AI model predicts cancer patient mortality risk from facial images

US researchers have developed an artificial intelligence model capable of analyzing facial images to estimate biological age and predict mortality risk in cancer patients, potentially helping doctors select the most appropriate treatments for each individual.

According to researchers at Mass General Brigham, the model, named FAHR-Face, was trained on more than 40 million facial images. Two tools were developed using this model: one estimates biological age with greater accuracy than previous algorithms, while the other predicts survival outcomes for cancer patients.

The study found that the difference between estimated biological age and actual age is linked to survival outcomes. Using both tools together improved prediction accuracy compared to using either tool alone.

Researchers explained that the two tools analyze different facial regions: the first focuses on the folds of the nose, lips, forehead, and temples to estimate biological age, while the second relies on the area below the eyes and around the bridge of the nose to predict mortality risk.

Radiation oncologist Raymond Mak and co-researcher Hugo Aerts noted that this technology could help physicians assess patients' health more precisely and choose treatments suited to their ability to tolerate intensive interventions such as surgery, radiotherapy, and medication.

The researchers emphasized that the model is intended to support, not replace, clinical assessment, and stressed the need for further studies and validation before it is widely adopted in medical practice.