Two-Phase and Graph-Based Clustering Methods for Accurate and Efficient Segmentation of Large Mass Spectrometry...

Area of Science:

• Biomedical Imaging
• Computational Biology
• Data Science

Background:

• Clustering is crucial for segmenting anatomical features and tissue types in mass spectrometry imaging (MSI).
• Current MSI clustering methods are computationally expensive, restricting their use to small datasets.
• There is a need for scalable and efficient algorithms for analyzing large MSI data.

Purpose of the Study:

• To develop a novel, computationally efficient clustering algorithm for mass spectrometry imaging (MSI).
• To enable the segmentation of anatomical features in large-scale MSI datasets.
• To improve the differentiation of tissue types using MSI data analysis.

Main Methods:

• A graph-based algorithm utilizing a two-phase sampling strategy was developed.
• The algorithm was tested on diverse sample types and synthetic MSI data.
• Data acquired with varying laser fluence and designed-in variance were used to assess robustness.

Main Results:

• The proposed algorithm successfully segmented anatomical features missed by conventional MSI algorithms.
• Validation on synthetic data confirmed the algorithm's accuracy.
• The method demonstrated robustness to data quality variations, including differing laser fluence.
• Accurate segmentations of large MSI datasets were achieved, comparable to histopathology.

Conclusions:

• The novel graph-based clustering approach offers a scalable and efficient solution for MSI data analysis.
• This method overcomes the computational limitations of existing techniques, enabling the analysis of large MSI datasets.
• The algorithm's robustness and accuracy provide a valuable tool for MSI-based tissue segmentation and analysis.