive use of CAD for malignant masses on mammograms may be more effective than the current use of CAD, which is aimed at the prevention of perceptual oversights.

Objective:To evaluate an interactive computer-aided detection (CAD) system for reading mammograms to improve decision making.Methods:A dedicated mammographic workstation has been developed in which readers can probe image locations for the presence of CAD information. If present, CAD findings are displayed with the computed malignancy rating. A reader study was conducted in which four screening radiologists and five non-radiologists participated to study the effect of this system on detection performance. The participants read 120 cases of which 40 cases had a malignant mass that was missed at the original screening. The readers read each mammogram both with and without CAD in separate sessions. Each reader reported localized findings and assigned a malignancy score per finding. Mean sensitivity was computed in an interval of false-positive fractions less than 10%.Results:Mean sensitivity was 25.1% in the sessions without CAD and 34.8% in the CAD-assisted sessions. The increase in detection performance was significant (p = 0.012). Average reading time was 84.7 ± 61.5 s/case in the unaided sessions and was not significantly higher when interactive CAD was used (85.9 ± 57.8 s/case).Conclusion:Interactive use of CAD in mammography may be more effective than traditional CAD for improving mass detection without affecting reading time.

When reading mammograms, radiologists do not only look at local properties of suspicious regions but also take into account more general contextual information. This suggests that context may be used to improve the performance of computer-aided detection (CAD) of malignant masses in mammograms. In this study, we developed a set of context features that represent suspiciousness of normal tissue in the same case. For each candidate mass region, three normal reference areas were defined in the image at hand. Corresponding areas were also defined in the contralateral image and in different projections. Evaluation of the context features was done using 10-fold cross validation and case based bootstrapping. Free response receiver operating characteristic (FROC) curves were computed for feature sets including context features and a feature set without context. Results show that the mean sensitivity in the interval of 0.05-0.5 false positives/image increased more than 6% when context features were added. This increase was significant ( p < 0.0001). Context computed using multiple views yielded a better performance than using a single view (mean sensitivity increase of 2.9%, p < 0.0001). Besides the importance of using multiple views, results show that best CAD performance was obtained when multiple context features were combined that are based on different reference areas in the mammogram.

In computer-aided diagnosis (CAD) research, feature selection methods are often used to improve generalization performance of classifiers and shorten computation times. In an application that detects malignant masses in mammograms, we investigated the effect of using a selection criterion that is similar to the final performance measure we are optimizing, namely the mean sensitivity of the system in a predefined range of the free-response receiver operating characteristics (FROC). To obtain the generalization performance of the selected feature subsets, a cross validation procedure was performed on a dataset containing 351 abnormal and 7879 normal regions, each region providing a set of 71 mass features. The same number of noise features, not containing any information, were added to investigate the ability of the feature selection algorithms to distinguish between useful and non-useful features. It was found that significantly higher performances were obtained using feature sets selected by the general test statistic Wilks' lambda than using feature sets selected by the more specific FROC measure. Feature selection leads to better performance when compared to a system in which all features were used.

False positive (FP) marks represent an obstacle for effective use of computer-aided detection (CADe) of breast masses in mammography. Typically, the problem can be approached either by developing more discriminative features or by employing different classifier designs. In this paper, the usage of support vector machine (SVM) classification for FP reduction in CADe is investigated, presenting a systematic quantitative evaluation against neural networks, k-nearest neighbor classification, linear discriminant analysis and random forests. A large database of 2516 film mammography examinations and 73 input features was used to train the classifiers and evaluate for their performance on correctly diagnosed exams as well as false negatives. Further, classifier robustness was investigated using varying training data and feature sets as input. The evaluation was based on the mean exam sensitivity in 0.05-1 FPs on normals on the free-response receiver operating characteristic curve (FROC), incorporated into a tenfold cross validation framework. It was found that SVM classification using a Gaussian kernel offered significantly increased detection performance (P = 0.0002) compared to the reference methods. Varying training data and input features, SVMs showed improved exploitation of large feature sets. It is concluded that with the SVM-based CADe a significant reduction of FPs is possible outperforming other state-of-the-art approaches for breast mass CADe.