Preserving full spectrum information in imaging mass spectrometry data reduction

Area of Science:

• Biomedical Imaging
• Analytical Chemistry
• Data Science

Background:

• Imaging mass spectrometry (IMS) is crucial for molecular tissue characterization.
• IMS datasets are large, posing significant data size challenges.
• Traditional data reduction methods like peak picking can omit low-abundance species and introduce bias.

Purpose of the Study:

• To develop an alternative data reduction technique for IMS.
• To achieve data size reduction while preserving full spectral information.
• To mitigate biases and information loss inherent in traditional methods.

Main Methods:

• Utilized a low-rank matrix completion model.
• Employed a randomized sparse-format-aware algorithm for data approximation.
• Developed a method to approximate IMS datasets, reducing dimensionality.

Main Results:

• The new method achieves data size reduction comparable to peak picking.
• It better conserves ion intensity profiles and preserves full spectrum information.
• Demonstrated improved preservation of low signal-to-noise ratio signals and near-isobars, mitigating selection bias.

Conclusions:

• The proposed method offers a comprehensive approach to IMS data compression.
• It enables more thorough analysis by retaining complete spectral profiles.
• This technique reduces information loss and selection bias in IMS data analysis.