AI System For Blood Test Of Brain Tumor Invented In UK

An artificial intelligence (AI) system that analyzes a drop of blood could soon help doctors identify thousands of patients with brain tumors, British media said.


The technology could be available in the NHS within 18 months, researchers said. An earlier trial showed the technique had a 90 percent success rate in identifying gliomas, the most common brain tumor.

More than 9,000 people are diagnosed with primary brain tumours in the UK each year, about half of them malignant. They often describe ambiguous symptoms to GPS, such as headaches, memory problems and changes in temperament. Brain scans are the only reliable way to identify tumors.

However, most patients who see a gp are not referred for a scan in a timely manner, leading to worsening conditions. Meanwhile, only 2 percent of patients who were sent for a scan ended up with tumors.

The new system involves taking a patient’s blood sample and spinning a centrifuge to separate out whole blood cells, the report said. The remaining liquid is a mixture of various compounds, including proteins, lipids, and components of cells that are broken down during the rotation.


Three drops of the liquid are analysed using an existing technique called infrared spectroscopy to read its chemical composition. An AI system then studies the results, looking for subtle differences in blood composition patterns that are linked to the presence of brain tumors. The system, which runs on a regular laptop, is designed to determine which patients are most at risk and should be scanned first.

In a recent trial, the system analysed blood samples from 400 patients who showed signs of brain tumours and were referred to western general hospital in Edinburgh for scans. Forty of them were found to have tumors. The trial correctly identified 82 percent of brain tumors. The test correctly identified 84 per cent of people who did not have brain tumours, meaning a lower rate of “false positives”.