Sparse Regularization-Based Unmixing Methods for Photoacoustic Image Analysis

Photoacoustic imaging leverages the specific differences in light absorption by biomacromolecules to reflect biochemical composition features within biological tissues, demonstrating significant potential for early disease diagnosis. High-precision photoacoustic unmixing algorithms can parse out the contributions from individual constituents from multi-wavelength photoacoustic imaging, thereby enabling the identification, localization, and quantification of biomarkers associated with disease onset and progression. This study introduces a refined sparse regularization-based unmixing method (USR) for unmixing photoacoustic images, which integrates an L1 regularization constraint mechanism on the basis of least squares. This enhancement improves the model’s capacity to represent sparse information about chromophores in biological tissues, thereby improving the accuracy and fidelity of component unmixing. To verify the feasibility of this algorithm, we set up two different sets of simulation models to simulate the distribution of chemical components in biological tissues under different conditions. Experiments with simulated data demonstrated that this algorithm notably enhances unmixing accuracy when compared to traditional algorithms. It effectively mitigates the risk of overfitting and showcases superior performance in processing complex photoacoustic signals from biological tissues.