Quadratic blind linear unmixing: A graphical user interface for tissue characterization.

Area of Science:

• Spectroscopy and Imaging
• Biophysics
• Computational Biology

Background:

• Spectral unmixing decomposes complex sample data into constituent components and their abundances.
• Prior research focused on blind unmixing for multi-spectral fluorescence lifetime imaging microscopy (m-FLIM) datasets using linear and quadratic models.
• Existing methods can be limited by the number of components or end-members.

Purpose of the Study:

• To present an interactive software tool for spectral unmixing.
• To implement blind end-member and abundance extraction (BEAE) and quadratic blind linear unmixing (QBLU) algorithms.
• To provide a flexible solution for spectral data decomposition with or without prior knowledge of component numbers.

Main Methods:

• Development of interactive software in Matlab implementing BEAE and QBLU algorithms.
• Software allows estimation of end-members and abundances when the number of components is known.
• Software provides a blind solution to estimate component number, end-members, and abundances when no prior knowledge is available.

Main Results:

• The software successfully performs spectral unmixing for multi/hyper-spectral data.
• Validated performance across diverse biological samples: ex-vivo human coronary arteries, human breast cancer cells, and in-vivo hamster oral mucosa.
• Demonstrated efficiency and ease-of-use for spectral data decomposition.

Conclusions:

• The developed software offers a powerful and accessible tool for spectral unmixing.
• It provides both guided and completely blind solutions for component and abundance extraction.
• The software is freely available, facilitating advancements in multi/hyper-spectral data analysis.