A Statistical Approach for Improving Image Quality and Cell Quantification from Imaging Mass Cytometry

Imaging Mass Cytometry (IMC), a multiplexed imaging technology, has become a valuable tool in biomedical research due to its capability to measure over 100 markers theoretically. However, the presence of noise in IMC images can affect cell phenotyping and the precision of clinical analysis. To address this challenge, we propose IMCell, a highly effective and generalizable method that enhances image quality and cell quantification by leveraging biological priors. IMCell begins by generating decoy cells using Monte Carlo sampling and incorporates cell shape information for each protein channel. Subsequently, it estimates background noise from the expression of decoy cells by fitting Gaussian mixture models with the EM algorithm. Additionally, it identifies positive cell expressions by comparing distribution patterns between real cells and decoy cells. Based on the analysis results from IMCell, we further introduce an image quality index for comprehensive image quality assessment by combining noise and positive cell expressions. Experiment results demonstrate the efficacy of our proposed approach in enhancing image quality and facilitating downstream analysis. By improving the quality of IMC images, IMCell holds the potential to enhance our understanding of complex biological systems and disease mechanisms.