AI-powered system could transform cancer care

A new AI-powered computer system can detect cancer signs with “remarkable” accuracy, its developers claim.

The new system, called ‘Histomorphological Phenotype Learning’ (HPL), could accelerate cancer diagnosis and help predict patient outcomes.

A team led by researchers from the University of Glasgow and New York University is behind the discovery.

The predictions made by the HPL system correctly assessed the likelihood and timing of cancer’s return 72 per cent of the time - almost 10 per cent more accurate than human counterparts.

Professor John Le Quesne from Glasgow University, co-senior author of the paper, said: “It takes many years to train human pathologists to identify the cancer subtypes they examine under the microscope and draw conclusions about the most likely outcomes for patients. It’s a difficult, time-consuming job and even highly-trained experts can sometimes draw different conclusions from the same slide.”

To trial the system, researchers used images from more than 450 samples of lung adenocarcinoma stored in the United States National Cancer Institute’s Cancer Genome Atlas database.

Lung adenocarcinoma is among the top three most common types of lung cancers in the UK.

The system first used an algorithm to analyse the images and break them down into different tiny tiles, with each representing a section of human tissue.

It later analysed the tiles and learned to classify any visual features shared across the cells in each tile.

When the team added analysis of slides from squamous cell lung cancer to the HPL system, it was capable of correctly distinguish between their features with 99 per cent accuracy.

Le Quesne added: “It could prove to be an invaluable tool to aid pathologists in the future, augmenting their existing skills with an entirely unbiased second opinion. The insight provided by human expertise and AI analysis working together could provide faster, more accurate cancer diagnoses and evaluations of patients’ likely outcomes. That, in turn, could help improve monitoring and better-tailored care across each patient’s treatment.”

Once the algorithm identified patterns in the samples, the researchers used it to analyse links between the phenotypes and the clinical outcomes stored in the database, including how long patients lived after having surgery.

Phenotypes are the physical and biochemical traits of a cell.

The system found some phenotypes, like less invasive tumour cells, were more common in those who lived longer after treatment, while others like aggressive tumour cells forming solid masses, were more closely linked with the recurrence of tumours.

Similar levels of accuracy were found when the research was expanded to include analysis of thousands of slides across 10 other types of cancers, including breast, prostate and bladder cancers, the results were similarly accurate.

Adalberto Claudio Quiros, co-first author of the paper, said: “This kind of self-learning algorithm will only become more accurate as additional data is added, helping it become more fluent in the language of cancer. Unlike humans, it brings no preconceived ideas to its work, so it may even find patterns across the datasets that haven’t been fully explored before.

“Ultimately, our aim is to provide doctors and patients with a tool that can help provide them with an improved understanding of their prognosis and treatment.”

Researchers from the University College London and the Karolinska Institute also contributed to the paper.

The team’s paper, titled ‘Mapping the landscape of histomorphological cancer phenotypes using self-supervised learning on unlabeled, unannotated pathology slides’, has been published in the scientific journal Nature Communications.

The Engineering and Physical Sciences Research Council (EPSRC), the Biotechnology and Biological Sciences Research Council (BBSRC), and the National Institutes of Health funded the research.