In this thesis, we explore the effectiveness of Deep Learning techniques in attempting to automatize and improve the diagnosis of atheroma plaques. To achieve so we tackle the following problems: ultrasound image segmentation and plaque tissue classification.
- Data Cleaning
- Automatic GT generation
- Data exploration
- Comun Carotid Artery segmentation
- Tissue class 5EAD ification of atherosclerosi plaque
This project has been mostly implemented in python, only a part of the GT generation is implemented in MATLAB. For the segmentation task we used a keras fork running on Theano. The Fully Convolutional Network implemented to explore the carotid artery is based on the semantic segmentation in Keras.
For the classification of the plaque we used keras running tensorflow on the backend.
- Arnau Escapa
- Enric Sarlé
- Jonatan Piñol