Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PyRadiomics-cuda: a GPU-accelerated 3D features extraction from medical images within PyRadiomics

Jakub Lisowski, Piotr Tyrakowski, Szymon Zyguła, Krzysztof Kaczmarski

TL;DR

PyRadiomics-cuda presents a GPU-accelerated extension that transparently speeds up 3D shape feature extraction in PyRadiomics by offloading mesh generation and diameter computations to CUDA while preserving API compatibility. It integrates via a build-time dispatcher that falls back to CPU when no GPU is available, enabling seamless adoption in existing pipelines. Benchmarking on KITS19 across multiple hardware setups shows substantial speedups (up to several orders of magnitude on modern GPUs) with remaining bottlenecks in data loading and transfer, indicating strong potential for high-throughput radiomics in AI workflows. The work demonstrates meaningful reductions in feature extraction time, enabling scalable analyses for large-scale clinical datasets such as the xLUNGS project, with practical implications for infrastructure cost and turnaround time.

Abstract

PyRadiomics-cuda is a GPU-accelerated extension of the PyRadiomics library, designed to address the computational challenges of extracting three-dimensional shape features from medical images. By offloading key geometric computations to GPU hardware it dramatically reduces processing times for large volumetric datasets. The system maintains full compatibility with the original PyRadiomics API, enabling seamless integration into existing AI workflows without code modifications. This transparent acceleration facilitates efficient, scalable radiomics analysis, supporting rapid feature extraction essential for high-throughput AI pipeline. Tests performed on a typical computational cluster, budget and home devices prove usefulness in all scenarios. PyRadiomics-cuda is implemented in Python and C/CUDA and is freely available under the BSD license at https://github.com/mis-wut/pyradiomics-CUDA Additionally PyRadiomics-cuda test suite is available at https://github.com/mis-wut/pyradiomics-cuda-data-gen. It provides detailed handbook and sample scripts suited for different kinds of workflows plus detailed installation instructions. The dataset used for testing is available at Kaggle https://www.kaggle.com/datasets/sabahesaraki/kidney-tumor-segmentation-challengekits-19

PyRadiomics-cuda: a GPU-accelerated 3D features extraction from medical images within PyRadiomics

TL;DR

PyRadiomics-cuda presents a GPU-accelerated extension that transparently speeds up 3D shape feature extraction in PyRadiomics by offloading mesh generation and diameter computations to CUDA while preserving API compatibility. It integrates via a build-time dispatcher that falls back to CPU when no GPU is available, enabling seamless adoption in existing pipelines. Benchmarking on KITS19 across multiple hardware setups shows substantial speedups (up to several orders of magnitude on modern GPUs) with remaining bottlenecks in data loading and transfer, indicating strong potential for high-throughput radiomics in AI workflows. The work demonstrates meaningful reductions in feature extraction time, enabling scalable analyses for large-scale clinical datasets such as the xLUNGS project, with practical implications for infrastructure cost and turnaround time.

Abstract

PyRadiomics-cuda is a GPU-accelerated extension of the PyRadiomics library, designed to address the computational challenges of extracting three-dimensional shape features from medical images. By offloading key geometric computations to GPU hardware it dramatically reduces processing times for large volumetric datasets. The system maintains full compatibility with the original PyRadiomics API, enabling seamless integration into existing AI workflows without code modifications. This transparent acceleration facilitates efficient, scalable radiomics analysis, supporting rapid feature extraction essential for high-throughput AI pipeline. Tests performed on a typical computational cluster, budget and home devices prove usefulness in all scenarios. PyRadiomics-cuda is implemented in Python and C/CUDA and is freely available under the BSD license at https://github.com/mis-wut/pyradiomics-CUDA Additionally PyRadiomics-cuda test suite is available at https://github.com/mis-wut/pyradiomics-cuda-data-gen. It provides detailed handbook and sample scripts suited for different kinds of workflows plus detailed installation instructions. The dataset used for testing is available at Kaggle https://www.kaggle.com/datasets/sabahesaraki/kidney-tumor-segmentation-challengekits-19

Paper Structure

This paper contains 12 sections, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related works
  3. Architecture of the system
  4. PyRadiomics integration
  5. Efficiency experiments and runtime results
  6. Test framework
  7. Hardware used and testing data set
  8. Discussion of the results
  9. Conclusions
  10. Competing interests
  11. Author contributions statement
  12. Acknowledgments

Figures (2)

  • Figure 1: Time efficiency of the KITS dataset feature extraction using PyRadiomics and PyRadiomics-cuda on the various machines. (See Supplementary Information for exact data values of this plot.)
  • Figure 2: Available speedup of 3D features processing when using PyRadiomics-cuda on various GPU devices compared to the original implementation working on CPU processors. Intel Xeon with PyRadiomics taken as a base reference time. (See Supplementary Information for exact data values of this plot.)