Accelerated C-Arm Reconstruction by Out-of-Projection Prediction

3D image reconstruction from C-arm projections is a computationally demanding task. However, for interventional procedures a fast reconstruction is desirable. We present a method to reduce the number of actually back-projected voxels and ultimately the processing time by 20–30% without affecting image quality. The proposed method detects and skips subvolumes that are not visible in the current view. It works with projection matrices and is thus capable of handling arbitrary geometries. It can easily be incorporated into existing algorithms and is also suitable for the back-projection step of iterative algorithms.