Use of a priori information for anatomical segmentation of medical images
Thesis in external company
External reference persons Giovanni Valbusa (email@example.com)
Research Groups DAUIN - GR-06 - ELECTRONIC DESIGN AUTOMATION - EDA
Thesis type EXPERIMENTAL, IN A COMPANY
Description Bracco Imaging is an Italian multinational active in the healthcare sector which operates in the field of Diagnostic imaging, with a large portfolio of products and solutions for X-ray imaging, including computed tomography (CT), magnetic resonance imaging (MRI), ultrasound and nuclear medicine. Bracco Imaging is currently carrying out research activities on the use of AI for the analysis and interpretation of medical images at the Italian research center located in Colleretto Giacosa (TO).
The proposed thesis is about the use of a priori information for the problem of biomedical image segmentation. Segmentation plays a crucial role in many applications of computer-based medical images analysis in particular on those requiring the automatic identification of anatomical structures or regions of interests.
Integrating anatomical priors in the segmentation process is tricky and particularly challenging when the segmentation engine is based on Convolutional Neural Networks (CNN) a class of deep neural networks. We propose a thesis work where the student is asked to apply and compare several state-of-the-art approaches to integrate a priori anatomical information with CNNs for segmentation problems.
These approaches will be tested using both images from public datasets and on more than 800 chest x-rays images manually segmented by an expert radiologist collected for the study AIforCOVID. The study AIforCOVID, promoted by the Centro Diagnostico Italiano an hospital of the Bracco Group during the COVID-19 crisis, is about the use of AI for the management of COVID-19 patients.
Dedicated hardware is available at Bracco for the development of deep-learning based projects.
Deadline 18/10/2020 PROPONI LA TUA CANDIDATURA