Automating a Process Convolution Approach to Account for Spatial Information in Imaging Mass Spectrometry Data

Area of Science:

• Computational biology
• Statistical modeling
• Bioinformatics

Background:

• Imaging techniques like imaging mass spectrometry (IMS) generate large, spatially-referenced datasets.
• Existing analytical tools often neglect spatial information, leading to vague interpretations.
• Automated modeling of spatially-referenced imaging data is needed.

Purpose of the Study:

• To develop automated methods for modeling spatially-referenced imaging data using a process convolution (PC) approach.
• To optimize PC parameter specification for computational efficiency and accurate spatial analysis.
• To provide clear interpretations of covariate effects while maintaining statistical power and type I error control.

Main Methods:

• Developed an automated process convolution (PC) approach for spatially-referenced imaging data.
• Conducted simulation studies to determine optimal PC parameter specification.
• Tested methods using simulations with real spatial information and spiked-in effects.
• Detailed the imagingPC R package for accessibility.

Main Results:

• The proposed methods effectively incorporate spatial information into the analysis.
• Clear interpretations of covariate effects were achieved.
• The methods maximized statistical power and maintained nominal type I error rates.
• The framework is flexible and scalable for various imaging techniques.

Conclusions:

• The automated PC approach offers a robust framework for analyzing spatially-referenced imaging data.
• The imagingPC R package makes these advanced methods accessible to researchers.
• This approach enhances the utility of big data from imaging techniques by leveraging spatial context.