Related Experiment Video
Updated: Dec 25, 2025
06:08
Quantitative Visualization and Detection of Skin Cancer Using Dynamic Thermal Imaging
Published on: May 5, 2011
17.1K
Predictive model for the quantitative analysis of human skin using photothermal radiometry and diffuse reflectance
Nina Verdel1,2,3, Jovan Tanevski4, Sašo Džeroski4,5
1Department of Complex Matter, Jožef Stefan Institute, Jamova cesta 39, 1000 Ljubljana, Slovenia.
Biomedical Optics Express
|March 25, 2020
Summary
A new method noninvasively analyzes skin structure using pulsed photothermal radiometry (PPTR) and diffuse reflectance spectroscopy (DRS). Machine learning significantly speeds up analysis, improving accuracy and robustness.
Area of Science:
- Biomedical Optics
- Dermatology
- Medical Imaging
Background:
- Noninvasive skin analysis is crucial for diagnosing and monitoring various dermatological conditions.
- Accurate assessment of skin chromophores, scattering properties, and layer thicknesses is essential.
- Existing methods often face computational challenges, limiting their practical application.
Purpose of the Study:
- To develop a computationally efficient method for noninvasive human skin analysis.
- To improve the accuracy and robustness of skin structure and composition assessment.
- To reduce the computational burden associated with complex numerical models.
Main Methods:
- Combined pulsed photothermal radiometry (PPTR) and diffuse reflectance spectroscopy (DRS) for data acquisition.
- Developed a machine learning-based predictive model (PM) using random forests, trained on ~9,000 examples.
- Implemented a hybrid approach (HA) integrating the PM's speed with the IMC procedure's versatility.
Main Results:
- The predictive model (PM) demonstrated highly satisfactory performance in cross-validation and direct comparisons.
- The hybrid approach (HA) significantly improved accuracy and robustness compared to the inverse Monte Carlo (IMC) procedure.
- The HA substantially reduced computation times while maintaining high analytical performance.
Conclusions:
- The developed machine learning predictive model offers a fast and accurate alternative for noninvasive skin analysis.
- The hybrid approach effectively balances computational speed with analytical precision and robustness.
- This novel methodology holds significant potential for advancing dermatological diagnostics and research.
