Room: Exhibit Hall | Forum 1
Purpose: As compared with mammography, cone-beam breast CT(CBBCT) improves breast cancer diagnosis at the cost of increased imaging dose. In this work, we propose an iterative reconstruction algorithm for low-dose CBBCT from very few projections with the capability of accurate calcification detection.
Methods: As the projection number decreases, conventional iterative CBBCT reconstruction loses small structures and calcification detection becomes challenging. Based on the fact that calcifications are small and sparsely distributed, we propose to add a new L1 term in the objective function of the iterative CT reconstruction, which models only the calcification signals. The optimization problem is then solved by a standard gradient-based method. By separating a sparse distribution of calcifications from the background image, our method significantly reduces the data requirement for accurate calcification detection. The method performance is evaluated using both simulation and patient studies.
Results: We compare the FBP reconstruction on a full scan, the iterative reconstruction using total variation regularization (TVR) and the proposed reconstruction on reduced projections. In the simulation study, one hundred calcifications are simulated at random locations on ten water discs. The average contrasts of the calcifications are 94.5% and 32.71% of the full-scan FBP reconstruction for the TVR result with 50 and 15 projections, respectively. The proposed algorithm improves these calcification contrasts to 104.5% and 91.8%. A similar performance is found on the patient images. One calcification is first identified on the full-dose CBBCT. The calcification is clearly seen on the image reconstructed by the proposed algorithm even when the projection number drops to 15, while the TVR reconstruction loses the calcification at 20 projections.
Conclusion: Compared with the TVR method, the proposed iterative CT reconstruction algorithm has a superior performance on calcification detection on CBBCT, especially when projections are very few.