Dragonfly algorithm-support vector machine approach for prediction the optical properties of blood

Area of Science:

• Biomedical Optics
• Artificial Intelligence in Medicine
• Computational Biology

Background:

• Optical properties of blood are vital for medical diagnostics and laser-based therapies.
• Accurate estimation of blood's absorption and scattering coefficients is essential for these applications.

Purpose of the Study:

• To develop a rapid and accurate artificial intelligence model for estimating blood's optical properties.
• To build highly accurate Dragonfly Algorithm-Support Vector Regression (DA-SVR) models for predicting absorption and scattering coefficients.

Main Methods:

• Utilized Dragonfly Algorithm/Support Vector Machine (SVM) models for estimation.
• Developed DA-SVR models using wavelength, hematocrit, and oxygen saturation as key parameters.
• Trained and tested models on 1000 datasets within the 250-1200 nm wavelength range and 0-100% hematocrit.

Main Results:

• Achieved high accuracy with correlation coefficients (R) of 0.9994 for absorption and 0.9957 for scattering.
• Reported low Root Mean Squared Error (RMSE) values (0.972 and 2.9193) and Mean Absolute Error (MAE) values (0.2173 and 0.2423).
• Demonstrated a strong match between model predictions and experimental data.

Conclusions:

• The developed DA-SVR models accurately predict blood's absorption and scattering coefficients.
• These models offer a reliable reference for future research on human blood's optical properties.
• The AI approach provides a rapid and precise tool for optical property estimation in biomedical applications.