Biological Age Prediction Tool from Selfies
Researchers at Mass General Brigham have introduced a deep learning algorithm named Faceage, which provides intriguing insights into our biological age just from photographs of our faces. It’s fascinating to think that a simple selfie could be more than just a snapshot but might actually reveal how quickly we age biologically compared to our chronological age.
What’s more, as detailed in a press release from MGB, Faceage can also estimate survival rates for cancer patients, providing another layer of importance to this technology.
The AI was trained using 58,851 images sourced from public data sets representing healthy individuals. To evaluate its effectiveness, researchers tested it on photos of 6,196 cancer patients taken before they began radiation therapy. Astonishingly, the results indicated that the biological age predicted by the tool was, on average, around five years older than their actual chronological ages.
The accuracy of Faceage was further validated through its ability to forecast the life expectancy of patients receiving palliative care, outperforming predictions made by a panel of clinicians based on the same photographs.
This research, published in Lancet Digital Health, was spearheaded by Dr. Hugo Aerts, who pointed out that this technology signifies a shift in how we can utilize seemingly mundane images for significant clinical insights. He emphasized that such self-portraits could profoundly influence clinical decisions and patient care strategies.
Dr. Aerts remarked, “It genuinely matters how old someone appears compared to their chronological age. In fact, individuals who look younger tend to have better outcomes post-cancer treatment.” The broader ambition here is to mitigate biases that can arise in medical decisions based on how a patient looks or how old they seem.
However, anticipation is tempered with caution; the researchers acknowledged that more work is needed before Faceage can be deployed in clinical settings. Future endeavors aim to extend testing to various hospitals and assess patients at different disease stages, along with exploring its potential in predicting various health outcomes.
Dr. Harvey Castro, an emergency physician and AI expert, shared his perspective on this tool. He noted the potential it has but also the pitfalls. While the promise lies in being able to quantify a physician’s instinctive assessments, he highlighted the critical need for diverse training data to avoid biases in results. Castro warned against the ethical implications, stating that questions about data ownership, patient understanding of their information, and privacy are paramount as we tread carefully into this new territory.
In this rapidly evolving realm of AI, it remains clear that while technologies like Faceage can enhance medical care, they cannot replicate the empathy, context, and human touch that are vital components of effective healthcare.
