Modular deep neural networks for automatic quality control of retinal optical coherence tomography scans

Josef Kauer-Bonin, Sunil Kumar Yadav, Ingeborg E. Beckers, Kay Gawlik, Seyedamirhosein Motamedi, Hanna Gwendolyn Zimmermann, Ella Maria Kadas, Frank Haußer, Friedemann Paul, Alexander Ulrich Brandt

Abstract: Highlights•Automatic quality analysis of retinal optical coherence tomography images (AQUA-OCT).•Analysis of centering, image and signal quality as well as segmentation plausibility.•Modular, light-weight, and efficient deep neural networks, runnable on average consumer grade hardware for the analysis.•Utilizing transfer learning to make the proposed method equally effective on the two most widely used OCT devices.•Quality visualization in form of color-coded quality maps.