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

Area of Science:

• Biomedical imaging
• Computational pathology
• Biotechnology

Background:

• Imaging Mass Cytometry (IMC) enables high-plex analysis of tissue microenvironments.
• IMC data can suffer from low signal-to-noise ratios and artifacts, hindering analysis.
• Existing noise reduction methods are insufficient for complex IMC datasets.

Purpose of the Study:

• To develop an automated pipeline, IMC-Denoise, for restoring IMC images.
• To improve image quality and enable more accurate downstream analyses.
• To address challenges in analyzing patient tissue specimens with IMC.

Main Methods:

• Developed IMC-Denoise, an automated content-aware pipeline.
• Implemented differential intensity map-based restoration (DIMR) for hot pixel removal.
• Utilized self-supervised deep learning (DeepSNiF) for shot noise filtering.

Main Results:

• IMC-Denoise significantly improves image quality in modeled and real-world IMC data.
• Achieved 87% noise reduction and a 5.6-fold higher contrast-to-noise ratio in human bone marrow.
• Demonstrated enhanced manual gating and automated phenotyping, revealing subtle cellular differences in diseased tissues.

Conclusions:

• IMC-Denoise effectively removes noise and artifacts from IMC images.
• The pipeline enhances the accuracy of cell-scale downstream analyses.
• IMC-Denoise is expected to benefit various mass cytometry applications for tissue microenvironment characterization.