Medical_Image

Structure

• performance: Images with the results of the different approaches
• project: The code of the project
  • Autoencoder_aproach: Files used to train and test the different autoencoders. Also, to generate images with the autoencoder
  • CNN_approach_files: Files used on the fine-tuning of CNNs approach to classify patients
  • Classifier_patches_pacients: Folder with the scripts to classify patches in the autoencoder approach and patients in both approaches
  • models: Folder with a script with different autoencoders
  • setups: Setups used to train the autoencoders
  • setups_clf: Setups used to fine-tune the classifiers
  • autoncoder_vs_CNN.ipynb: Comparison of parameters and inference time of both models.
  • dataset.py: different pytorch datasets used
  • metrics.py: Metrics used as reconstruction error in the autoencoder approach
  • partition_dataset.ipynb: Notebook to split the patients in train and test
  • preprocessing.ipynb: Notebook to do the image preprocessing
  • utils.py: Different helper functions used
  • visualize_annotated_classes.py: Visualization of the autoencoder generated images per class

Results

• Results on patches
• Results on patients

Performance Autoencoder approach:

Patch classification

Patient classification

Performance CNN approach:

Patch classification

Patient classification