Rapid artefact removal and H&E-stained tissue segmentation

Area of Science:

• Digital Pathology
• Computational Pathology
• Image Analysis

Background:

• Whole-slide imaging (WSI) is crucial for digital pathology.
• Artefacts in H&E-stained WSIs can hinder accurate analysis.
• Existing segmentation methods may struggle with diverse artefacts.

Purpose of the Study:

• To develop a rapid and robust method for segmenting H&E-stained tissue in WSIs.
• To eliminate common artefacts like pen marks and scanning issues.
• To provide an alternative to complex machine learning approaches.

Main Methods:

• Utilized a single-channel representation of a low-magnification RGB overview of WSIs.
• Employed pixel value bimodality for distinguishing tissue from artefacts.
• Applied Otsu thresholding on the manipulated colour space representation.

Main Results:

• Successfully segmented H&E-stained tissue and removed artefacts in 29 out of 30 diverse WSIs.
• Outperformed Otsu thresholding and Histolab tools in artefact removal.
• Demonstrated high efficacy across WSIs from various institutions and scanners.

Conclusions:

• The proposed method offers a simple yet effective solution for H&E-stained tissue segmentation in WSIs.
• Artefact removal is achieved rapidly without machine learning or parameter tuning.
• This approach enhances the reliability of WSI analysis in digital pathology.