Medicine of the future: with AI, the algorithm under the skin

LibeCare. Think health tomorrow: survey

LibéCARE. Think health tomorrowcase

Thanks to American scientific advances in artificial intelligence, particularly in the field of medical imaging, doctors could predict distant diseases using data-fed computers.

How to take good care of yourself, live well, age well? Come to Caen on December 9 and 10 at the MoHo with LibéCare to debate with doctors, intellectuals and experts. While waiting for the event, carried out in partnership with the Normandy region, MGEN and ADMD, Release will publish in a dedicated space articles, forums and testimonials.

A single chest x-ray is enough to predict the future. It is not submitted to the radiologist, but to the computer: the software analyzes it with its expert eye and announces the risk for the patient of having a heart attack in the next ten years… This kind of scenario still looked like a pure science fiction a few years ago, but it becomes perfectly real, and even banal, thanks to artificial intelligence. Algorithms are making giant leaps in medicine, particularly in the field of medical imaging where they excel. Some software detects retinopathy as well as an ophthalmologist, others find suspicious masses in mammograms or microcracks in wrist x-rays…

But this Tuesday, November 29, it was therefore heart attacks that were at the heart of the discussions at the annual meeting of the Radiological Society of North America. It presented the results of a project carried out at Massachusetts General Hospital, attached to the Harvard Medical School, with this new algorithm capable of predicting the risk of stroke. It is not so obvious for the radiologist because there are several organs to examine and several factors to combine. We can notice “if the heart is very large or has a malformation”, as American cardiologist Nicole Weinberg explains to the media Medical News Todaybut also “see if the aorta looks enlarged or if there are calcium deposits, which show up on the x-ray. You can also see in the lung tissue if there is an accumulation of fluid […]which may indicate heart failure.

“Unimaginable volume of data”

To scrutinize all these places at once and know what to conclude, the algorithm was trained by a machine learning process. We “fed” him by showing him more than 147,000 chest X-rays from 40,000 patients, informing him for each picture about the existence – or not – of a cardiovascular pathology. After ingesting this mountain of examples, which represent the equivalent of a radiologist’s entire career, the software knows how to manage on its own. It spots clues in images and makes reliable predictions. “The beauty of this approach is that it only requires an X-ray image, of which millions are made daily around the world,” rejoices radiologist Jakob Weiss, lead author of the Massachusetts General Hospital study. “A single image can predict future episodes of major cardiovascular events with the same efficiency as current clinical assessments,” based on a whole list of criteria such as age, sex, hypertension, diabetes, tobacco consumption, blood tests…

Artificial intelligence has gigantic potential for such distant medical predictions. “To really understand a patient’s long-term needs, and not just their transitory problems, requires having an unimaginable volume of data: the genome, demographic data, medical history, environmental factors… details in the magazine Forbes Tony Ambrozie, chief information officer for the Baptist Health hospital network in Florida. Realistically, it is impossible for healthcare professionals to do these analyzes manually. Machine learning is evolving towards solutions that perform this vast data processing automatically, to help practitioners develop safe and personalized care pathways for their patients.”

Digital copy of the brain

In France, the Aramis project is developed at the National Institute for Research in Computer Science and Automation to test a computer model of the human brain that combines several sources of data acquired on patients monitored over time: genetics, medical imaging and clinical data (from neurological tests, for example). We thus recreate a kind of digital copy of their brain, a brain of “virtual patient” that can be computer-aged to see what diseases it is likely to develop. The hope is that in a few years, the algorithm will be sufficiently trained to predict a scenario of brain aging from a few data on a patient, and estimate for example what are their chances of developing Alzheimer’s disease on a horizon two, three or five years. Far from dehumanizing medicine, artificial intelligence used as a prognostic aid should, on the contrary, allow earlier and fairer treatment of patients.

We wish to say thanks to the author of this article for this awesome material

Medicine of the future: with AI, the algorithm under the skin

Visit our social media profiles as well as other related pages