Cancer Informatics Collaboratory
Home / Tools / PORTS radiomics test suite
Tool · PET radiomics · Open source

PORTS: PET Oncology Radiomics Test Suite

Radiomics turns medical images into large sets of numbers that may predict how a patient will do. PORTS was built to test whether those numbers can be trusted, starting with PET scans.

The research behind this page

  • Paul E. Kinahan: Professor of Radiology and Bioengineering, University of Washington; created the PORTS group. ✓ UW Radiology profile

These researchers created the original work. They did not write or review this page. ✓ marks a credential or role confirmed on an official profile; "as listed" roles come from the original page. All researchers featured on this site.

Based on: 1 official project or publisher site, 1 archived original page. See all sources

Get the code

What it is
Test objects, image databases, an open code base, reliability tests and best-practice guidance for PET radiomics
Code
PORTS 3D image texture metric calculation package (MATLAB)
Hosted on
MATLAB Central File Exchange
Group created
14 Nov 2015
Links checked 21 Nov 2026.

The problem PORTS addresses

Radiomics extracts many quantitative features from images, such as texture, shape and intensity, then looks for links to outcomes. Its promise is early prediction of how a tumour will respond and help choosing targeted therapies. The PORTS team warned that many published radiomics findings may not repeat, because imaging markers are hard to validate: scans are taken and analysed in different ways, and the data have the usual problems of very high-dimensional "omics" work.[1]

What PORTS provides

  • Standardised test objects and image databases
  • A validated, open-science code base for computing features
  • Tools to test how reliable each feature is
  • Best-practice guidance for PET radiomics studies[1]

How to use it

  • Compute 3D texture features on PET volumes with the same code other groups used, so results can be compared.
  • Run the reliability tests before trusting a feature in a prediction model.
  • Cite the version you used; papers have cited "PORTS version 1.01".[2]

Sources

  1. Archived copy of the original page (2 Aug 2022) · web.archive.org · Archived original
  2. MATLAB File Exchange listing · www.mathworks.com · Official site

About this page

This was the overview of the PORTS group on the hub, created by Paul E. Kinahan in November 2015. We rewrote it to explain the problem and point to where the code is available today.