Dissertation Cristiana Ferreira Tiago
On September 6th 2023, Cristiana Ferreira Tiago defended her thesis titled "Deep Generative Models Applied to 2D and 3D Echocardiography: Image Generation and Analysis" for the degree of Philosophiae Doctor at the Department of Informatics, Faculty of Mathematics and Natural Sciences, University of Oslo (UiO). The trial lecture was titled "Super-Resolution for medical images using AI: state of the art in CT/MRI and limitations for extending to echocardiography".
Photo: Private/UiO
Disputation: Cristiana Ferreira Tiago - Institutt for informatikk
Adjudication committee
- Professor Elsa Angelini,Telecom Paris, Institut Polytechnique de Paris, France
- Dr. Andreas Østvik, Department of Circulation and Medical Imaging, Norwegian University of Science and Technology (NTNU) and SINTEF Digital
- Associate Professor Andreas Kleppe, Department of Informatics, University of Oslo, Norway
Chair of defence:
- Professor Petter Nielsen, Department of Informatics, University of Oslo
Supervisors
- Adjunct Professor Eigil Samset, Department of Informatics, University of Oslo
- Data science leader Kristin Sarah McLeod, GE Vingmed Ultrasound AS
- AI technical leader Jurica Sprem, GE Vingmed Ultrasound AS
- Engineering leader Sten Roar Snare, GE Vingmed Ultrasound AS
Summary of the thesis
Echocardiography, i.e. ultrasound imaging of the heart, is the most frequently used modality and to be able to save clinicians’ time when analyzing such exams, automatic algorithms are being developed mainly using Deep Learning techniques. However, privacy concerns, limited data availability and variability in echocardiography images pose significant challenges for developing such models. Therefore, this thesis described the developed and applied deep generative models that can efficiently and accurately generate synthetic 2D and 3D echocardiography images, with a high realism level. Results showed that the synthetic images are realistic and are a helpful and relevant resource which can be used to develop the Deep Learning algorithms for echocardiography image analysis, this way facilitating clinical workflows, as they represent a promising step towards more efficient and effective medical imaging and diagnosis.