RT-LAMP-DETR

This repository contains the code for RT-LAMP-DETR, an artificial intelligence (AI) operated-tool to enable a more precise and rapid result analysis in large scale testing presented in One-step colorimetric isothermal detection of COVID-19 with AI-assisted automated result analysis: a platform model for future emerging point-of-care RNA/DNA diagnosis. In this project, we developed a novel ultrasensitive and specific dual RT-LAMP assay targeting the Nsp9 segment of SARS-CoV-2 ORF1ab gene and human 18S rRNA gene (internal control) for decentralized COVID-19 screening with two modes of analysis: naked-eye observation for the highest convenience, and an AI-operated automated analysis to accommodate high-throughput testing.

This repository is based on the original DETR code from https://github.com/facebookresearch/detr

RT-LAMP-DETR dataset

The dataset is under _data folder of this repository. The data is already split into 3 folders for training (train), validation (val) and testing (test).

Install the required packages

pip install -r requirements.txt

Training RT-LAMP-DETR from scratch

./train_rt_lamp_detr.sh

Evaluate RT-LAMP-DETR

Please see eval_rt_lamp_detr.ipynb notebook for the code to evaluate RT-LAMP-DETR based on the trained model.

Our pretrained RT-LAMP-DETR model

You can download our pretrained RT-LAMP-DETR model from here. If the link does not work, please file an issue and we will update it.

Using docker (https://www.docker.com/)

git clone [email protected]:peer-ai/rt-lamp-detr.git
cd rt-lamp-detr
docker build . -t rt-lamp-detr

# train 
docker run -v `pwd`:/workspace --gpus 0 --shm-size=2g -it rt-lamp-detr ./train_rt_lamp_detr.sh

# run jupyterlab
docker run -v `pwd`:/workspace --gpus 0 --shm-size=2g -p 9001:9001 -it rt-lamp-detr jupyter-lab --allow-root --port 9001 --ip "*"

Then, please visit this url [http://localhost:9001/lab?token=...] (fill in with token shown after running jupyter-lab command above) with your browser. This will bring up jupyterlab enviroment for you to run eval_rt_lamp_detr.ipynb notebook to evaluate the trained model.

Citation

Please cite this work if you find it useful.

@journal{rt-lamp-detr,
  author = {...},
  title = {One-step colorimetric isothermal detection of COVID-19 with AI-assisted automated result analysis: a platform model for future emerging point-of-care RNA/DNA diagnosis},
  year = {2021},
  publisher = {...},
  journal = {...},
  howpublished = {...}
}