Fast, robust dynamic field-of-view adjustment for iterative reconstruction of dedicated breast CT

Dedicated breast CT (bCT) produces high-resolution cone-beam CT breast images at radiation doses comparable to that of a screen-film mammographic exam, potentially allowing for breast cancer screening with bCT. Early indicators for breast cancer include microcalcifications clusters (MC), which consist of groups of calcium specks ranging from 200μm to 500μm in size. Iterative image reconstruction (IIR) is well-suited for low-dose CT image reconstruction, as it can alleviate the impact of data noise on the reconstructed images, but IIR is computationally demanding. In addition, the need to visualize MC requires high resolution and consequently large array sizes. Adjusting the reconstructed image array to tightly fit the actual breast, rather than the scanner field-of-view (FOV), improves computational efficiency. Here, we describe an automated method for identifying the breast bounding-box based on the analysis of two projection views that are 90 degrees apart and aligned with the reconstructed image array. The method was shown to be robust for 15 patient cases. Depending on breast size, computation time was reduced by 30% to 40%, compared to the full FOV reconstruction.