Newly developed image registration method for whole-body PET/CT images is applied and evaluated in oncology patients

A paper describing a new image registration method for whole-body PET/CT images and proof-of-concept for applications in oncology patients was recently published in Scientific Reports.

The study was performed as a collaboration between researchers from Uppsala University, Aarhus University Hospital and Antaros Medical. It primarily sought to develop a method for image registration that would enable analysis of global PET/CT imaging features that could hold important information related to treatment outcomes.

The registration method was then used, with Imiomics, to look at metabolic tumour volume changes in classical Hodgkin lymphoma (cHL) patients and therapy-related tissue volume changes in head and neck cancer (HNC) patients. The proof-of-concept found in these applications supports the potential use of this registration method and Imiomics in other disease areas where whole-body information from PET/CT images may be associated to patient outcomes, such as metabolic diseases and other types of cancer.

Originally developed using whole-body magnetic resonance (MR) images, Imiomics was conceived as an image analysis concept to address the under-utilisation of the vast amounts of data that is collected in whole-body imaging. Imiomics, in brief, involves deforming whole-body images, changing them to fit into a common coordinate system that is based based on statistically normal compositions.

The way that the source image is deformed to fit the common coordinate system is then captured using statistics and will describe how the source image differs from a normal composition. This has potential applications across many research avenues and disease areas. You can read more about our previous work with Imiomics via this publication.

This study highlighted the potential benefit of using this image registration method and Imiomics to quantify treatment outcomes using whole-body PET/CT images. As the method is almost fully automatic and requires little manual input, it can be scaled up for use in larger datasets for future research. Furthermore, the method should generalise well to the clinical setting, as it was evaluated in over 250 clinical images from multiple centres and different patient populations, with the only exclusion criterion being arm positioning. In applications where global imaging features are likely to be associated to patient outcomes, this method could also provide insights into which image regions are most informative.

Title: An image registration method for voxel-wise analysis of whole-body oncological PET-CT

Authors: Jönsson H, Ekström S, Strand R, Pedersen MA, Molin D, Ahlström H, Kullberg J

Find the publication here.

Share on Linkedin Share on Twitter