Artificial intelligence analysis of brain imaging in MS

Dr Minh-Son To

Flinders University

| Better treatments | Neurobiology | Incubator | 2021 | Investigator Led Research |


Magnetic resonance imaging (MRI) plays a central role in the clinical management of MS. The aim of this research is to develop artificial intelligence (AI) tools to analyse MRI scans, to assist clinicians and radiologists with decision making.

Dr Minh-Son To and his team will focus on detecting lesion changes on images taken over an extended period, since the appearance of new lesions or evidence of new activity can influence choice of therapy. Once they have validated this on a retrospective dataset, they will develop a software interface for deploying this tool in a clinical environment. This will enable future prospective trials to take place.

Project Outcomes

Dr Minh-Son To and his team worked on a project employing advanced AI to analyse MRI scans of individuals diagnosed with MS. Their primary goal was to develop algorithms that could automatically detect changes in lesions over time, aiding radiologists in interpreting and reporting the scans. Additionally, they aimed to link these changes with the treatments people with MS were undergoing to potentially personalise their care.

The team successfully engineered a groundbreaking AI algorithm specifically designed to detect alterations in lesions. However, they encountered challenges when attempting to forecast these changes over time and establish connections with the therapies being administered. As a result, they didn't pursue correlation with treatment.

The team used a special type of smart computer system called 'generative adversarial networks' (GANs) for this project. These networks have two parts that work against each other - one to create realistic images and the other to tell real from fake images. They cleverly used one of these parts to spot differences between two pictures, which had never been done before in this kind of computer learning.

This part they used could make a kind of map showing where new or changing spots appeared in the scans. It helped doctors quickly see the areas that were different in the pictures. However, it was not able to guess how the spots might change over time. The team thought this may be because the computer was trained for one job but not performing as well in other related tasks in this setup.

Despite the challenges, they built a software system that could retrieve old MRI images from hospital records, process them for analysis by AI, and then convert them back to a format that doctors normally use. This was designed to test how well their AI can detect changes in these images. Currently, they're in the process of preparing this evaluation with radiologists and trainees in their local radiology department, although this phase isn't yet completed.

Ultimately, Dr To and the team hope this technology will assist medical professionals in better understanding and treating MS.

Updated: 31 March 2023

Updated: 16 November, 2021

Stages of the research process

Fundamental laboratory

Laboratory research that investigates scientific theories behind the possible causes, disease progression, ways to diagnose and better treat MS.

Lab to clinic timeline: 10+ years

Research that builds on fundamental scientific research to develop new therapies, medical procedures or diagnostics and advances it closer to the clinic.

Lab to clinic timeline: 5+ years
Clinical Studies
and Clinical Trials

Clinical research is the culmination of fundamental and translational research turning those research discoveries into treatments and interventions for people with MS.

Lab to clinic timeline: 1-5 years


Grant Awarded

  • Incubator Grant

Total Funding

  • $15,000


  • 1 year

Read More

Newsletter subscription

  • Enter your details

Artificial intelligence analysis of brain imaging in MS