Welcome to HoverFast’s documentation!

Introduction

Welcome to the official documentation for HoverFast, a high-performance tool designed for efficient nuclear segmentation in Digital Pathology images.

Overview

HoverFast was designed to provide fast and accurate nuclear segmentation for brightfield digital pathology images. It can take full advantage of multiple CPU cores and provide an output that you can drag-and-drop onto QuPath to visualize your segmentation results.

Complementary tools

As useful as HoverFast is for nuclear segmentation, it can greatly benefit with the use of other tools to help delinate the part of the tissue that should be analyzed. By default, HoverFast will just use a white threshold to detect tissue regions. This can often be suboptimal. For example, in an ideal situation, you would not want to spend time segmenting nuclei on artefacts such as coverslips or dust. For that reason, we highly recommend looking into other tools such as HistoQC (https://github.com/choosehappy/HistoQC) or HistoBlur (https://github.com/choosehappy/HistoBlur) to remove those by generating accurate tissue masks.

Contents:

• Installation Guide
  • Prerequisites
  • Using Docker
  • Using Singularity
  • Local Installation with Conda
  • Advanced Installation Options
  • Troubleshooting
• Infer WSI Module
  • Infer WSI Parser
  • Usage Examples
• Infer ROI Module
  • Infer ROI Parser
  • Usage Examples
• Train Module
  • Dataset Generation
  • Train Parser
  • Usage Examples
• Unit testing
  • Setting Up the Environment
  • Running the Tests
  • Special Considerations
  • Conclusion