AI algorithm more accurate than radiologists at detecting a rare cancer, study shows

AI algorithm more accurate than radiologists at detecting a rare cancer, study shows

Artificial intelligence (AI) tools are increasingly being studied to help health professionals with diagnosis.

Researchers from The Royal Marsden NHS Foundation Trust and The Institute of Cancer Research have now used AI to improve diagnoses of a rare cancer.

They found that AI could be twice as accurate as a biopsy - an invasive procedure where a sample is taken using a needle - at grading the risk of some sarcomas, according to the study published in The Lancet Oncology.

Sarcomas are a rare form of cancer with roughly 5,300 cases in the UK per year. This study focused on retroperitoneal sarcomas, soft tissue sarcomas located deep in the abdomen and pelvis, that can be challenging to biopsy.

“Through this early research, we’ve developed an innovative AI tool using imaging data that could help us more accurately and quickly identify the type and grade of retroperitoneal sarcomas than current methods,” said Dr Amani Arthur, the study’s first author, in a statement.

How can AI help with diagnosis?

The researchers used a method called radiomics, which uses computers to extract quantitative information from medical images such as X-rays, MRIs and CT scans.

These features are then used to diagnose diseases or predict how a disease will progress. By using algorithms, radiomics can be used to analyse details that could not be spotted with the naked eye.

Researchers used the CT scans of 170 patients treated at The Royal Marsden to create an AI algorithm, they explained in a statement.

Then, the algorithm was tested on 89 patients from centres across Europe and the US.

The model was able to accurately predict how aggressive a tumour was likely to be for 82 per cent of them, while only 44 per cent of tumours were correctly graded using a biopsy.

The model was also able to accurately predict the type of sarcoma for 84 per cent.

The reporting radiologist was "only able to correctly diagnose 73 per cent of liposarcoma [one form of sarcoma] and 43 per cent of leiomyosarcoma," the study authors said.

Retroperitoneal sarcomas are especially difficult to diagnose because they often have common symptoms like abdominal pain, bloating, or a decrease in appetite.

They can also grow considerably before the appearance of the symptoms, experts say.

Hope for other cancers

“The disease is very rare – clinicians may only see one or two cases in their career – which means diagnosis can be slow. This type of sarcoma is also difficult to treat as it can grow to large sizes and, due to the tumour’s location in the abdomen, involve complex surgery,” said Arthur.

“[This AI tool] could improve patient outcomes by helping to speed up diagnosis of the disease, and better tailor treatment by reliably identifying the risk of each patient’s disease,” Arthur added.

The team also hope to refine the algorithm so it can be used for other types of cancer.