COVID-Net USPro: An Explainable Few-Shot Deep Prototypical Network for COVID-19 Screening using Point-of-Care Ultrasound
The COVID-Net USPro is an explainable few-shot deep prototypical network that is designed to detect COVID-19 cases from very few ultrasound images. As part of the COVID-Net initiative, and to promote reproducibility and foster further innovation, the network is open-sourced and available to the public.
For complete environment setup, please refer to the environment.yml file in this repository.
The main requirements include:
- PyTorch 1.10.0
- OpenCV 4.5.5
- Python 3.9.9
- Numpy
- Scikit-Learn
- Matplotlib
For an example on how to use COVID-Net USPro, please refer to example notebook here: notebook
The COVIDx-US dataset is an open-access benchmark dataset of ultrasound imaging data for AI-driven COVID-19 analytics. The github repo can be accessed here: https://github.com/nrc-cnrc/COVID-US. For the complete description, please refer to paper. This dataset is used to train and evaluate COVID-Net USPro.
The analysis workflow for COVID-Net USPro is outlined below.
The prototypical network structure of COVID-Net USPro is also illustrated below.
Full test results are detailed in experiment_outputs. The following plots illustrate the testing results of two encoders ResNet18L1 and ResNet50L4.
