UniFORM: A New Method for Normalizing Multiplex Tissue Imaging Data
Author Information
Author(s): Wang Kunlun, Ait-Ahmad Kaoutar, Kupp Samuel D, Sims Zachary, Cramer Eric, Sayar Zeynep, Yu Jessica, Wong Melissa H., Mills Gordon B., Eksi S. Ece, Chang Young Hwan
Primary Institution: Oregon Health & Science University (OHSU)
Hypothesis
Can UniFORM provide a better normalization method for multiplex tissue imaging data compared to existing methods?
Conclusion
UniFORM outperforms existing normalization methods in preserving biological signal fidelity and mitigating batch effects in multiplex tissue imaging data.
Supporting Evidence
- UniFORM consistently outperformed existing methods in aligning marker distributions across multiple datasets.
- UniFORM achieved the highest average kBET score of 0.60843, indicating effective batch correction.
- Downstream analyses showed improved clustering and visualization after normalization with UniFORM.
Takeaway
UniFORM is a new tool that helps scientists make sure their images of tissues look right by fixing differences in how bright or dark they are, so they can see the real biology better.
Methodology
UniFORM uses a non-parametric, automated pipeline for normalizing both feature-level and pixel-level data, preserving marker intensity distribution shapes while maintaining positive population proportions.
Potential Biases
Potential biases may arise from the reliance on linear normalization factors, which may not be suitable for all types of imaging data.
Limitations
The effectiveness of UniFORM depends on the quality of the cell/nuclei segmentation masks used to extract signals.
Participant Demographics
The study involved two cancer datasets: CRC-ORION (6 samples) and PRAD-CyCIF (20 samples).
Statistical Information
P-Value
0.25835
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website