Harvard study finds AI tool accuracy better than traditional methods to predict paediatric cancer relapse

A team of Harvard researchers have found that an AI tool does a better job of predicting a relapse risk in paediatric cancer patients with far better accuracy than the traditional methods. 

The research collected nearly 4000 MR (Magnetic Resonance) scans from 715 paediatric patients. Experts then employed a technique called temporal learning to analyse the scan reports. 

“Many paediatric gliomas are curable with surgery alone, but when relapses occur, they can be devastating. It is very difficult to predict who may be at risk of recurrence, so patients undergo frequent follow-up with magnetic resonance imaging for many years, a process that can be stressful and burdensome for children and families. We need better tools to identify early which patients are at the highest risk of recurrence," said  Benjamin Kann, Assistant Professor of Radiation Oncology at Harvard Medical School.

What is temporal learning?

The AI model is trained to understand the findings from multiple brain scans taken over a period of several months post-surgery. What makes the study different from other AI models is the use of the temporal learning model, because typically AI models that are used for medical imaging are trained to conclude single scans. With the temporal learning, images acquired over time inform the algorithm’s prediction of cancer relapse. 

Researchers concluded that with the introduction of the new technique, the predictions of either low- or high-grade glioma by one year post-treatment, with an accuracy of 75-89 per cent — substantially better than the accuracy associated with predictions based on single images, which they found to be roughly 50 per cent. 

“We have shown that AI is capable of effectively analyzing and making predictions from multiple images, not just single scans. This technique may be applied in many settings where patients get serial, longitudinal imaging, and we’re excited to see what this project will inspire," said first author Divyanshu Tak of the AIM Program at Mass General Brigham and the Department of Radiation Oncology at the Brigham.

The team has also cautioned that further validation across additional settings is important before clinical application. The results of the study were published in The New England Journal of Medicine AI.

Health