Print
Univis
Search
 
FAU-Logo
Techn. Fakultät Willkommen am Institut für Informatik FAU-Logo

Stefan Ploner M. Sc.

Researcher in the Image Fusion (IMF) group at the Pattern Recognition Lab of the Friedrich-Alexander-Universität Erlangen-Nürnberg

Joint Iterative Reconstruction and Motion Compensation for Optical Coherence Tomography Angiography

Optical Coherence Tomography Angiography (OCTA) is a new modality that allows access to angiographic imaging and flow measurements in 3-D without the use of invasive contrast agents. However, image quality is degraded due to noise and inaccurate modelling thereof and many sources of artifact of which motion is the most dominant one. Motion compensation methods exist, but they do not reach their full potential in accuracy and reliability, and they are not tailored towards OCTA imaging. Still, 73% of current clinically acquired OCTA volumes suffer from significant artifacts. Hence, there is an urgent need for accurate noise and motion modelling in order to improve image quality and to reduce the number of failed scans which constitute a limiting factor to clinical workflow, diagnosis, and research.

While OCTA imaging has shown great promise for improving patient care in ophthalmology, as outlined, its full potential has been limited. There is an urgent need for robust and accurate motion detection and correction and improved OCTA processing algorithms that will decrease noise levels and improve image quality and resolution. In order to address these needs, the following objectives shall be tackled in this project:

  1. Improved motion compensation including the use of the OCTA signal for improved data consistency, more accurate and realistic motion models that operate both on a global affine scale to compensate for rotational and scaling effects and on fine scale to correct distortions, and exact modeling of the repeated scans present in OCTA scan protocols.
  2. Physically motivated OCTA signal extraction and noise modeling that employs compressive sensing-based regularization approaches in the OCTA signal extraction and integrates the full 3-D motion model from Objective 1.
  3. Precision Learning Reconstruction that augments the physically correct model from Objective 2 with additional deep learning techniques to learn data-optimal sparse domains for signal extraction.