Scholarlink

Introduction <p>Facial hyperpigmentation is a primary marker of skin aging in Asian populations. While many artificial intelligence (AI)-based aging simulators exist on social media, they often lack scientific transparency and dermatological validation. This study introduces and validates a novel facial aging simulator specifically engineered to provide personalized, evidence-based predictions of pigmentary spot progression according to varying levels of ultraviolet (UV) exposure and photoprotection.</p> Methods <p>The simulator utilizes a dual-tool framework: Elicitation Based Aging Simulator (EBAS) and AgingMapGAN (AMGAN). EBAS leverages the collective expertise of 28 dermatologists via a structured Delphi process and causal Bayesian belief networks (BBNs) to model skin aging trajectories. AMGAN, a conditional generative adversarial network, provides high-resolution visual representations of these trajectories. The model was trained on a standardized dataset of 600 individuals and focused on the density of pigmentary spots (DPS) on the cheek. Model performance was benchmarked against expert clinical grading using correlation metrics.</p> Results <p>The simulator demonstrated high accuracy in replicating expert logic (Pearson’s correlation 0.96). In a case study of a 38-year-old female of Chinese descent, the model predicted the 15-year probability of a clinically significant progression (a two-grade increase in DPS). In the absence of photoprotection, the probability of reaching this elevated grade by age 53 was 71.35%, whereas regular daily application of SPF 50+ sunscreen reduced this probability to 39.24%.</p> Conclusions <p>This framework represents the application of a scientifically validated image-generation tool for predicting age-related hyperpigmentation based on individual exposome factors. By integrating dermatological knowledge, this simulator provides a robust educational and research resource for personalized skincare strategies. The findings confirm that consistent photoprotection significantly mitigates the 15-year trajectory of premature facial aging, offering a scientifically validated foundation for public health communication on prevention.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting Age-Related Facial Hyperpigmentation via Dermatologist Knowledge Elicitation and Generative Modeling

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月20日

查看全文
  • Edouard Raynaud,
  • Laudine Bertrand,
  • Frederic Flament,
  • Jennifer Bourland,
  • Emmanuelle Tancrède-Bohin,
  • Tao Li,
  • Hussein Jouni

摘要

Introduction

Facial hyperpigmentation is a primary marker of skin aging in Asian populations. While many artificial intelligence (AI)-based aging simulators exist on social media, they often lack scientific transparency and dermatological validation. This study introduces and validates a novel facial aging simulator specifically engineered to provide personalized, evidence-based predictions of pigmentary spot progression according to varying levels of ultraviolet (UV) exposure and photoprotection.

Methods

The simulator utilizes a dual-tool framework: Elicitation Based Aging Simulator (EBAS) and AgingMapGAN (AMGAN). EBAS leverages the collective expertise of 28 dermatologists via a structured Delphi process and causal Bayesian belief networks (BBNs) to model skin aging trajectories. AMGAN, a conditional generative adversarial network, provides high-resolution visual representations of these trajectories. The model was trained on a standardized dataset of 600 individuals and focused on the density of pigmentary spots (DPS) on the cheek. Model performance was benchmarked against expert clinical grading using correlation metrics.

Results

The simulator demonstrated high accuracy in replicating expert logic (Pearson’s correlation 0.96). In a case study of a 38-year-old female of Chinese descent, the model predicted the 15-year probability of a clinically significant progression (a two-grade increase in DPS). In the absence of photoprotection, the probability of reaching this elevated grade by age 53 was 71.35%, whereas regular daily application of SPF 50+ sunscreen reduced this probability to 39.24%.

Conclusions

This framework represents the application of a scientifically validated image-generation tool for predicting age-related hyperpigmentation based on individual exposome factors. By integrating dermatological knowledge, this simulator provides a robust educational and research resource for personalized skincare strategies. The findings confirm that consistent photoprotection significantly mitigates the 15-year trajectory of premature facial aging, offering a scientifically validated foundation for public health communication on prevention.