Simulating cardiac signals on 3D human models for photoplethysmography development

Area of Science:

• Biomedical Engineering
• Computer Vision
• Medical Imaging

Background:

• Contactless healthcare monitoring using image-based heart rate estimation is promising.
• Existing methods require extensive datasets with varied conditions (motion, lighting, physiology).
• Collecting high-quality, parameter-rich datasets is a significant challenge.

Purpose of the Study:

• To introduce a 3D bionic human model for generating synthetic data.
• To evaluate traditional and deep learning-based imaging photoplethysmography (iPPG) methods.
• To overcome limitations in real-world data acquisition for vital signs monitoring.

Main Methods:

• Development of a 3D bionic human model with integrated synthetic cardiac signals and motion.
• Rendering videos of the 3D model under various scenarios (stillness, talking, lighting changes, activity).
• Testing five traditional and four deep learning iPPG extraction methods on rendered videos.

Main Results:

• The 3D model can be customized with different appearances and skin tones.
• Performance of iPPG methods on synthetic data showed high agreement with real human data.
• The study identified performance advantages and disadvantages of selected iPPG methods consistent across synthetic and real data.

Conclusions:

• The 3D bionic human model provides a controllable method for generating synthetic data for vital signs monitoring.
• This approach is valuable for testing and optimizing image-based vital signs methods in challenging scenarios.
• Potential applications include situations where real-world data is difficult to obtain, such as in drone-assisted rescue.