Related Experiment Video
Updated: Oct 20, 2025
05:22
Author Spotlight: Enhanced Multiplex Immunofluorescent Microscopy Protocol for Neuroscience Research
Published on: June 21, 2024
567
HistoClean: Open-source software for histological image pre-processing and augmentation to improve development of
Kris D McCombe1, Stephanie G Craig1, Amélie Viratham Pulsawatdi1
1Patrick G Johnston Centre for Cancer Research, Queen's University Belfast, Belfast, Northern Ireland.
Computational and Structural Biotechnology Journal
|September 15, 2021
Summary
HistoClean is a new, user-friendly graphical tool that simplifies digital pathology image pre-processing and augmentation for deep learning models. It improves convolutional neural network accuracy for cancer prediction tasks without requiring coding knowledge.
Area of Science:
- Digital Pathology
- Computational Pathology
- Artificial Intelligence in Medicine
Background:
- Digital pathology has advanced cancer research, driving the use of deep learning (DL) for image analysis.
- Image pre-processing and augmentation are crucial for robust DL models, but often require coding expertise.
- A gap exists between pathology and computer science, hindering the adoption of advanced image analysis techniques.
Purpose of the Study:
- To introduce HistoClean, a graphical user interface (GUI) for streamlined image pre-processing and augmentation in digital pathology.
- To bridge the knowledge gap between pathologists, biomedical scientists, and computer scientists.
- To demonstrate the utility of HistoClean in improving DL model performance for cancer-related tasks.
Main Methods:
- Developed HistoClean, a user-friendly GUI integrating multiple image processing modules.
- Utilized HistoClean for pre-processing digital pathology images for a convolutional neural network (CNN).
- Applied classical image augmentation and pre-processing techniques through HistoClean.
Main Results:
- HistoClean successfully pre-processed images, enhancing a simple CNN model's accuracy.
- Improvements in model accuracy were observed at tile, region of interest, and patient levels.
- The study confirmed HistoClean's effectiveness in improving standard DL workflows.
Conclusions:
- HistoClean provides an accessible solution for image pre-processing and augmentation in digital pathology.
- The tool facilitates the application of advanced techniques without prior coding knowledge, benefiting interdisciplinary collaboration.
- HistoClean is a valuable, free, and open-source resource for the digital pathology and AI research community.
