An Automatic Method Using Hybrid Genetic Algorithms to Fit Zernike Polynomials for Fluorescence Removal in Raman Spectra

ABSTRACT Removing fluorescence from the Raman spectra of biological samples poses significant challenges because of the diverse baseline variations. Current methods often fail to accurately remove fluorescence, compromising the results of Raman spectroscopy analyses. We introduce a novel iterative numerical optimization method using Zernike polynomials (INOM‐ZP) as its foundation. By integrating the minimum point algorithm (MPA) with a genetic algorithm (GA), our approach minimizes an objective function in the real domain, avoiding premature convergence to local optima and achieving solutions close to the global minimum. This method addresses the pressing need for robust fluorescence removal techniques, thereby enhancing the accuracy of Raman spectroscopy analysis. The INOM‐ZP algorithm optimally fits Zernike polynomials to experimental data associated with fluorescence, enabling automatic validation of baseline curve reconstruction and quantitative characterization of fluorescence intensity variations. Our method effectively removes fluorescence from Raman spectra across various sample types, representing a significant advancement in fluorescence removal techniques for Raman spectroscopy. The best fluorescence extraction result for a synthetic Raman spectrum in reliability was 0.99 ± 6.8203 × 10 −9 , with a precision of 0.00633 ± 2.1092 × 10 −7 and a repeatability of 0.0125 ± 6.9984 × 10 −8 . Similarly, for biological spectra, such as bone, the reliability was 0.99 ± 1.835 × 10 −8 , with a precision of 0.0577 ± 3.889 × 10 −8 and a repeatability of 0.0775 ± 1.935 × 10 −7 .