Let us know about free updates
Scientists use artificial intelligence analysis of the facial features of cancer patients to predict survival outcomes, and in some cases exceed the clinician’s short-term life expectancy forecast.
Researchers used deep learning algorithms to measure subjects’ biological age and found that cancer patients’ characteristics were on average about five years older than the time series.
A new technical tool known as Face is part of the driving force for the increase using estimates of body organ aging as a so-called biomarker of potential disease risk. Advances in AI have boosted these efforts because of their ability to learn from large health data sets and make risk predictions based on them.
The study said information derived from facial photographs could be “clinically meaningful,” said Hugo Aerts, co-author of a paper on the study published Thursday in Lancet Digital Health.
“The study shows that simple selfie-like photographs contain important information that helps inform patients and clinicians’ clinical decision-making and care plans,” says Aerts, AI director of medicine at Mass General Brigham in Massachusetts.
“It really matters how old someone looks compared to chronological order. Individuals with faces younger than chronological order are significantly better after cancer therapy,” he added.
Scientists trained 58,851 photographs of healthy people estimated from the public dataset. The algorithms of 6,196 cancer patients were then tested using photographs taken at the start of radiation therapy.
Among cancer patients, older ages result in worse survival outcomes, even after adjusting for age, gender, and cancer type. This effect was particularly pronounced for people over 85 years of age.
Scientists then asked 10 clinicians and researchers to predict whether patients receiving palliative radiation therapy for advanced cancer would be alive in six months. Human raters were around 61% of the time when only patients were able to access photos, but that improved to 80% when they underwent face agitation analysis.
Possible face limitations include data bias and measurements that reflect model errors rather than actual differences between time series and biological age, the researchers said.
Scientists are now testing a wider range of patient technologies and assessing their ability to predict disease, general health and life expectancy.
Research into biomarkers for aging is the subject of intense research activities. In February, scientists announced a simple blood test to detect how quickly the organs age and help flag the increased risk of 30 diseases, including lung cancer.
Recommended
Facial aging is a field of growing interest, and scientists are exploring a variety of techniques. One is the concept of perception of aging. So it’s not just how old you are biologically, but how old you look to an experienced medical professional.
Perceived aging has emerged as a potential predictor of mortality and several age-related diseases, researchers say. The drawback is that data is generated through human observation, which is time consuming and expensive.
Faith’s assessment was thought to be “very thorough,” says Jaume Bacardit, an AI specialist at Newcastle University, who has applied technology to recognize aging.
However, he added that further explanations need to be made about how AI technology worked to check for potential distortion factors.
“So, which part of the face is based on predictions?” Bacardit said. “This will help identify potential confounding factors that may otherwise be undetected.”
