Unmixing Biological Fluorescence Image Data with Sparse and Low-Rank Poisson Regression

Area of Science:

• * Multispectral biological fluorescence microscopy
• * Advanced image analysis and deconvolution techniques

Background:

• * Accurate identification of multiple fluorophores in complex biological samples is crucial but challenged by increasing fluorophore numbers and decreasing signal-to-noise ratios.
• * Existing spectral unmixing methods struggle with highly overlapping fluorophores and low signal conditions, leading to degraded accuracy.
• * Prior knowledge of fluorophore spatial distributions can enhance identification and abundance estimation.

Approach:

• * Proposed a regularized sparse and low-rank Poisson unmixing approach (SL-PRU) for deconvolving spectral images with highly overlapping fluorophores.
• * SL-PRU incorporates multi-penalty terms for sparseness and spatial correlation, and utilizes Poisson regression for improved photon abundance estimation.
• * Developed a parameter tuning method for SL-PRU applicable even without ground truth abundance information.

Key Points:

• * SL-PRU effectively handles highly overlapping fluorophores in low signal-to-noise regimes.
• * The method simultaneously enforces sparseness and spatial correlation in abundance maps.
• * Poisson regression offers a more accurate estimation of photon abundance compared to least squares.
• * Parameter tuning method enables practical application without ground truth data.

Conclusions:

• * The proposed SL-PRU method significantly improves the accuracy of fluorophore identification and abundance estimation in challenging spectral imaging scenarios.
• * SL-PRU demonstrates superior performance on both simulated and real-world biological images compared to existing methods.
• * This approach offers a valuable tool for quantitative analysis in multispectral fluorescence microscopy.