Evaluation of Digital Image Recognition Methods for Mass Spectrometry Imaging Data Analysis.

Area of Science:

• Analytical Chemistry
• Biotechnology
• Pharmacology

Background:

• Mass spectrometry imaging (MSI) data analysis is often manual, time-consuming, and struggles with increasing dataset complexity.
• The need for automated and efficient methods is critical for extracting meaningful molecular distribution information from MSI data.

Purpose of the Study:

• To develop a comprehensive, semi-targeted method for discovering molecular distributions of interest in mass spectrometry imaging data.
• To implement and validate a fast, batch-searchable algorithm for ranking images based on spatial correlation.

Main Methods:

• Utilized structural similarity (SSIM) index scoring, a widely available image similarity algorithm.
• Developed a MATLAB-based batch search method for efficient processing of MSI datasets.
• Applied the method to two distinct datasets: plant metabolite analysis (Artemisia annua leaf) and drug distribution study (maraviroc in macaque tissue).

Main Results:

• Demonstrated the effectiveness of the SSIM-based approach for ranking images by spatial correlation.
• Successfully identified molecular distributions of interest in both plant and animal tissue MSI datasets.
• The method proved to be fast and powerful for batch searching large MSI datasets.

Conclusions:

• The presented semi-targeted discovery method offers a robust and automated solution for analyzing complex mass spectrometry imaging data.
• This approach significantly reduces the labor and time required for identifying spatial molecular patterns.
• The SSIM-based batch search is a valuable tool for researchers working with large-scale MSI datasets.