AGILE: An open source library for image reconstruction using graphics card hardware acceleration

Introduction: Iterative image reconstruction of undersampled data from multiple coils has shown a huge potential for a wide range of applications during the last years [1,2]. While pronounced improvements in imaging speed were achieved with the proposed methods, the reconstruction times of most implementations were too long for online image reconstruction, which is necessary in order to facilitate application in clinical routine. Parallelized implementations on graphics hardware were recently established as a feasible method to significantly speedup image reconstruction [3]. While programming graphics hardware was simplified significantly with the introduction of dedicated libraries for general purpose computing like CUDA [4] or OpenCL [5], efficient implementation, especially concerning memory management, is still a challenging task. However, the mathematical operations needed for image reconstruction are often very similar, and can be re-used in future implementations. These include basic matrix-vector operations, algorithms to solve linear systems of equations, (non-uniform)-FFTs and solvers for constrained optimization problems. The goal of this work is to introduce the open source library AGILE (Environment for Linear and non-linear Image reconstruction using Gpu Acceleration), designed for image reconstruction on GPUs in biomedical sciences. It is based on highly efficient implementations of numerical methods, but also includes code for iterative MR image reconstruction as well as a framework for finite element calculations and applications for Fluorescence Tomography [6].