Denoising influence on discrete frequency classification results for quantum cascade laser based infrared microscopy

Area of Science:

• Biomedical Optics
• Spectroscopy
• Medical Imaging

Background:

• Fourier Transform Infrared (FT-IR) imaging is too slow for clinical applications.
• Quantum Cascade Laser Infrared (QCL IR) microscopy in Discrete Frequency (DF) mode offers a faster alternative.
• Effective denoising is crucial for optimizing QCL IR data quality and analysis speed.

Purpose of the Study:

• To investigate the impact of various denoising algorithms on the classification accuracy of QCL IR images.
• To determine the most effective denoising schemes for QCL IR data, specifically for pancreatic tissue histology.
• To assess the transferability of findings to other DF imaging modalities.

Main Methods:

• Comparison of spectral and spatial denoising techniques on FT-IR imaging data.
• Application of multivariate denoising methods like Principal Component Analysis (PCA) and Minimum Noise Fraction (MNF).
• Histologic classification of QCL IR images of pancreatic tissue using Random Forest algorithms.

Main Results:

• Denoising significantly impacts the classification accuracy of QCL IR data.
• Multivariate denoising methods generally show high effectiveness for datasets with multiple spectral bands.
• The study identified optimal denoising strategies for QCL IR image classification.

Conclusions:

• Denoising is essential for enhancing the clinical utility of QCL IR microscopy.
• The findings provide a framework for optimizing QCL IR data processing and analysis.
• This research is likely applicable to other DF imaging techniques like AFM-IR and CARS/SRS imaging.