New AI tool could enhance and personalize brain tumor imaging

A new artificial intelligence (AI) tool developed by researchers at UCL and UCLH can analyze scans more quickly than experienced neuroradiologists and provide greater patient-personalized detail to aid treatment.

The research, published in NeuroImage: Clinical, shows that the tool can analyze scans in just three seconds—compared with the typical five minutes needed by experienced neuroradiologists—and it could be used to more accurately predict an individual patient’s treatment outcome than with a doctor alone.

Additionally, a workforce analysis forecast that if deployed nationally, it could save more than £1.5million in NHS costs within the next three years.

Approximately 100,000 people in the U.K. live with brain tumors, who are both diagnosed and monitored following treatment by using scans of their brain.

Just as individual patients are unique, so too are the imaged appearances of their tumors. Current NHS care for brain tumors addresses these differences by having specialist imaging doctors (radiologists) analyze the images to assess and describe where the tumor is, its approximate size, and its proximity to vital healthy brain structures.

The new research shows that the AI-based tool can perform this analysis effectively and efficiently. In a large group of 1,172 patients, the UCL Queen Square Institute of Neurology and UCLH team also showed the tool is accurate across patients of all ages and sexes.

Lead author Dr. James Ruffle (UCL Queen Square Institute of Neurology) said, “Given that the imaging appearances of an individual’s brain tumor vary greatly from one patient to another, artificial intelligence technologies provide an innovative solution to enhance health care workers’ data-driven decision making, improving and personalizing care for each individual affected, and at near-zero additional cost to the NHS—and with savings to the NHS over the medium term.”

Senior author, Dr. Harpreet Hyare, Lead Teenage and Young Adult Neuroradiologist at UCLH, said, “This work shows real promise in enabling neuroradiologists to provide accurate and quantified descriptors of brain tumors.

“By providing objective quantitative assessments of different brain tumor components that can be monitored over time, we can enable clinicians to plan treatment more effectively at critical time points.”

The AI-based imaging tool is one of several the team are currently developing specifically for neuro-oncology applications that they hope to bring into the clinical arena in the near future.