Computer‐aided detection of masses in digital tomosynthesis mammography: Comparison of three approaches

Abstract

The authors are developing a computer-aided detection ͑CAD͒ system for masses on digital breast tomosynthesis mammograms ͑DBT͒. Three approaches were evaluated in this study. In the first approach, mass candidate identification and feature analysis are performed in the reconstructed three-dimensional ͑3D͒ DBT volume. A mass likelihood score is estimated for each mass candidate using a linear discriminant analysis ͑LDA͒ classifier. Mass detection is determined by a decision threshold applied to the mass likelihood score. A free response receiver operating characteristic ͑FROC͒ curve that describes the detection sensitivity as a function of the number of false positives ͑FPs͒ per breast is generated by varying the decision threshold over a range. In the second approach, prescreening of mass candidate and feature analysis are first performed on the individual two-dimensional ͑2D͒ projection view ͑PV͒ images. A mass likelihood score is estimated for each mass candidate using an LDA classifier trained for the 2D features. The mass likelihood images derived from the PVs are backprojected to the breast volume to estimate the 3D spatial distribution of the mass likelihood scores. The FROC curve for mass detection can again be generated by varying the decision threshold on the 3D mass likelihood scores merged by backprojection. In the third approach, the mass likelihood scores estimated by the 3D and 2D approaches, described above, at the corresponding 3D location are combined and evaluated using FROC analysis. A data set of 100 DBT cases acquired with a GE prototype system at the Breast Imaging Laboratory in the Massachusetts General Hospital was used for comparison of the three approaches. The LDA classifiers with stepwise feature selection were designed with leave-one-case-out resampling. In FROC analysis, the CAD system for detection in the DBT volume alone achieved test sensitivities of 80% and 90% at average FP rates of 1.94 and 3.40 per breast, respectively. With the 2D detection approach, the FP rates were 2.86 and 4.05 per breast, respectively, at the corresponding sensitivities. In comparison, the average FP rates of the system combining the 3D and 2D information were 1.23 and 2.04 per breast, respectively, at 80% and 90% sensitivities. The difference in the detection performances between the 2D and the 3D approach, and that between the 3D and the combined approach were both statistically significant ͑p = 0.02 and 0.01, respectively͒ as estimated by alternative FROC analysis. The combined system is a promising approach to improving automated mass detection on DBTs.

Digital Breast Tomosynthesis: Lessons Learned from Early Clinical Implementation

“…Conflicting data have been reported regarding detection of microcalcifications at DBT (37,38). In addition, there currently is no commercially available computeraided detection (CAD) system for DBT, although active research in this field suggests improved performance of 3D CAD compared with DM CAD (39)(40)(41)(42)(43).…”

Digital Breast Tomosynthesis: State of the Art

“…DBT projections (137,138), DBT reconstructions (138)(139)(140)(141), and the combination of 2D projections and reconstructions (138,142) (145). Considering that the field is gravitating toward a combination of DBT and 2D mammography (either SM or FFDM), the benefit of applying CAD to DBT versus CAD to 2D mammography that is widely used is unclear and needs investigation.…”

Characterization of masses in digital breast tomosynthesis: Comparison of machine learning in projection views and reconstructed slices

“…Although the information content in the original PV images is the same, the PV and DBT processing analyze the information in different ways; the extracted information may represent different characteristics of the lesions and provide complementary information to improve the overall accuracy, as demonstrated in our previous mass detection study. 18 Although development of CAD for DBT is similar to that for mammography to a certain extent, the additional 3D information and the flexibility of using different approaches pose greater challenges as well as opportunities for designing computerized lesion detection and characterization algorithms. Many different computer-vision techniques can be designed at each step in each approach.…”

“…[13][14][15][16][17][18] Very few studies have been conducted for computerized lesion characterization in DBT to date. Chan et al 19,20 investigated the feasibility of classification of malignant and benign masses using the reconstructed DBT slices as input.…”