This project is still under construction

The code is based on MONAI (AI Toolkit for Healthcare Imaging).

Env configuration

pip install monai
conda create -n unest python=3.7
conda activate unest
pip install -r requirements.txt

Datasets

Figure 1 Lung cavity weakly supervised semantic segmentation dataset.

Our dataset is available in the dataset. As the paper is still under review, the dataset has been encrypted.

Please look at the Data Description below for detailed information about the dataset.

After decompressing the dataset, you can get the following directory:

├── Train_data
│ ├── Training_Dataset
│ │ ├── TRN_00.nii.gz
│ │ ├── TRN_000.nii.gz
│ │ ├──  …

Training Data (ImageCLEF training data) Preparation

We need to download the following datasets:

ImageCLEF 2022 Tuberculosis - Caverns Report and Caverns Detection

and unzip the Cavern Detection Train CT files (1 to 7) and the Cavern Report Train CT files (1 to 2):

-------------------------------------------------------------------------------------
|                     Zip files                                    Unzip files      |
-------------------------------------------------------------------------------------
| de0e8772-594d-41ce-9e85-578c4b59e9f3_detection_train_CT_1    (TRN_000 - TRN_099)  |
| 29cc320d-1c9d-4c7c-8d12-a006499c2f2f_detection_train_CT_2    (TRN_100 - TRN_179)  |
| 3b9da027-6015-4806-bbe9-04efb760ee53_detection_train_CT_3    (TRN_180 - TRN_269)  |
| c45b6a20-e3dc-49b5-bd93-1a86dc10721c_detection_train_CT_4    (TRN_270 - TRN_339)  |
| 447557ed-45dd-4468-8381-dbf0642b4312_detection_train_CT_5    (TRN_340 - TRN_419)  |
| f9da57c9-dbb5-4a3d-bebb-0618a8aef99f_detection_train_CT_6    (TRN_420 - TRN_499)  |
| eac9b86a-7673-4b24-924a-29529de6f130_detection_train_CT_7    (TRN_500 - TRN_558)  |
------------------------------------------------------------------------------------
| 45037ba5-e1e7-4011-98c6-35ed190e204a_cavern_report_train_CT_1   (TRN_00 - TRN_29) |
| e222bfea-28db-4d3a-b7cf-b68a7e18992b_cavern_report_train_CT_2   (TRN_30 - TRN_59) |
-------------------------------------------------------------------------------------

they will be placed in the following directory:

├── Train_data
│ ├── Original_Image_Dataset
│ │ ├── TRN_00.nii.gz
│ │ ├── TRN_000.nii.gz
│ │ ├──  …

SwinUNeLCsT

For supervised paradigm training

python main.py

For weakly supervised paradigm training

coming soon

SwinUNeLCsXt

Run the example script:

sh scripts/train_r50_SwinUNeLCsXt.sh

Contact

If you have any questions, please feel free to contact me via tan.joey@pelajar.upm.edu.my

LICENSE

This repo is under the Apache-2.0 license. For commercial use, please contact the authors.