Skip to content

Latest commit

 

History

History
120 lines (107 loc) · 5.61 KB

README.md

File metadata and controls

120 lines (107 loc) · 5.61 KB
Raw

Docker

Here we present the full flow of creating docker image that supports the following features:

  1. Full preprocessing for dicom and nifti files including:
    1. dicom to nifti
    2. segmentation
    3. cropping
  2. anomaly detection
  3. risk classification
  4. origin classification

Provide weights

Create a folder called web_service_models it should be like the following:

├── web_service_models
│   ├── anomaly_detection
│   │   ├── model_1
│   │   │   ├── best_val_model.pt
│   │   │   ├── config.json
│   │   ├── model_2
│   │   │   ├── ...
│   ├── risk_classification
│   │   ├── model_1
│   │   │   ├── best_val_model.pt
│   │   │   ├── config.json
│   │   ├── model_2
│   │   │   ├── ...
│   ├── origin_classification
│   │   ├── model_1
│   │   │   ├── best_val_model.pt
│   │   │   ├── config.json
│   │   ├── model_2
│   │   │   ├── ...
│   ├── cardiac_segmentation
│   │   ├── plans.json
│   │   ├── dataset.json
│   │   ├── fold_0
│   │   │   ├── checkpoint_final.pth
│   │   ├── ...

Docker Image

Build the image using the following command:

docker build . -t narco_script:latest

Run Docker

Create two folders namely input and output. Then, put input image in the input directory and then run one of the following commands.

Running cropped nifti file

docker run --rm  -v ./input:/app/input -v ./output:/app/output --gpus all narco_script:latest --input_path img-path.nii.gz --is_cropped --is_nifti

Running cropped nifti file with threshold

docker run --rm  -v ./input:/app/input -v ./output:/app/output --gpus all narco_script:latest --input_path img-path.nii.gz --is_cropped --is_nifti --threshold 0.5

Running non cropped nifti file

docker run --rm  -v ./input:/app/input -v ./output:/app/output --gpus all narco_script:latest --input_path img-path.nii.gz --is_nifti

Running with dicom zip file

docker run --rm  -v ./input:/app/input -v ./output:/app/output --gpus all narco_script:latest --input_path dicom-path.zip

Output

The output reports are in output/output.txt and output/output.json.

Expected Output for the samples:

Download the samples from the Inference on samples data and put them in the input directory, then run the following:

docker run --rm  -v ./input:/app/input -v ./output:/app/output --gpus all narco_script:latest --input_path samples/normal/10064059/img.nii.gz --is_nifti
docker run --rm  -v ./input:/app/input -v ./output:/app/output --gpus all narco_script:latest --input_path samples/narco/11943667/img.nii.gz --is_nifti
docker run --rm  -v ./input:/app/input -v ./output:/app/output --gpus all narco_script:latest --input_path samples/narco/12017913/img.nii.gz --is_nifti
docker run --rm  -v ./input:/app/input -v ./output:/app/output --gpus all narco_script:latest --input_path samples/narco/12065293/img.nii.gz --is_nifti
docker run --rm  -v ./input:/app/input -v ./output:/app/output --gpus all narco_script:latest --input_path samples/narco/12076929/img.nii.gz --is_nifti

Expected Outputs:

Note: The range is from 0 to 100.

10064059: {"anomaly": 0.0, "risk": 0, "origin": 0, "report": "No coronary anomalies(AAOCA) have been detected."}
11943667: {"anomaly": 59.41, "risk": 0.0, "origin": 0.84, "report": "R-AAOCA has been detected with anatomical low-risk features."}
12017913: {"anomaly": 100.0, "risk": 100.0, "origin": 0.49, "report": "R-AAOCA has been detected with anatomical high-risk features."}
12065293: {"anomaly": 100.0, "risk": 99.99, "origin": 0.07, "report": "R-AAOCA has been detected with anatomical high-risk features."}
12076929: {"anomaly": 82.36, "risk": 0.0, "origin": 100.0, "report": "L-AAOCA has been detected with anatomical low-risk features."}

The report.txt is like the following:

############################################################################################################################
##########################################     Model Outputs for threshold: 70    ##########################################
                              |   Anomaly Detection          |     Origin Classification    |Anatomical Risk Classification
         Probability          |            82.36             |            100.0             |             0.0              
        Final Results         |            AAOCA             |           L-AAOCA            |           Low Risk           
############################################################################################################################
################################################## Final Diagnosis Report ##################################################
R-AAOCA has been detected with anatomical low-risk features.
############################################################################################################################

Save docker image:

docker save narco_script:latest -o narco_script.tar

Load docker image:

docker load -i narco_script.tar

Large files:

For large files, increase the shared memory size of the docker container if you face any sudden docker failure or Core Dump Error.

Note: The default shared memory size is 64MB.

docker run --rm --shm-size 128MB -v ./input:/app/input -v ./output:/app/output --gpus all narco_script:latest --input_path samples/normal/10064059/img.nii.gz --is_nifti