Phasor-Based Spatio-Spectral Segmentation for Hyperspectral Fluorescence Microscopy.

Fluorescence microscopy is essential for visualizing cellular and tissue structures. When combined with a hyperspectral modality, it enables the acquisition of emission spectra that provide insight into the composition of fluorophores. These spectra can be used for fluorophore-based image segmentation; however, accurate spectral separation becomes challenging when fluorophore emission spectra overlap and mix within individual pixels. We present a spatially-corrected density peak clustering (SC-DPC) method that combines phasor transformation with histogram-based density-peak detection and additional spatial correction to achieve accurate and fast clustering. SC-DPC effectively handles large datasets and clusters that are not separable linearly in phasor space while maintaining high clustering accuracy and computational efficiency. Results with synthetic data and experimental multiphoton microscopic data from the mouse retina demonstrate that SC-DPC provides a robust, high-performance approach for hyperspectral fluorescence data segmentation.