Moles on human skin

Skin cancer and other diseases detected earlier with AI

Image credit: Dreamstime

Researchers at the University of Waterloo in Canada have developed artificial intelligence (AI) technology to help detect a type of deadly skin cancer earlier.

Melanoma is a common type of skin cancer, which can spread to other parts of the body. It is easily treatable if detected early – often by changes in appearance of moles – but can be deadly if noticed too late.

“There can be a huge lag time before doctors even figure out what is going on with the patient,” said Professor Alexander Wong, a systems design engineer at the University of Waterloo. “Our goal is to shorten that process.”

Currently, consultants carry out visual examinations of moles and other potentially affected areas of skin to decide whether patients require biopsies, which can lead to diagnosis.

The technology uses machine-learning techniques to analyse photographs of the skin, then provides doctors with data of biomarkers of melanoma.

The software is trained on a dataset of tens of thousands of images of patches of skin, and their corresponding levels of eumelanin – a type of natural skin pigmentation – and haemoglobin, a protein in red blood cells. When shown new images of potentially cancerous patches of skin, the machine learning program is able to suggest the quantities of eumelanin and haemoglobin in the patch.

Changes in the concentration and distribution of these natural chemicals are strong indicators of melanoma.

“This could be a very powerful tool for skin cancer clinical decision support,” said Professor Wong. “This more interpretable information there is, the better the decisions are.”

The researchers suggest that the AI system, if deployed to doctors, could save significant healthcare costs by preventing unnecessary biopsies, which also often leave scarring. Cases flagged up by the system as potentially cancerous could be followed up with human inspections and interventions.

The researchers hope that the technology could become available to doctors in 2018.

Introducing AI tools to medical research and practice can provide greater objectivity than human doctors and patients can hope to achieve, and allows for huge amounts of data to be analysed in a short period of time. This reduced the amount of tedious, repetitive work clinicians are expected to do, allowing them to instead focus on aspects of their work that requires human experience and judgement.

Meanwhile, researchers across the Atlantic at McGill University are also harnessed the power of AI to help with earlier diagnosis. Scientists from the Douglas Mental Health University Institute’s Translational Neuroimaging Laboratory are using AI techniques to develop an algorithm to recognise the warning signs of dementia up to two years before its onset, based on a single PET scan of the brain of at-risk patients.

Machine learning tools have also been taught to recognise signs of diabetic eye disease, based on basic photographs.

Recent articles

Info Message

Our sites use cookies to support some functionality, and to collect anonymous user data.

Learn more about IET cookies and how to control them

Close