Stimulated Raman histologic images of diffuse astrocytoma (left) and meningioma (right) © Daniel Orringer, NYU Langone Health

AI trained to spot brain tumours faster than humans

Artificial Intelligence-based diagnosis was 94.6 per cent accurate, compared to 93.9 per cent for human doctors.

Scientists at New York University have developed an artificial intelligence (AI) that can diagnose brain tumours faster and more accurately than human doctors.

Advertisement

First, the researchers used an advanced form of imaging called stimulated Raman histology (SRH) that uses lasers to highlight areas of the brain that would not usually be visible in a scan. These images were then processed and analysed by an artificially intelligent system, which generated an accurate brain tumour diagnosis in less than 150 seconds.

According to the research, which was published in the journal Nature Medicine, the AI-based diagnosis was 94.6 per cent accurate, compared to 93.9 per cent for human doctors.

“As surgeons, we’re limited to acting on what we can see; this technology allows us to see what would otherwise be invisible, to improve speed and accuracy in the OR [operating room], and reduce the risk of misdiagnosis,” said senior author Dr Daniel A Orringer, associate professor of neurosurgery at NYU Grossman School of Medicine, who co-led the study. “With this imaging technology, cancer operations are safer and more effective than ever before.”

Read more about how AI is changing medicine:

To create the AI tool, researchers trained a type of neural network with 2.5 million SRH images to classify tissue into 13 categories that represented the most common types of brain tumour.

Next, brain tumour biopsies were taken from patients during operations and analysed using the new technique. As a control, other samples were sent for analysis by the pathology lab – the standard practice – which takes 20 to 30 minutes.

Advertisement

When the new technique was used, the whole process could be completed while the operation was taking place, therefore streamlining and improving cancer diagnosis.