CERN/TIGRE

What it solves

TIGRE provides a fast and accurate way to perform 3D tomographic reconstruction. It addresses the challenge of processing large amounts of x-ray absorption data to create high-quality 3D images, specifically optimizing these computationally expensive operations for GPUs to reduce reconstruction time.

How it works

It combines high-level languages (Python and MATLAB) for ease of use with low-level CUDA performance for speed. The toolbox implements a wide variety of projection and backprojection operations that can be distributed across multiple GPUs. It supports flexible CT geometries, including Cone Beam, Parallel Beam, and Digital Tomosynthesis, where geometric parameters can be defined per projection.

Who it’s for

It is designed for the tomographic community, including algorithm developers, imaging researchers, and clinicians who need a high-performance tool for CT reconstruction.

Highlights

• GPU Acceleration: Optimized for single and multi-GPU setups to handle large datasets.
• Diverse Algorithms: Includes Filtered Backprojection (FBP), gradient-based iterative algorithms (SART, SIRT), Krylov subspace methods, statistical reconstruction (MLEM), and variational methods (FISTA).
• Flexible Geometry: Supports various scan geometries and allows per-projection parameter definition.
• Comprehensive Tooling: Includes TV denoising for 3D images, image quality metrics, and loaders for Nikon, Varian, and Phillips DICOM data.