A user-friendly tool for cloud-based whole slide image segmentation with examples from renal histopathology

Area of Science:

• Digital pathology
• Computational biology
• Machine learning in medicine

Background:

• Image-based machine learning tools offer significant potential for clinical pathology research.
• Pathologists and biological scientists often lack the programming skills for command-line-based computational tools.
• A user-friendly interface is needed to bridge the gap between advanced tools and end-users.

Purpose of the Study:

• To develop an accessible, cloud-based machine learning tool for whole slide image (WSI) segmentation.
• To provide a graphical user interface (GUI) for image analysis, eliminating the need for programming expertise.
• To facilitate feature extraction from segmented regions for further pathological research.

Main Methods:

• Developed Histo-Cloud, a cloud-based tool with a GUI for WSI segmentation.
• Utilized a state-of-the-art convolutional neural network (CNN) for segmentation tasks.
• Demonstrated segmentation of glomeruli, interstitial fibrosis, tubular atrophy, and vascular structures.

Main Results:

• Successfully segmented key histological structures in renal and non-renal WSIs.
• Demonstrated the scalability and best practices for transfer learning with Histo-Cloud.
• Evaluated the effects of dataset variability on segmentation performance.
• Applied Histo-Cloud to analyze glomerular features in murine models for animal research.

Conclusions:

• Histo-Cloud is an open-source, web-accessible tool for histological structure segmentation.
• The tool is adaptable for segmenting any histological structure, irrespective of staining.
• Histo-Cloud democratizes advanced image analysis for pathology researchers.