AI twice as accurate as a biopsy when grading severity of sarcomas

The new AI model could potentially be applied to other cancers

New research from the Institute of Cancer Research (ICR) and The Royal Marsden NHS Foundation Trust has revealed that artificial intelligence (AI) could be twice as accurate at grading the aggressiveness of some sarcomas as a biopsy.

The new AI algorithm could speed up the diagnosis of sarcoma patients and could potentially be applied to other cancers.

A biopsy, which is currently standard practice, is an invasive medical procedure that involves taking a small sample of body tissue for examination.

Around 4,295 people in England are diagnosed every year with soft tissue sarcomas, a rare type of cancer that develops in the connective tissues in the body, including fat, muscles and blood.

With over 50 forms of the condition, the RADSARC-R study analysed the CT scans of 170 patients treated at The Royal Marsden, with two common forms of a rare soft tissue sarcoma that develops in the back of the abdomen: retroperitoneal sarcoma – leiomyosarcoma and liposarcoma.

After producing an AI algorithm, researchers tested it on 90 patients across Europe and the US, using a technique called radiomics to extract information about patients’ diseases from medical images.

Results showed that the model accurately graded the aggressiveness of 82% of the tumours, compared to 44% which were accurately graded using biopsies, and accurately predicted the disease type of 84% of the sarcomas tested, compared to radiologists who could not diagnose 35% of the cases.

As well as speeding up diagnosis, the AI technology could improve the management of the disease and its outcomes and could spare low-risk patients from unnecessary treatments, scans and hospital visits.

“In the next phase… we will test this model in clinic on patients with potential retroperitoneal sarcomas to see if it can accurately characterise their disease and measure the performance of the technology over time,” said Dr Amani Arthur, clinical research fellow, ICR, and registrar, The Royal Marsden NHS Foundation Trust.