Inference of Angiography Flow Information from Structural Optical Coherence Tomography Images in Cynomolgus Monkeys Using Deep Learning

Author(s): Peter M Maloca, Philippe Valmaggia, Nadja Inglin, Beat Hörmann, Sylvie Wise, Philipp Müller, Pascal W Hasler, Nicolas Feltgen, Nora Denk

Introduction: To adapt and replicate deep learning-based methods for the automated detection of flow signals in OCT imaging, focusing on their application to the retinas of healthy cynomolgus monkeys. Methods: From 193 healthy cynomolgus monkeys, an unprecedented number of 382 coregistered OCT and OCTA stack pairs were obtained for training, evaluation, and separate testing. An adapted U-Net architecture with an additional max-pooling layer to account for the large spatial input format was used. The net was trained with an Adam Optimizer and a Mean Squared Error loss function until the loss on the validation set reached a plateau (21,000 steps).The following metrics were calculated for each OCT and OCTA B-scan pair in the test set: mean-squared error (MSE), structural similarity index (SSI), and peak signal to noise ratio (PSNR). Results: The developed deep learning method allowed to automatically detect the flow signal within the native structural OCT scans in animals. The average MSE over all test set image pairs was 0.00370368 with a standard deviation of 0.000825. Average SSI was 0.88339 with a standard deviation of 0.02167 and the average PSNR was 24.43170 dB with a standard deviation of 1.08154 dB. No large difference in the distribution of MSE, SSI, and PSNR were found among eyes and among individual cynomolgus monkeys. Conclusion: Deep learning can reliably detect retinal flow signals from standard OCT scans in healthy cynomolgus monkeys, offering a viable alternative to OCTA imaging and enabling broader access to vascular analysis in preclinical research.