In a groundbreaking advancement, researchers have created an artificial intelligence (AI) tool capable of predicting the survival chances of cancer patients based solely on a photograph of their face. This innovative system was developed by experts at Mass General Brigham, a leading nonprofit healthcare organization in Massachusetts, with the aim of estimating biological age and survival outcomes for individuals diagnosed with cancer.
This remarkable project demonstrates the potential of AI in health assessments derived from simple selfies, utilizing a vast pool of over 40 million facial images representing diverse ethnicities and age groups.
The findings were recently published in the Lancet Digital Health journal, illuminating the capabilities of this AI model, which distinguishes between biological age—an appearance-based age—and chronological age to predict survival likelihood among cancer patients.
Doctors have devised two specialized tools using this technology: FAHR-FaceAge and FAHR-FaceSurvival. The former estimates biological age and predicts cancer survival, while the latter assesses survival predictions directly from a facial image, bypassing biological age as a step.
While sharing foundational principles, these algorithms extract different information from the facial image. FAHR-FaceAge analyzes features such as the creases from the nose to the mouth corners and the areas around the forehead and temples, whereas FAHR-FaceSurvival focuses specifically on the region beneath the eyes and around the nasal area.
“FAHR-FaceSurvival represents a significant leap forward in the field of facial health recognition AI by deriving survival risks directly from cancer patients’ facial photographs,” explained Hugo Aerts, a co-senior author of the study. “This is, to our knowledge, the first instance of an AI model reliably predicting cancer survival from a mere photograph.”
In a study involving cancer patients undergoing radiotherapy in the United States, those who appeared five or more years older than their actual age faced a 21% increase in mortality risk over a four-year average follow-up period. Conversely, patients appearing five or more years younger had a 17% lower risk.
The impressive technical advancement lies in how FAHR-Face learns a generalized representation from over 40 million images, allowing it to adapt this knowledge to various health outcomes like biological age and survival probabilities. This foundational model approach significantly enhances the scalability for developing multiple facial health biomarkers, contrasting with traditional models that require independent training.
Research outcomes indicate a strong correlation between biological age estimates and overall physical health, potentially reflecting patients’ capacity to endure specific treatments. This development could revolutionize the way AI-assisted facial analysis aids healthcare professionals in making informed treatment decisions.
Further implications of this tool could include improved predictions of complications or mortality following major surgeries or other cancer treatments. Dr. Raymond Mak, a radiation oncologist at Mass General Brigham, stated, “As the population ages, there is a crucial need for better methods to assess physical reserve and frailty prior to suggesting aggressive treatments, such as radiation, surgery, or systemic therapy.”
“Our aim is to create a more quantitative, consistent, and scalable health assessment process based on facial photographs to support personalized treatment decisions, enhancing rather than replacing physician judgment,” he added.
Researchers also discovered various lifestyle factors affecting facial aging, including heavy alcohol use, drug consumption, and smoking, all of which were linked to an appearance of greater biological age in the AI’s analysis.
Another unexpected result was the consistency of FAHR-FaceAge across diverse real-world photo variations such as pose, lighting, facial expressions, and makeup, demonstrating stability even post-plastic surgery, except for minor adjustments following rhinoplasty.
While the prospect of utilizing single facial photos presents an accessible and cost-effective medium for personalized healthcare decisions, experts emphasize that further research is essential before the widespread application of this tool.
Editor’s Take
This development in AI technology signifies an exciting frontier in personalized medicine, particularly for cancer treatment strategies. By leveraging facial analysis, healthcare providers may soon gain invaluable insights into patient prognosis, potentially leading to more tailored and effective treatment approaches. Nonetheless, careful validation is critical to ensure reliability and efficacy in clinical settings.
Source: www.theguardian.com
