A deep convolutional neural network for Recurrent Abnormality Detection in chromosomes!
Explore the docs »
View Demo
·
Report Bug
·
Request Feature
Table of Contents
This work has been published by BMC Bioinformatics. Please check out the paper and reach out with any questions or comments!
There are many great neural networks for classifying chromosomes, however, a great deal of clinical cytogenetics involves analyzing samples with abnormal chromosomes. So, we created Chromosome Recurrent Abnormality Detector (ReAD) to automatically classify many important abnormalities found in patients with blood cancer.
Here's why:
- Currently, cytogeneticists spend the majority of their time manually classifying chromosomes but can only inspect a tiny fraction of the cells collected from imaging.
- Recurrent abnormalities can heavily inform the diagnosis, prognisis, and treatment plan.
- An automated classifier could screen hundreds of cells for abnormalities and flag cells requiring further inspection.
Certainly, this neural network does not cover all types of recurrent abnormalities in hematopathology. So I'll be updating the model in the near future to incorporate more. If you have your own dataset, you may also suggest changes by forking this repo and creating a pull request or opening an issue.
- python 3.7
- fastai 2.1.4
- fastai2 0.0.30
- fastasi2-extensions 0.0.31
- torch 1.7.0
- jupyter 1.0.0
- numpy 1.19.4
A dataset of individual chromosome images.
Clone the repository:
git clone https:github.com/DaehwanKimLab/Chromosome-ReAD.
The directory contains 4 Jupyter notebooks, one for each convolutional neural network architecture tested, that can be ran to repeat training network from scratch and inferring on the test set. If you would like to preserve the orginal experimental results but would like to rerun the experiments yourself, simply make a copy a notebook and run the cells in the duplicate.
See the open issues for a list of proposed features (and known issues).
Contributions are what make the open source community such an amazing place to be learn, inspire, and create. Any contributions you make are greatly appreciated.
- Fork the Project
- Create your Feature Branch (
git checkout -b feature/AmazingFeature) - Commit your Changes (
git commit -m 'Add some AmazingFeature') - Push to the Branch (
git push origin feature/AmazingFeature) - Open a Pull Request
Distributed under the MIT License. See LICENSE for more information.
Your Name - @and_m_cox - Andrew.Cox@UTSouthwestern.edu
Project Link: https://github.com/DaehwanKimLab/Chromosome-ReAD
UT Southwestern Bioinformatics and Pathology Departments.