A Comparative Study on Signal Decomposition Techniques for Stimulated Raman Photoacoustic Microscopy.

Area of Science:

• Biomedical Optics
• Photoacoustic Imaging
• Spectroscopy

Background:

• Multispectral photoacoustic microscopy (MS-PAM) offers functional imaging capabilities.
• Stimulated Raman spectroscopy (SRS) can be integrated with MS-PAM for enhanced molecular contrast.
• Accurate signal decomposition is crucial for quantitative MS-PAM analysis, particularly for oxygen saturation estimation.

Purpose of the Study:

• To comparatively evaluate the performance of different signal decomposition techniques in MS-PAM systems utilizing SRS.
• To assess the accuracy of fast Fourier transform (FFT), least squares (LSQ), cross-correlation (XCorr), and deep learning (DL) methods.
• To determine the optimal method for oxygen saturation estimation in MS-PAM.

Main Methods:

• Developed MS-PAM system incorporating SRS with dual-wavelength excitation (532 nm and 558 nm).
• Implemented and tested four signal decomposition algorithms: FFT, LSQ, XCorr, and a DL approach (CNN-autoencoder).
• Validated methods against ground truth photoacoustic signals from mouse brain tissue.

Main Results:

• The FFT method demonstrated superior accuracy and consistency in signal decomposition compared to LSQ, XCorr, and DL.
• LSQ, XCorr, and DL methods exhibited performance variability, particularly at longer time delays.
• Oxygen saturation estimation using the FFT-decomposed signals yielded reliable results.

Conclusions:

• FFT is the most accurate and reliable method for signal decomposition in SRS-based MS-PAM.
• The findings support the use of FFT for quantitative functional imaging, including oxygen saturation mapping.
• Further optimization of LSQ, XCorr, and DL methods may be needed for specific MS-PAM applications.