neuro-ml / low-resolution

Code release for the paper B. Shirokikh, A. Shevtsov et al. "Accelerating 3D Medical Image Segmentation by Adaptive Small-Scale Target Localization"

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Accelerating 3D Medical Image Segmentation by Adaptive Small-Scale Target Localization

Code release for the paper

Table of Contents

Repository Structure

[tbd]

Time performance diagrams could be built notebook/results.

The pre-trained models' weights can be found in the model folder. She source code for each of them is located at the lowres/model folder. The hyperparameters for these models (e.g., patch_size, batch_size, etc.) are stored in *.config files at config/assets/model.

Pre-trained ModelX8 (for LUNA16) could be found in ~/low-resolution/model/model_x8.pth.

Finally, the script lowres/benchmark/benchmark_time.sh can estimate the time, the given model spend to process the chosen amount of scans. By default the whole dataset used, so the desired number can be specified inside lowres/benchmark/model_predict.sh. The single argument of the script -- the desired number of threads (e.g., 8).

Installation

Execute from the directory you want the repo to be installed:

git clone https://github.com/neuro-ml/low-resolution
cd low-resolution
pip install -e .

Later we assume that the directory is ~/

LUNA16 Challenge Dataset

One of the datasets we used for models training and comparison is a LUNA16 Challenge dataset [1]. If you want to reproduce the results you need to download and preprocess it in a certain way.

Downloading

To download it use the script that will get not only the data from the challenge, but also additional lung nodules annotaions from the LIDC-IDRI database [2], that contains all the series from the LUNA16 Challenge dataset.

The only argument must be given is the absoulte path to the folder, the data will be downloaded, e.g. /mount/hdd/LUNA16_raw.

python ~/low-resolution/lowres/dataset/luna/downloader.py /mount/hdd/LUNA16_raw

If there were some problems during download, the script can be run one more time to get the missing or corrupted archives.

Extraction

After all the archives were succesfully downloaded, they need to be unpacked inside the same folder, uisng e.g. 7za:

7za x *.zip

Arranging

Experimention with the raw data is not so handy, so we will use bev library in order to prepare it. The upsides of using the library are given in the README.

First you need to setup bev storage and repository ("Creating a repository" page in the wiki). After that you can execute the prepared scripts for raw data processing:

python ~/low-resolution/lowres/dataset/luna/images.py /mount/hdd/LUNA16_raw
python ~/low-resolution/lowres/dataset/luna/lungs_mask.py /mount/hdd/LUNA16_raw
python ~/low-resolution/lowres/dataset/luna/lung_nodules_mask.py /mount/hdd/LUNA16_raw

Every script will process the corresponding instances and put the result in the folder LUNA16_processed near the folder with the raw files (in our example it is /mount/hdd/LUNA16_raw)

Finally, we need to add the procesed data to the created storage:

bev /mount/hdd/LUNA16_processed/image ~/low-resolution/assets
bev /mount/hdd/LUNA16_processed/lungs_mask ~/low-resolution/assets
bev /mount/hdd/LUNA16_processed/lung_nodules_mask ~/low-resolution/assets

After that you will see modified .hash files that points to your local storage. At this poit all the datawork is finished.

Experiment Reproduction

To run a single experiment please follow the steps below:

First, the experiment structure must be created:

python -m dpipe build_experiment --config_path "$1" --experiment_path "$2"

where the first argument is a path to the .config file e.g. "~/low-resolution/config/exp_holdout/unet3d.config" and the second is a path to the folder, where the experiment structure will be organized e.g. "~/unet3d_experiment/".

Then, to run an experiment please go to the experiment folder inside the created structure:

cd ~/unet3d_experiment/experiment_0/

and call the following command to start the experiment:

python -m dpipe run_experiment --config_path "../resources.config"

where resources.config is the general .config file of the experiment.

References

[1] Setio A. A. A. et al. Validation, comparison, and combination of algorithms for automatic detection of pulmonary nodules in computed tomography images: the LUNA16 challenge //Medical image analysis. – 2017. – Т. 42. – С. 1-13.

[2] Armato III, S. G., McLennan, G., Bidaut, L., McNitt-Gray, M. F., Meyer, C. R., Reeves, A. P., Zhao, B., Aberle, D. R., Henschke, C. I., Hoffman, E. A., Kazerooni, E. A., MacMahon, H., Van Beek, E. J. R., Yankelevitz, D., Biancardi, A. M., Bland, P. H., Brown, M. S., Engelmann, R. M., Laderach, G. E., Max, D., Pais, R. C. , Qing, D. P. Y. , Roberts, R. Y., Smith, A. R., Starkey, A., Batra, P., Caligiuri, P., Farooqi, A., Gladish, G. W., Jude, C. M., Munden, R. F., Petkovska, I., Quint, L. E., Schwartz, L. H., Sundaram, B., Dodd, L. E., Fenimore, C., Gur, D., Petrick, N., Freymann, J., Kirby, J., Hughes, B., Casteele, A. V., Gupte, S., Sallam, M., Heath, M. D., Kuhn, M. H., Dharaiya, E., Burns, R., Fryd, D. S., Salganicoff, M., Anand, V., Shreter, U., Vastagh, S., Croft, B. Y., Clarke, L. P. (2015). Data From LIDC-IDRI [Data set]. The Cancer Imaging Archive. https://doi.org/10.7937/K9/TCIA.2015.LO9QL9SX

About

Code release for the paper B. Shirokikh, A. Shevtsov et al. "Accelerating 3D Medical Image Segmentation by Adaptive Small-Scale Target Localization"


Languages

Language:Jupyter Notebook 58.7%Language:Python 41.1%Language:Shell 0.3%