Related Experiment Video
Updated: Jul 15, 2026
06:44
Automated Dissection Protocol for Tumor Enrichment in Low Tumor Content Tissues
Published on: March 29, 2021
Tissue-Preserving Artifact Quality Control for Whole-Slide Computational Pathology
Lin Wei1, Yue Lu1, Yucheng Shi2
1School of Cyber Science and Engineering, Zhengzhou University, No. 100 Science Avenue, Zhengzhou, Henan, 450001, China.
Journal of Imaging Informatics in Medicine
|July 13, 2026
Summary
GuardQC is a new computational pathology pipeline that controls artifacts in whole-slide imaging (WSI) while preserving essential tissue. This quality control method improves diagnostic model reliability by minimizing false positives and tissue loss.
Area of Science:
- Computational pathology
- Digital pathology
- Medical image analysis
Background:
- Reliable whole-slide imaging (WSI) analysis requires robust diagnostic models and effective image-quality control.
- Existing high-sensitivity artifact masks can remove relevant tissue, particularly in dense regions, impacting diagnostic accuracy.
- Artifacts in WSI can be mistaken for benign histology, leading to misinterpretation.
Purpose of the Study:
- To introduce GuardQC, a conservative preprocessing pipeline for artifact control and tissue preservation in computational pathology.
- To develop a quality control method that minimizes artifact masking while maintaining tissue integrity.
- To enhance the reliability of diagnostic models by improving WSI preprocessing.
Main Methods:
- GuardQC employs a coarse-to-fine strategy using a high-recall proposal network and a dual-stream ResNet-18 verification model.
- The verification model integrates local texture and tissue context across multiple resolutions after dynamic alignment.
- Iterative hard negative mining, localized Otsu re-segmentation, and PDE-based inpainting are used for artifact neutralization and tissue preservation.
Main Results:
- GuardQC achieved 98.99% accuracy and 100% specificity on an expert-curated subset.
- The pipeline masked significantly less tissue (0.32%) compared to a deep learning baseline (0.94%) in a WSI cohort.
- Feature ablations demonstrated substantial recovery of representation consistency for various machine learning models.
Conclusions:
- GuardQC serves as an effective quality-control layer for WSI preprocessing.
- The method successfully balances artifact control with critical tissue preservation.
- GuardQC supports scanner-aware analysis and inspectable processing for computational pathology applications.
