This is a software developed for the classification of osteoarthritis. Automatic grading of knee osteoarthritis severity combined with joint space narrowing. After importing the original image of the posterior anterior view of the knee X-ray, the software automatically grades the severity of osteoarthritis of the knee (KOA) and provides knee joint detection, knee bone segmentation and joint space narrowing.
This repository provides a visual (GUI) version of the executable software for display analysis of individual cases and non-display batch processing of multiple case treatments. In addition, this executable software is user-friendly and does not require a coding background or python environment; we believe these can be used as reference and learning software for radiologists in their daily diagnoses.
In addition to the executable version, we provide the source code as well as the dockerfile and python environment requirements. We welcome any further development and modifications under our non-commercial use license.
Developed by: Hanxue Gu.
step 1: go to the exe folder , for example: path to /image_final_pipeline
step 2: type in the command line:
./image_final_pipeline --img-name --png-path --txt-path
notice: --img-name: the path and the name of the dicom image. (** we also support png image version now is raw dicom file is not available) --png-path: the path to save the result as an png image. --txt-path: save the result in an txt file.
for exmaple:
./image_final_pipeline \
--img-name "102451" \
--png-path "img_folder/result.png" \
--txt-path "txt_folder/result.txt" \
step 1： go to the exe folder, for example: /image_final_gui
step 2: type in the command line:
./image_final_gui
step 3: it will generate a window with two button on the top (The GUI is shown below, the blue contor is the segmentation results and the red lines are the joint space distance on both medial side and lateral side), then click "import" to import the dicom image.
step 4: wait for about 30 seconds, the results will show on window.
step 5: click "save" button to choose the path to save the png and txt result.
step 1： go to the exe folder, for example: /image_final_gui
step 2: type in the command line:
image_final_gui.exe
the next steps are the same as Mac version.
step 1: setup docker environment on linux. run dockerfile, modify the path under run_anydevice.sh, and use command line to run the bashfile.
export DOCKER_DIRECTORY='path to test_folder'
export PROJECT_DIRECTORY='path to test_folder'
export DOCKER_IMG=image_new
export CONTAINER_NAME=knee_project_container_new
export DATA_FOLDER='where the test_images is'
cd $DOCKER_DIRECTORY
docker build -t $DOCKER_IMG -f Dockerfile .
nvidia-docker run -d -it \
--name $CONTAINER_NAME \
--shm-size=14g \
-v $DOCKER_DIRECTORY:/Dockerspace \
-v $PROJECT_DIRECTORY:/workspace \
-v $DATA_FOLDER:/data \
$DOCKER_IMG
docker exec -it $CONTAINER_NAME /bin/bash
step 2: go to the root folder.
run command:
python image_final_gui.py
run:
python image_final_pipeline.py --img-name --png-path --txt-path
If you make some modifications on the source codes and want to bundle them into an updated executable version, you could use pyinstaller
pip install pyinstaller
Then modify the pathex under *.spec file and then pack the software by yourself, for example
pyinstaller image_final_gui.spec
@misc{https://doi.org/10.48550/arxiv.2203.08914,
doi = {10.48550/ARXIV.2203.08914},
url = {https://arxiv.org/abs/2203.08914},
author = {Gu, Hanxue and Li, Keyu and Colglazier, Roy J. and Yang, Jichen and Lebhar, Michael and O'Donnell, Jonathan and Jiranek, William A. and Mather, Richard C. and French, Rob J. and Said, Nicholas and Zhang, Jikai and Park, Christine and Mazurowski, Maciej A.},
keywords = {Computer Vision and Pattern Recognition (cs.CV), Artificial Intelligence (cs.AI), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {Automated Grading of Radiographic Knee Osteoarthritis Severity Combined with Joint Space Narrowing},
publisher = {arXiv},
year = {2022},
copyright = {Creative Commons Attribution Non Commercial Share Alike 4.0 International}
}