AI Cancer Prognosis: Predicting Patient Survival with New Tools

How does AI mortality risk prediction work?

Researchers have created an AI system that predicts cancer survival odds and mortality risk through a single facial photograph. According to research published in The Lancet Digital Health on October 19, 2026, this tool aims to identify patients who are most likely to pass away within 90 days.

A separate study published in BMC Cancer on October 5, 2026, also details a machine learning model designed to predict general and cancer-specific survival for patients who underwent surgery for gastrointestinal stromal tumors. Investigators at Tianjin Union Medical Center and Nankai University trained this model using data from 3,511 patients identified from the Surveillance, Epidemiology, and End Results database.

The researchers split this cohort into a training set and an internal validation set at a ratio of seven to three, while also using an external validation cohort of 219 patients from the Tianjin facility. To refine their predictions, the team applied three distinct feature selection methods to distinguish between variables relevant to death from any cause and death directly linked to the tumor.

Limitations of AI cancer prognosis models

To learn how to assess patient health, the AI model was trained on more than 40 million general face images and photographs of over 30,000 cancer patients, many of whom have already passed away. These findings were released in The Lancet Digital Health.

Raymond Mak, a radiation oncologist at the Brigham and Women’s Cancer Institute at Massachusetts General Hospital, explains that the algorithm specifically analyzes the bridge of the nose and the area under the eyes. He notes that this focus mirrors ancient observations by Hippocrates regarding external signs that a patient is nearing the end of their life.

Who can this technology be used for?

The model is designed for broad application across diverse populations. According to reports from sg.headtopics.com and inspire2rise.com, the tool functions effectively across different ages, races, genders, and cancer types.

It also includes an age estimation feature. The technology determines a person's age with a margin of error of about five years.

What is the goal of this AI prediction?

The goal of the tool is to prevent over-treatment by providing a sobering second opinion for terminal patients. Raymond Mak, a radiation oncologist at Mass General Brigham Cancer Institute, explains that doctors are generally not very good at estimating end-of-life expectancy.

The model assigns risk scores to patients, and those with the highest scores were more likely to die within 90 days. Mak notes that the AI captures physiological reserve and mortality risk from a photo to provide a sobering second opinion that prevents over-treatment. This approach is not intended to deny treatment or give up on patients.

Is this AI tool ready for clinical use?

Experts caution that this technology is not yet ready for clinical use. Miroslaw Bober, a professor of video processing and machine learning at the University of Surrey, warns against rushing the tool into clinics.

Bober argues that the current study does not prove the AI provides value beyond what a skilled doctor can see, noting that prediction is not the same as explanation. The tool requires further testing against professional medical judgment and validation across different hospitals and patient groups before it can be used in medical practice.