A platform-independent framework for phenotyping of multiplex tissue imaging data.
Mansooreh Ahmadian1, Christian Rickert2, Angela Minic2
1Department of Biostatistics and Informatics, Colorado School of Public Health, University of Colorado Anschutz Medical Campus, Aurora, Colorado, United States of America.
Plos Computational Biology
|September 21, 2023
Summary
Multiplex imaging analysis is improved by a new pixel classifier preprocessing step. This method reduces platform-specific noise and artifacts, enhancing data reliability for various analysis pipelines.
Area of Science:
- Computational Biology
- Biomedical Imaging
Background:
- Multiplex imaging offers deep insights into cellular states but presents analysis challenges due to data complexity.
- Existing computational pipelines often lack cross-platform applicability, hindering reproducible results.
Purpose of the Study:
- To introduce a platform-independent image preprocessing method for multiplex imaging data.
- To enhance the reliability and reproducibility of multiplex image analysis across different techniques.
Main Methods:
- A pixel classifier-based approach was developed to distinguish signal from noise and artifacts.
- This method generates feature maps with improved signal-to-noise ratios and normalized pixel values.
- The preprocessing step was validated using two established image analysis pipelines.
Main Results:
- The proposed method effectively reduces platform-specific characteristics in multiplex images.
- Feature representation maps showed improved signal-to-noise ratios and normalized data.
- Downstream analysis pipelines yielded comparable results after preprocessing.
Conclusions:
- The pixel classifier preprocessing step minimizes platform dependency in multiplex imaging analysis.
- This approach enhances the generalizability and robustness of computational pipelines for multiplex imaging data.
