Advanced echocardiographic image processing with machine learning Advanced echocardiographic image processing with machine learning

Echocardiography produces some of the most challenging medical data to work with due to the inherent spatial resolution. Nevertheless, it is arguably the most used images in any cardiac unit since they are non-invasive and cheap to obtain for every patient, as well as providing the best temporal resolution in 2D echo. Due to the difficulties of processing echo data comparing, most state-of the-art machine learning techniques have focused on other type of images with better resolution (e.g. CT or MRI), with a few exceptions. The main goal of this TFG is to develop an entire pipeline of cardiac echo data processing based on machine learning techniques, including tasks such as segmentation of the left atria in 2D and 3D acquisitions, detection and classification of Doppler signals, and automatic echo view recognition, among others. The output of these algorithms will be integrated in our current user interface tools processing echo data. This TFG will be in collaboration with Hospital Clínic de Barcelona and Hospital Sant Pau.


Supervisors: Oscar Camara, Jordi Mill