/Brain-MRI-generation-using-GANs

Generation missing MRI using GANs - master thesis from Politechnic of Milan

Primary LanguageJupyter Notebook

Politecnico di Milano - Thesis Repository

Brain Magnetic Resonance Imaging Generation using Generative Adversarial Networks - master thesis

Python PyPI

In Details

Folder structure

├──  data                       - this folder contains a batch of T1c predictions from our models. 
│   ├── batch_from_MIGAN_t1c.npy 
│   ├── batch_from_MIpix2pix_t1c.npy 
│   └── batch_from_pix2pix_t1c.npy
│ 
│
│
├──  documentation/thesis
│   ├── template LaTeX          - this folder contains src LaTex code of the thesis.
│   └── 2020_04_Alogna.pdf      - this file contains the master thesis pdf.
│ 
│
│
├── generation        - this folder contains any source of code related to the MRI generation.
│   ├── experiments   - this folder contains the experiments conducted: Skip and Internal connection analyses
│   ├── models        - this folder contains the code needed to train the 14 models of our work. 
│   ├── tests         - this folder contains the notebooks to test, qualitatively and quantitavely, the models.
│   │
│   ├── compute_baselines.ipynb            - this file contains the code to compute baselines score between modalities.
│   ├── dataset_helpers.py                 - this script contains the code to read the tf.record.
│   ├── evaluation_metrics.py              - this script contains the implemented evaluation metrics.
│   ├── generate_images_to_segment.ipynb   - this file stores all the predictions inside a tf.record.
│   └── write_tfrecord_script.py           - this script is needed to convert .mha files in tf.record.
│
│
│
├── images      - this folder contains images used in the documentation.
│
│
│
├── segmentation      - this folder contains all the files needed to segment the predictions
│   ├── models        - this file contains the models implemented and trained by [1].
│   │
│   ├── DeepMRI.py                                  - script from [1].
│   ├── SegAN_IO_arch.py                            - script from [1].
│   ├── SeganCATonColab.ipynb                       - this file allows to test the predictions contained in data folder
│   ├── dataset_helpers.py                          - script with some utility functions.
│   ├── dsc_from_generated_samples.ipynb            - this file computes DSC from the segmentation of the predictions.
│   └── segmentations_from_generated_samples.ipynb  - this file shows the qualitative results of segmentation of the predictions.

Models overview

Generated MRI samples


alt text

Segmentations using GANs predictions


alt text

References

[1] E. Giacomello, D. Loiacono, and L. Mainardi, “Brain mri tumor segmentation with adversarial networks,” 2019