hippo

Main repository for the hippo project.

JIRA Project can be found here.

Introduction

The Hippo app allows users to run inference on custom medical images and get diagnoses. The app will also include online training as a future feature.

Look forward to all the cool new features!

Test Instructions

To run tests using the poetry virtual environment run the following command in the project root directory:

poetry run pytest .\tests\

To run tests and generate coverage reports run the following commands in the project root directory:

poetry run coverage erase
poetry run coverage run -m pytest -v -s
poetry run coverage report -m

Build Instructions

Project Build

TBC

Documentation Build

Building the documentation runs in two stages. Run the following commands in a terminal in the project root folder:

1. Auto build documentation RST files for all modules in the hippo project:

   • poetry run sphinx-apidoc -f -o docs\source hippo --implicit-namespaces

2. Build final documentation set:

   • poetry run sphinx-build -c docs\source -b html docs\source docs\build\html

The documentation can be viewed by opening the .\docs\build\html\index.html file.

Contribution Guidelines

TBC

Pre-commit setup

Setting up the pre-commit library to run the pre-commit hooks whenever you commit your code is done as follows:

1. To setup pre-commit to run using poetry run the command :

   • poetry run pre-commit install

2. If you want to run the pre-commit checks manually run the command:

   • poetry run pre-commit run --all-files

Dataset

The dataset used for training the model is located here. The aim of the model is to identify malignant skin lesions. The dataset has 33126 images coming from 2056 different patients. Of the total images, 98% of them represent healthy skin lesions while 2% represent diseased ones.

The data is in DICOM format. Each .dicom file holds both the image and image metadata. The metadata of each image contains the following fields:

• Age
• Patient_id
• Sex
• Anatomy site
• Diagnosis (this is the type of lesion, while the malignant lesions are all melanomas, the benign lesions can be of different types)
• Benign/malignant flag

More thoughts on this can be found in a google doc

Contributors

• Matei Moldoveanu ([email protected])
• Yohahn Ribeiro ([email protected])