Government-backed researchers claim AI cancer breakthrough

Technology developed by UK academics is claimed to be able to detect a likely recurrence of the disease with an accuracy rate 10 points in advance of expert human clinicians

A new artificial intelligence-powered computer system backed by government funding can detect cancer signs with “remarkable” accuracy, its developers claim.

The new system, Histomorphological Phenotype Learning’(HPL), could accelerate cancer diagnosis and help predict patient outcomes, according a team of researchers led by academics from the University of Glasgow and New York University. Researchers from the University College London and the Karolinska Institute also contributed to the paper.

The predictions made by the HPL system correctly assessed the likelihood and timing of returning cancer in existing patients in 72% of instances – a figure which researchers claim is almost 10 points higher than the rate typically achieved by human analysis.

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. The technology uses an algorithm to analyse images and break them down into different tiny tiles, with each representing a section of human tissue. The system then analyses these tiles and, using machine learning, classified any visual features shared across cells.

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%, according to the research team.

Once the algorithm identified patterns in the samples, the researchers used it to analyse links between various phenotypes – physical and biochemical traits of a cell – and the clinical outcomes stored in the database, including how long patients lived after having surgery. 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.

Related content

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.

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.”

He 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.”

The research project was backed by financial support from government bodies including: the Engineering and Physical Sciences Research Council; the Biotechnology and Biological Sciences Research Council; and the National Institutes of Health.

The research paper – 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.