Deep Learning-based Denoising of Monte Carlo Simulation for Laser-Tissue Interaction Modeling

Key Points

• The aim is to develop a deep learning method to reconstruct high-sampling Monte Carlo simulation data from low-sampling data for laser-tissue interaction modeling.
• Generated low-sampling and high-sampling Monte Carlo simulation data based on optical properties of laser-treated skin.
• Extracted surface morphology from optical coherence tomography (OCT) images.
• Trained and compared three denoising neural networks: U-Net, Restormer, and NAFNet.
• Achieved a mean squared error (MSE) of 14.52 against high-resolution data.
• Obtained a peak signal-to-noise ratio (PSNR) of 49.46, indicating high accuracy.
• Total computation time reduced by 4.65 times compared to high-sampling simulations, demonstrating improved efficiency.