Chapter 1: Introduction to Facial Recognition Technology

My eldest daughter is spending a few weeks at sleepaway camp, where a photographer captures moments throughout the day, uploading them to a private Facebook group. In previous years, I would have spent considerable time online, checking for photos featuring her. However, this year, I simply uploaded her picture to an application that utilizes AI. It identifies her face among a multitude of cheerful campers, allowing me to view only the relevant images. It’s quite remarkable—and a little unsettling.

This development in facial recognition at summer camps signifies how pervasive this technology has become. Today, we will delve into the capabilities of AI in analyzing facial images, as the implications go far beyond mere appearance.

Section 1.1: The Intersection of Facial Analysis and Health

What piques my interest in facial imaging is a recent study published in Cell Metabolism, which introduces a novel aspect to traditional facial analysis—facial temperature.

To comprehend this study, it's crucial to grasp the expanding field of age-related research, which utilizes various biological "clocks" to assess age.

It has been discovered that age is more than just a number. Our cellular and biochemical markers can present varying age estimates. These "clocks" may offer more insightful predictions than the calendar age we typically rely upon.

Numerous molecular indicators exist—such as telomere length—that correlate with age and may even surpass calendar age in forecasting age-related health issues. While assessing telomere length usually requires a blood sample and can be expensive, other methods, including photographs, could serve as alternatives.

We often make age assessments based on visual cues when meeting new people or, in my case, when examining a new patient. It’s not uncommon to note that someone appears younger than their chronological age. Just look at Sharon Stone—66 years old and still turning heads! (Though to be fair, she may have some help.)

Section 1.2: Advancements in Facial Age Prediction

Returning to the Cell Metabolism study, the researchers employed multiple algorithms to derive an "age" from facial photographs.

The initial algorithm was straightforward. Researchers gathered 2,811 images of Han Chinese individuals aged 20 to 90 years and created a 3D facial model from these images. They then trained a convolutional neural network to estimate individuals' ages from the photos, achieving impressive accuracy.

In this era of AI, such results may not seem extraordinary. A quick online search reveals numerous applications claiming to estimate age from photos. When I uploaded a rather unflattering image of myself to ChatGPT, it initially hesitated but eventually estimated my age to be between 35 and 45—I'll take that as a win!

However, the Cell Metabolism study delves even deeper. Researchers employed thermal imaging to capture temperature readings at 54 distinct facial points, adding a new layer to facial analysis.

This is where the intrigue intensifies. While visible facial features can be altered through cosmetics, expressions, or cosmetic surgery, temperature is a more challenging variable to manipulate.

Interestingly, the temperature patterns of our faces evolve as we age, such as a noticeable cooling effect on the nose and cheeks.

Researchers found that by combining this thermal data, they could estimate an individual’s chronological age with reasonable accuracy, although not as precisely as the visual analysis alone.

Chapter 2: Implications of Facial Temperature Analysis

However, predicting age isn't the most fascinating aspect of facial thermal imaging. The real potential lies in predicting metabolic states. Participants in the study underwent comprehensive metabolic assessments and detailed lifestyle evaluations, allowing facial images to inform predictions about their health.

For instance, the 3D reconstructions could indicate dietary habits, revealing that seafood consumers often look younger than their actual age, while those who primarily eat poultry and meat tend to appear older. Thermal imaging could also suggest sleep patterns, indicating that more sleep correlates with a younger appearance.

Furthermore, facial temperature patterns could highlight individuals with higher body mass index (BMI), elevated blood pressure, or increased fasting glucose levels.

The researchers utilized the difference between predicted and actual ages as an illness metric, revealing that individuals with conditions such as hypertension, diabetes, or liver cysts appeared "older" based on facial temperature.

The study even explored facial temperature as a form of biofeedback. In a small trial, researchers measured the age discrepancy between facial temperature and actual age before and after a two-week jump rope regimen. Remarkably, participants appeared about five years younger post-exercise, as assessed by thermal imaging.

While this technology is undeniably fascinating, it's important to note that widespread facial temperature assessments are not imminent. What this research underscores is the wealth of information we hold that can be deciphered by those equipped with the right tools. Whether it’s from our facial wrinkles, the curvature of our smiles, or even the speed at which we type, this data has always existed, but we lacked the means to analyze it effectively—until now.

As a child, I was captivated by Star Trek, particularly the iconic tricorder—a device that could reveal everything about a person's health in mere seconds from a distance. I once envisioned a future where medicine would evolve similarly. While my medical education dispelled that notion, the advent of AI and data analytics suggests that the tricorder era may not be as far off as I once believed.

A version of this commentary first appeared on Medscape.com.