Collaborative Computational Project in Synergistic Reconstruction for Biomedical Imaging

Synergistic Symposium 2019 Software

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Christopher Syben PYRO-NN: Python reconstruction operators in neural networks Recently, several attempts were conducted to transfer deep learning to medical image reconstruction. An increasingly number of publications follow the concept of embedding the computed tomography (CT) reconstruction as a known operator into a neural network. However, most of the approaches presented lack an efficient CT reconstruction framework fully integrated into deep learning environments. As a result, many approaches use workarounds for mathematically unambiguously solvable problems. PYRO-NN is a generalized framework to embed known operators for CT-Reconstruction into the prevalent deep learning framework Tensorflow. The current status includes state-of-the-art parallel-, fan-, and cone-beam projectors, and back-projectors accelerated with CUDA provided as Tensorflow layers. On top, the framework provides a high-level Python API to conduct FBP and iterative reconstruction experiments with data from real CT systems. To demonstrate the capabilities of the layers, the framework comes with baseline experiments, which are described in the supplementary material. The framework is available as open-source software under the Apache 2.0 licence. PYRO-NN comes with the prevalent deep learning framework Tensorflow and allows to setup end-to-end trainable neural networks in the medical image reconstruction context. We believe that the framework will be a step toward reproducible research and give the medical physics community a toolkit to elevate medical image reconstruction with new deep learning techniques.
Ander Biguri Accelerating tomographic reconstruction using multiple GPUs X-ray tomography reconstruction keeps requiring more and more computational power, specially when using iterative reconstruction. From data generation in optimization problems in medical applications, to scientific and industrial very large scale micro-tomography scans, the problems that researchers tackle require enormous amount of computing power. Generally, GPUs are used for this purpose, however the GPU hardware is not increasing its capabilities fast enough to cater the computing needs. In this work, we present a multi-GPU approach for iterative tomographic reconstruction for arbitrary amount of GPUs that is not bounded by the GPU hardware and that is included in the TIGRE toolbox. The software allows for very fast reconstruction, even on portable GPUs, and can reconstruct very large images, as large as the CPU RAM allows. The proposed multi-GPU splitting method can be easily applied to other tomographic modalities.
Richard Brown SIRF: Synergistic Image Reconstruction Framework The combination of PET with MR opens the way to more accurate diagnosis and improved patient management. At present, the data acquired by PET-MR scanners are essentially processed separately, but the opportunity to improve accuracy of the tomographic reconstruction via synergy of the two imaging techniques is an active area of research.

We will present the current status of the CCP-PETMR SIRF software suite, which provides an open-source software platform for efficient implementation and validation of novel reconstruction algorithms. SIRF offers user-friendly Python and MATLAB interfaces built on top of C++ libraries. SIRF uses advanced PET and MR reconstruction software packages and tools. Currently, for PET this is Software for Tomographic Image Reconstruction; for MR, Gadgetron and ISMRMRD; and for image registration tools, NiftyReg. SIRF aims to be capable of reconstructing images from acquired scanner data, whilst being simple enough to be used for educational purposes.

Evangelos Papoutsellis X-ray reconstruction and analysis with the CCPi Core Imaging Library The Core Imaging Library (CIL) is an object-oriented framework for optimisation-based problems focused on tomography reconstruction. It allows the user to read and preprocess raw data acquired from specific scanner configurations, e.g. parallel or cone beam geometries, and run state of the art iterative algorithms. In addition, CIL can setup a ”mix & match” framework for regularised problems with a flexible combination of different data fitting terms and non-smooth/ sparse priors, e.g., TV, TGV, DTV, NLTV. Currently, it is tested for various instruments such as IMAT (Neutron imaging & Diffraction instrument), HEXITEC (spectroscopic), Nikon Metrology X-ray CT, Diamond Light Source (Synchrotron) for 2D, 3D and 4D dynamic and energy resolved data.