cseka7 / chicken_ct_weights

Multi-atlas based weight estimation for chicken from CT images.

Case study of estimating the weights of valuable meat cuts from the CT images of chicken

About

This repository implements the analysis and application described in the paper

@article{Csoka2021,
    author={\'Ad\'am Cs\'oka and Szilvia Eszter Simon and Tam\'as P\'eter Farkas and S\'andor Sz\'asz and Zolt\'an S\"ut\''o and \"Ors Petneh\'azy and Gy\"orgy Kov\'acs snd Imre Repa and Tam\'as Donk\'o},
    title={ESTIMATION OF THE VALUABLE BROILER CHICKEN MEAT PARTS MASS FROM CT IMAGES USING ELASTIC REGISTRATION},
    year={2024}
}

Preprint:

Contents

Jupyter notebooks

1. `000_extract_training_features.ipynb` - segmentation by registration and feature extraction.
2. `001_training.ipynb` - feature subset and regressor parameter selection.
3. `002_analysis.ipynb` - the statistical analysis of the results, reproducing all the tables in the paper.
4. `003_orchestration.ipynb` - executes all notebooks

Other files

1. `config.py` - high level configuration parameters.
2. `requirements.txt` - package requirements.
3. `results.csv` - raw results of the regression analysis with feature selection.
4. `results.pickle` - raw results of the regression analysis with feature selection in pickle format.
5. `results.csv` - raw results of the regression analysis without feature selection.
6. `results.pickle` - raw results of the regression analysis without feature selection in pickle format.
7. `thigh.tex` - collected results of the thigh in tex (typesetting language) format.
8. `brest.tex` - collected results of the brest in tex (typesetting language) format.
9. `chicken_dissected_data.xlsx` - results of the dissection study.

Reproducing the results of the paper

Installation

Clone the `maweight` Python package (https://github.com/cseka7/maweight):

> git clone https://github.com/cseka7/maweight.git

Navigate into the root directory of the `maweight` repository and issue

> pip install .

Clone this package (chicken_ct_weights):

> git clone https://github.com/cseka7/chicken_ct_weights.git

Navigate into the root directory of this package, and issue

> pip install -r requirements.txt

Download the raw data

Download the CT images corresponding to the dissection study and the manual annotations from the link https://drive.google.com/file/d/1yz67G03E_avm-FGZ96m5gBxpHTy4DVRs/view?usp=sharing and extract its contents to the `data` directory.

Update the paths

Update the paths in the file `config.py` to match the environment the code is running in.

Execute the notebooks

Start a jupyter server in the active environment by issuing

> jupyter notebook

And run the notebook `003_orchestration.ipynb` to carry out all steps of the analysis.

Note that due to the large number of CT images and registered masks, the execution requires about 40Gb free space on the disk.