Artificial Intelligence-Based Health Assessment From Face Photographs - Healthy Volunteer Study

Overview

The investigators aim to examine how FaceAge estimates (a measure of biological age based on facial features) change across different time points and identify factors that may influence these changes. This will help the investigators understand the consistency and reliability of the FaceAge algorithm and allow them to make improvements.

Historical background The concept of biological age, as distinct from chronological age, has gained increasing attention in recent years. Biological age aims to quantify the physiological state of an individual, which can be influenced by genetic factors, lifestyle choices, and environmental exposures1-4. With advancements in artificial intelligence, new tools have emerged to estimate biological age and an individual's health from various imaging biomarkers, including facial features. One such tool is FaceAge, a deep learning system developed to estimate biological age from a single frontal face photograph5. FaceAge represents a step forward in quantifying biological age and an individual's health status, offering a non-invasive and potentially widely applicable method for assessing an individual's aging process. Various disease states and overall health conditions can impact a person's physical appearance. Additionally, research has established the existence of various biological rhythms in human physiology. These rhythms regulate numerous physiological functions and can influence an individual's appearance and well-being6. It is important to note that different photo settings, such as lighting conditions, facial expressions, and image quality, may lead to variations in AI algorithm performance, such as FaceAge estimates, highlighting the need for a comprehensive investigation of these factors to ensure the robustness and reliability of the algorithm. Previous pre-clinical or clinical studies leading up to and supporting the proposed research The first FaceAge study demonstrated the algorithm's effectiveness in estimating biological age across multiple clinical cohorts5. Trained on a dataset of 58,851 healthy individuals, FaceAge showed that cancer patients, on average, appeared approximately five years older than their chronological age. Moreover, FaceAge estimates were associated with survival outcomes and showed correlations with molecular mechanisms of senescence through gene analysis. Previous research has established that various factors, including health status, lifestyle, and environmental influences, can affect an individual's appearance7,8. Studies have shown that these factors can lead to variations in physical characteristics that may be captured by facial analysis tools5,9. The rationale behind the proposed research and potential benefits to patients and society Despite growing evidence supporting the impact of various factors on human physiology and appearance, there is limited research on how these variations might affect estimated biological age from facial features and the validation of AI algorithms like FaceAge require datasets that tests the biomarkers reproducibility, robustness across different imaging conditions and generalizability. The proposed study addresses this knowledge gap by collecting photographs from volunteers over time. By compiling a database of numerous photos per person, the investigators will generate a resource to test the reproducibility of AI algorithms developed to generate biomarkers from photographs. Additionally, this dataset will provide a resource to assess whether AI predictions such as FaceAge are consistent across individuals, disease type, and time periods. Furthermore, the investigators can rigorously test the model's reproducibility, generalizability and robustness across diverse real-world scenarios by including photos taken under various conditions and from different time points. This research could provide valuable insights into the reliability and variability of AI algorithms such as FaceAge estimates contribute to the understanding of how various factors may influence human appearance and perceived age and health. If major variations are observed, it could have important implications for using photographs in medical assessments and research studies utilizing FaceAge or similar algorithms. Understanding these potential variations could lead to more accurate and standardized use of AI-based assessments of health from photographs including facial age estimation tools in clinical settings, potentially improving their prognostic value and applicability in personalized medicine. Moreover, this research may open new avenues for studying the intersection of various health factors and aging, potentially leading to novel interventions to promote healthy aging and improve overall health outcomes.

Conditions

Eligibility

Locations (1)

Outcomes

Central Contacts