IMC-Denoise: a content aware denoising pipeline to enhance Imaging Mass Cytometry

Abstract

Imaging Mass Cytometry (IMC) is an emerging multiplexed imaging technology for analyzing complex microenvironments that has the ability to detect the spatial distribution of at least 40 cell markers. However, this new modality has unique image data processing requirements, particularly when applying this technology to patient tissue specimens. In these cases, signal-to-noise ratio for particular markers can be low despite optimization of staining conditions, and the presence of pixel intensity artifacts can deteriorate image quality and the subsequent performance of downstream analysis. Here we demonstrate an automated content-aware pipeline, IMC-Denoise, to restore IMC images. Specifically, we deploy a differential intensity map-based restoration (DIMR) algorithm for removing hot pixels and a self-supervised deep learning algorithm for filtering shot noise (DeepSNF). IMC-Denoise outperforms existing methods for adaptive hot pixel removal, and delivers significant image quality improvement and background noise removal to a diverse set of IMC channels and datasets. This includes a unique, technically challenging, human bone marrow IMC dataset; in which we achieve noise level reduction of 87% for a 5.6-fold higher contrast-to-noise ratio, and more accurate background noise removal with approximately two-fold improved F1 score. Our approach remarkably enhances both manual gating and automated phenotyping with cell-scale down-stream analysis on these complex data. We anticipate that IMC-Denoise will provide similar benefits in mass cytometry imaging domains to more deeply characterize the complex and diverse tissue microenvironment.