Deep learning for cardiac imaging
| PhD ceremony: | Mr L.B. (Daan) van den Oever |
| When: | January 17, 2022 |
| Start: | 14:30 |
| Supervisors: | prof. dr. ir. P.M.A. (Peter) van Ooijen, prof. dr. M. (Matthijs) Oudkerk, prof. dr. G.H. (Truuske) de Bock, prof. dr. R.N.J. Veldhuis |
| Where: | Academy building RUG |
| Faculty: | Medical Sciences / UMCG |
Deep learning for cardiac imaging
Cardiovascular disease (CVD) was diagnosed in 19.9 million people in Europe in 2017, causing a healthcare cost of over 210 billion euros. The burden of CVD is expected to keep increasing in following years. The European Society of Cardiology recommends to systematically screen people at risk for ischemic heart disease (IHD), so that the focus could be on prophylactic instead of reactive treatment.
A method for screening for CVD is the coronary artery calcium (CAC) score. This score can be semi-automatically calculated by segmenting the calcium spots in the coronary arteries on dual source non contrast CT images. However, individual scoring would add a significant burden to the already high workload of radiologists and technicians.
Automation of the scoring process could improve the cost-effectiveness of CVD screening and reduce the workload of radiologists and technicians. The advances in the field of artificial intelligence (AI) make AI a prime candidate for use in automation of CAC scoring. The primary aim of this thesis is to contribute to the automation of the test stage of CVD screening with CAC scoring on low-dose CT by using deep learning methods. A secondary aim of this thesis is to investigate how to evaluate such deep learning tools so that they can be implemented responsibly.