This study builds AI models that score diagnostic screening tests, and that predict screening results, clinical judgment, and life expectancy. Longevity Metrics collects a battery of clinical tests on each participant, in whole or in part, and follows every participant for life.
The sit-to-rise test and the timed walk are scored by hand today, from a person's count. A model scores the same test from video instead. It also measures what no one can count by eye - speed, asymmetry, steadiness - so one capture yields both the original score and additional measurements, intended to enrich the model and strengthen what it predicts.
Every test in a participant's record measures the same body, so the tests are correlated: a test that was performed carries information about one that was not. A model trained across the library learns those relationships and estimates a missing result from the results that are present. Each estimate is checked against records where that part was actually measured, and over decades against death and disease through linkage to the 100-Year Human Aging Study (NCT07563777).
The hypothesis is that the full battery can eventually be predicted across modalities with high accuracy using a few short video clips, replacing most in-person screening. That would let preventive screening reach people and places a physical laboratory cannot. How far the input can be reduced is the question this study exists to answer.
Every model is a physician-reviewed clinical decision aid until it is cleared by the FDA.
Age >= 18
willing to participate in the study
Age < 18 years
Participants enrolled in the 100-Year Human Aging Study (NCT07563777) through the unified consent, receiving any Longevity Metrics health screening service at any fixed, mobile, or remote (online) site, current and future. Participants are self-selected community members of all health statuses; no exclusions based on health status, diagnosis, or population group. This study adds no separate enrollment: all records analyzed and all prospective measurements collected under this protocol arise from participants consented under the associated protocols. Participants are followed for life, with mortality, incident disease, and functional-status outcomes ascertained through linkage to the 100-Year Human Aging Study.
Cohort
Other
AIML scored health screening battery
Diagnostic testVideo and audio-captured physical, cognitive, and observational screening tests scored by frozen, versioned AI models; models automatically score standard tests (sit-to-rise, timed walk, chair stand), predict non-performed results from performed ones, and predict clinical judgment; all outputs are physician-reviewed clinical decision aids, and no model output reaches a participant report without provider confirmation.
Measurement Accuracy, Non-Inferiority to Human Scoring.
Agreement between the model's value and the reference standard, established as non-inferiority to a qualified human scorer where a human reference exists. Gates deployment as a clinical decision aid.
Cross-Prediction of Non-Performed Results and Derived Scores
Accuracy of predicting, from video or from the performed part of a record, both the screening results the participant did not obtain, against the measured value; and the derived scores computed from a complete record, including the Longevity Score, biological age, and estimated age at death, against the value the scoring engine produces from the full measured record.
Prediction of Remaining Life Expectancy
Accuracy of predicting remaining life expectancy from video and library data, against observed mortality through linkage to the 100-Year Human Aging Study.
Prediction of Physician Clinical Judgment
Accuracy of predicting the physician's clinical judgment where no determining measurement exists, against the sealed record. Applies to models that predict a clinical determination rather than a measured value. Where a determining test exists but was not performed, the outcome falls under Primary Outcome 2.
Measurement Accuracy, Superiority to Human Scoring
Conditional on non-inferiority, whether the model scores more reliably or precisely than a qualified human scorer.
Uncertainty Calibration
Agreement between the model's stated confidence and its observed accuracy.
Within-Session Repeatability
Where a test is captured twice back to back, the stability of the model's output across the two captures. Not applicable where repeat capture is impractical, or where a second effort measures fatigue because the test is performed to failure.
Missing-Data and Completeness Robustness
Model performance on partial or incomplete input relative to complete input.
Click on the lead sponsor to view all of their trials.
