Structure-Preserving and Perceptually Consistent Approach for Visualization of Mass Spectrometry Imaging Datasets

Area of Science:

• Biomedical Imaging
• Molecular Imaging
• Computational Biology

Background:

• Mass spectrometry imaging (MSI) is crucial for visualizing compound distribution in biological tissues, aiding in molecular organization studies and abnormality detection.
• Analyzing large MSI datasets is challenging due to spectral complexity and sample heterogeneity, hindering effective exploratory visualization.
• Existing dimensionality reduction techniques, while useful, may not fully capture spatial and compositional information critical for MSI data interpretation.

Purpose of the Study:

• To explore and propose an advanced visualization approach for mass spectrometry imaging (MSI) data.
• To overcome limitations in analyzing complex MSI datasets and improve exploratory visualization.
• To enable visual comparison of different MSI datasets without requiring extensive prior knowledge.

Main Methods:

• Exploration of established dimensionality reduction techniques including principal component analysis, independent component analysis, non-negative matrix factorization, t-distributed stochastic neighbor embedding, and uniform manifold approximation and projection.
• Development of a novel approach combining structure-preserving visualization with nonlinear manifold embedding of normalized spectral data.
• Application of the proposed method to visualize molecular layers in chimpanzee and macaque cerebellum slices.

Main Results:

• The proposed method effectively preserves spatially overlapping signals while incorporating compositional spectral variations.
• It facilitates clear visualization of distinct tissue layers, such as the molecular layer, granular layer, and white matter, in cerebellum slices.
• The approach allows for visual comparison across different MSI datasets.

Conclusions:

• The novel visualization approach significantly enhances the exploratory analysis of mass spectrometry imaging data.
• This method aids in understanding molecular organization and identifying tissue structures without extensive prior chemical or histological knowledge.
• The technique offers a powerful tool for comparative analysis of diverse MSI datasets.