Pseudo-spectral angle mapping to improve immune cell classification in highly multiplexed fluorescence microscopy images

• DOI: 10.1117/12.3003486
• Publication Date: 2024-03-12T22:44:26Z
• Authors: Madeleine S. Durkee; Junting Ai; Gabriel Casella; Thao Cao; Marcus R. Clark; Maryellen L. Giger

Highly multiplexed fluorescence microscopy is an emerging technology that allows for spatial analysis of increasingly more classes cells within human tissue—state-of-the-art methods are now probing up to 60 different protein markers image. This level phenotypic resolution ideal uncovering the underpinnings immune cell interactions. However, defining types from this high-plex data non-trivial. We present a method borrows hyperspectral image improve accuracy and efficiency classification in highly images. Treating marker channels as spectral dimension images, we define reference "pseudospectra" representative expression all interest probed by panel. Cosine similarity computed each pseudo-spectra create class maps type question. Features extracted these maps—rather than compare decision-tree based classifying unsupervised K-means clustering mean pixel intensities across channels. demonstrate pSAM performs comparably, potentially outperforms with similar levels supervision.