Skip to content

Latest commit

 

History

38 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 

Repository files navigation

x-ray classification

Learning of chest radiographs in the application of pulmonary edema assessment.

Instructions

The main file is test_model.ipynb. This file contain preprocessed datasets, the model, and gui. The cl_model, fc_model - models with different architectures, just some useful files, they are not directly related to the model. The prep.py file contain a CustomDatases class that helps to make dataset with chest radiographs. In logs directory you can find test metrics of loss, auroc, f1-score and accuracy.

Usage

You need to import test_model.ipynb and prep.py. With IDE, download all libs, you find their in first code blocks. To load trained model, save the m4.pt file from disk in directory with this files. If app don't see m4 file, just write full path to this file.

Data

MIMIC-CHR-JPG

All data was took from MIMIC-CXR. It's large publicly available dataset of chest radiographs in DICOM format with free-text radiology reports. The dataset contains 377,110 images corresponding to 227,835 radiographic studies performed at the Beth Israel Deaconess Medical Center in Boston, MA. JPG files files was converted from this dataset.

Pulmonary edema severity

We aim to classify a given chest x-ray image into one of the four ordinal levels: no edema (0), vascular congestion (1), interstitial edema (2), and alveolar edema (3).

GUI

You can load any image on the BROWSE button. After that, model give you predcition 1 of 4 classes. You can transform image with filters in transformation tab.

Contact

If you have problems with application, contact me in tg(@dimitriy_lvov)

About

Previously, the task of classifying X-rays was solved only manually by specialists, which took a lot of time and required a lot of money. With the development of neural network architectures and the learning process, the ability to automatically classify X-ray images has improved significantly.

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Used by

Contributors

Languages