Reconstruction is the process by which a set of 2D images are processed to create a 3D volume. Tomosynthesis reconstruction is a far more difficult problem than conventional CT because in tomosynthesis an incomplete and sparse set of projection data is acquired. Breast tomosynthesis adds the additional burden of very high resolution projection images which pose a significant information processing challenge.
Recent advances in non-linear, iterative reconstruction algorithms have shown great promise in overcoming the image quality problems in tomosynthesis reconstruction. Computational techniques for non-linear iterative reconstruction are being developed further by Dexela to improve the quality of the reconstructions from the sparse and noisy data available. An optimised implementation of the reconstruction algorithm reduces the time taken to perform the reconstruction to a clinically acceptable level.
Improved reconstruction has also reduced the number of projection images required and allowed the use of a wider angular range. This leads to higher resolution and better contrast, making use of variable dose distribution and variable spacing between projections. This also impacts acquisition time.
The reconstructor can be implemented entirely in software or using a GPU for hardware accelerator. It can also be used in other medical applications.