Researchers from Harvard are utilizing artificial intelligence to examine the relationship between biological age and cancer prognosis, finding that appearing younger than one’s actual age and aging more slowly during treatment may improve survival rates. This study, building on a prior pilot study, underscores the potential of AI and digital facial photos in enhancing cancer screening and treatment outcomes.
The studies, which examine the clinical relevance of differences between biological and chronological age, suggest that the AI tool, FaceAge, could eventually allow for screening through a simple photo upload. This approach might enable doctors to give tailored advice based on biological age, recommending more aggressive treatment for biologically younger patients and less intense treatment for those who are biologically older.
Raymond Mak, a co-author of the studies and an associate professor at Harvard Medical School, questions the reliance on chronological age in medical assessments, proposing that biological age may offer a more accurate individual assessment. The research analyzed three metrics—FaceAge, FaceAge Deviation, and Face Aging Rate—across thousands of cancer patients.
The first study, published in the Journal of the National Cancer Institute, found that 65% of over 24,000 cancer patients appeared older than their chronological age, with a significant correlation between the gap in face age and cancer outcomes. Patients appearing five years younger than their actual age had better outcomes, while those looking 10 years older had worse outcomes.
The second study, featured in Nature Communications, assessed changes in face age over time, discovering that slower face aging is linked to better survival. Researchers evaluated 2,276 cancer patients undergoing radiation therapy, noting that those with a faster face-aging rate had poorer outcomes compared to those with a slower rate.
This second study introduced an updated version of the FaceAge algorithm, described by co-author Hugo Aerts as a “deep learning” tool capable of self-improvement through extensive data training. The algorithm was initially trained on 58,000 photos of known-age individuals and 6,000 images of cancer patients, but the new version incorporates 40 million global face images for enhanced accuracy.
Aerts explained that the algorithm’s flexibility allows it to be fine-tuned for specific cancers or diseases with minimal additional data. The research team is working on refining the tool’s accuracy across different skin types and conditions, recognizing that aging might affect organs in varied ways.
Original Source: news.harvard.edu
