Presentation
7 March 2022 Synthesizing degree of polarization uniformity from non-polarization sensitive OCT by convolutional neural network
Author Affiliations +
Proceedings Volume PC11941, Ophthalmic Technologies XXXII; PC1194109 (2022) https://doi.org/10.1117/12.2608395
Event: SPIE BiOS, 2022, San Francisco, California, United States
Abstract
In this research, we demonstrated a CNN-based DOPU estimation algorithm without polarization-sensitive OCT signals. The CNN was trained with pairs of retinal OCT (input) and DOPU (teaching) images. The recall and precision of RPE abnormal appearances between true DOPU and synthesized DOPU of pathological eyes were calculated. For normal eyes, the grader evaluated the soundness of the RPE appearance for true DOPU and synthesized DOPU. The recalls are relatively good (0.74-0.95), while the precisions highly depend on the types of abnormalities (0.37-0.98). Five RPE abnormalities are found in synthesized DOPU within 25 synthesized DOPU B-scans while there is no abnormality in the true DOPU.
Conference Presentation
© (2022) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Kensuke Oikawa, Shuichi Makita, Masahiro Miura, Takuya Iwasaki, Toshihiro Mino, Tatsuo Yamaguchi, and Yoshiaki Yasuno "Synthesizing degree of polarization uniformity from non-polarization sensitive OCT by convolutional neural network", Proc. SPIE PC11941, Ophthalmic Technologies XXXII, PC1194109 (7 March 2022); https://doi.org/10.1117/12.2608395
Advertisement
Advertisement
KEYWORDS
Optical coherence tomography

Convolutional neural networks

Polarization

Data acquisition

Evolutionary algorithms

Back to Top