anatomical-structures-classifier

A deep learning project

• Convolutional neural network made with PyTorch for multi-class classification of medical data.
• The dataset is composed of 20 CT scans images from 20 patients. There are 23 classes (anatomicals stuctures sush as muscles, bones and organs). Dataset is not shared.
• A cross-validation is used with 4 folds (15 training patients, 5 testing patients).
• A confusion matrix is computed and saved and accuracies / losses of each epoch (training and testing) are saved.
• Trained models are saved for anatomical structures detection.

Context

• This implementation is part of a research initiation project at INSA Lyon (1st year of telecom engineering) under supervision of researcher from CREATIS laboratory.
• This CNN model is only usable with a specific dataset. Therefore it can't be used as such.