An Aging Clock Derived from Images of the Lens of the Eye – Fight Aging!

Recent years have made it clear that any sufficiently large set of data derived from biochemistry or physiology can be fed into a machine learning process to develop an weighted combination of measures that reflects biological age. These aging clocks have been derived from many forms of omics data, from frequently measured blood biomarkers, from various other combination of common measures of health. Interestingly, photography of the face and, separately, of retinal structure also provide enough data for the development of clocks. In today’s open access paper, researchers report on a clock developed from photography of the lens of the eye.

All of these clocks are discovered, not designed. Thus once a given weighted combination of measures is in hand, the next, harder question is how exactly it relates to the underlying processes of aging. This is important because we want clocks that work well to assess the effects of novel interventions and potential interventions on the state of biological age. It we can’t be certain that a clock will, say, accurately reflect the contribution of senescent cells to aging, then one can’t trust that clock in testing the effects of senolytic therapies that clear senescent cells. One would have to calibrate the clocks against the therapy in life span studies, which somewhat defeats the point of having a clock in the first place. The development of sufficient data and understanding to circumvent this issue is the primary challenge in the ongoing development of aging clocks.

LensAge index as a deep learning-based biological age for self-monitoring the risks of age-related diseases and mortality

Assessing an individual’s aging process is important to evaluate one’s health status. As one ages, the human body becomes frail with regard to biological functions and the occurrence of chronic diseases, such as Alzheimer’s disease, cancer, diabetes, and cardiovascular diseases. Chronological age is defined as the time that an individual has experienced since birth. Since aging involves complex determinants, including genetic regulation, and the nutritional and environmental factors, peers with the same chronological age vary in aging and may have different health status and life expectancy. Thus, chronological age does not precisely reveal the true physiological age of individuals.

Biological age assessment based on various physiological biomarkers can quantitatively evaluate the degree of aging and predict the mortality and incidence of age-related diseases more accurately than chronological age. However, measuring biological age is challenging, largely due to obstacles in sample collection, variable aging rates of different tissues, and insufficient reliability of measuring tools and protocols. Intensive investigations of the biological indicators reflecting the overall aging pace of the human body are currently underway. For example, invasive methods measuring telomere length and DNA methylation status, profiling transcriptomics and proteomics, and the inflammatory aging clock have been used to generate biomarkers of aging at the molecular level using human blood cells. Furthermore, noninvasive techniques using machine learning and medical imaging, such as chest X-ray, magnetic resonance imaging (MRI) of the brain, and 3D facial imaging, were introduced to evaluate biological aging. However, these techniques are limited by high costs or instability in clinical practice. Therefore, a more objective, reliable, convenient, and noninvasive method that can accurately evaluate the biological age of an individual has yet to be developed for broader applications and self-management of health status.

The human lens, located in the anterior segment of the eye, is transparent under normal conditions and exchanges substances with the vitreous through the aqueous humor cycle. Age-dependent changes in the lens include nucleus enlargement, elasticity reduction, and increased opacity, all of which can be objectively and reliably observed through noninvasive imaging and rapidly assessed using digital photography. Thus, the human lens appears to be an optimal tissue with unique advantages for assessing biological age.

In this study, we used informative lens photographs to generate LensAge as an innovative indicator to reveal aging status of lens based on deep learning (DL) models. Under ideal physiological conditions (both genetic and environmental), biological age should be synchronized with chronological age. While in reality, there are almost always differences between biological age and chronological age, which is considered to result from individually different aging processes. Therefore, we measured the difference between LensAge and chronological age as the LensAge index to assess an individual’s aging rate relative to peers, and investigated its ability to evaluate the risks of age-related disease occurrence and all-cause mortality. Importantly, we tested whether our models can be generalized to smartphone-based lens photographs, which may have potential applications for self-monitoring the risks of age-related diseases and mortality during aging.