PyHIST: A Histological Image Segmentation Tool

Area of Science:

• Digital Pathology
• Computational Biology
• Machine Learning in Histopathology

Background:

• High-resolution biomedical imaging generates vast datasets, particularly in histopathology.
• Current machine learning workflows for histopathology analysis require extensive preprocessing.
• This preprocessing creates a significant overhead, hindering efficient data analysis.

Purpose of the Study:

• To introduce PyHIST, an open-source, user-friendly command-line tool.
• To streamline the segmentation and preprocessing of whole slide histological images.
• To facilitate the generation of image tiles for machine learning applications.

Main Methods:

• PyHIST offers an optional image rescaling capability to adjust resolution.
• It generates a mask to differentiate tissue from background in histopathological images.
• The tool produces individual image tiles containing relevant tissue content.

Main Results:

• PyHIST successfully segments tissue from background in whole slide images.
• It generates preprocessed image tiles suitable for machine learning input.
• The tool simplifies and automates key steps in histopathological image analysis.

Conclusions:

• PyHIST provides an efficient solution for histopathological image preprocessing.
• It reduces the computational burden associated with preparing data for machine learning.
• This tool enhances the accessibility and applicability of machine learning in digital pathology.