Optimization of denoising approaches in the context of ultra-fast LIBS imaging

Abstract

Laser-Induced Breakdown Spectroscopy (LIBS) has emerged as a powerful analytical tool capable of providing multi-elemental information from a single laser pulse with minimal sample preparation. This technique generates a laser-induced, transient plasma on the sample surface, whose spectral emission is analyzed to determine its elemental composition. μLIBS-Imaging, a variant offering spatially resolved elemental analysis, holds promise for applications in diverse fields such as industry, geology, forensics, and biomedicine. Our drive to go ever faster and analyze increasingly larger areas of interest in samples now compels us to use kHz lasers for this elemental imaging. Despite its potential, implementing such lasers in μLIBS-imaging would face diverse challenges mainly related to weak plasma emission and signal-to-noise ratio (SNR) degradation, particularly when applied to delicate biological samples. This paper investigates methods to enhance SNR in fast μLIBS imaging, particularly for biomedical applications. We focus on denoising techniques suitable for high-frequency laser applications, comparing methods like Savitzky-Golay smoothing, Fast Fourier Transform, wavelet-based filtering, Whittaker Filtering, and Principal Component Analysis (PCA). Our strategy optimizes denoising parameters for specific elemental emission peaks, enhancing SNR for individual elements of interest. The results demonstrate significant improvements in data quality, paving the way for more accurate and efficient elemental imaging in complex biomedical specimens.