AI proves twice as accurate as a biopsy at grading some sarcomas


New study from The Royal Marsden and The Institute of Cancer Research suggests new algorithm could have significant impact on grading and differentiating rare cancers

Artificial intelligence (AI) could be around twice as accurate as a biopsy at grading the aggressiveness of some sarcomas, according to new research from The Royal Marsden and The Institute of Cancer Research, London.

Results from the study, published in The Lancet Oncology, suggest that a new AI algorithm could help tailor the treatment of some sarcoma patients more accurately and effectively than a biopsy, an invasive procedure which is currently standard practice. The research also suggests that the technology could help clinicians diagnose subtypes of the rare disease, speeding up diagnosis as a result.

Researchers believe the technique could eventually be applied to other cancer types, potentially benefitting thousands of patients every year.

Soft tissue sarcoma, of which there are over 50 different types, is a rare form of cancer that develops in the body’s connective tissues, including fat, muscles, nerves and blood and lymph vessels. This study focused on retroperitoneal sarcoma, which develops in the back of the abdomen and, due to its location and rarity, is currently hard to diagnose and treat.

Researchers used the CT scans of 170 Royal Marsden patients with the two most common forms of retroperitoneal sarcoma – leiomyosarcoma and liposarcoma – to create an AI algorithm, which was then tested on nearly 90 patients from centres across Europe and the US. A technique called radiomics was used to analyse the CT scan data, which can extract information about the patient’s disease from medical images, including data which can’t be distinguished by the human eye.

The model accurately graded the risk (how aggressive a tumour is likely to be) of 82% of the tumours analysed, while only 44% were correctly graded using a biopsy. The model also accurately predicted the disease type of 84% of the sarcomas tested, meaning it can effectively differentiate between leiomyosarcoma and liposarcoma, compared with radiologists who identified 65% of the cases.

By offering a more accurate and effective way of grading tumours, researchers hope this technology will improve the management of the disease. Once identified, high-risk patients could be given amplified treatment, while low-risk patients could be spared unnecessary treatments, scans and hospital visits. It could also speed up diagnosis of the disease by supporting clinicians in more accurately identifying the subtype.

Study lead Professor Christina Messiou, consultant radiologist at The Royal Marsden Private Care and professor in imaging for personalised oncology at The Institute of Cancer Research, London, said: “This is the largest and most robust study to date that has successfully developed and tested an AI model aimed at improving the diagnosis and grading of retroperitoneal sarcoma using data from CT scans.

“We’re incredibly excited by the potential of this state-of-the-art technology, which could lead to patients having better outcomes through faster diagnosis and more effectively personalised treatment. As patients with retroperitoneal sarcoma are routinely scanned with CT, we hope this tool will eventually be used globally, ensuring that not just specialist centres – who see sarcoma patients every day – can reliably identify and grade the disease.

“In the future, this approach may help characterise other types of cancer, not just retroperitoneal sarcoma. Our novel approach used features specific to this disease, but by refining the algorithm, this technology could one day improve the outcomes of thousands of patients each year.”