This work focuses on making a novel ophthalmic Optical transmission tomography (OTT) device at the lowest possible cost and size. OTT prototype demonstrates images from all the layers in anterior human eye, while also benefiting from the cost-efficient design solutions: common-path architecture, mass-market CMOS cameras, latest USB data transfer standards, Arduino electronic control, etc. Notably, we show that the large degree of noise degradation (due to the use of low-cost optics/cameras) can be corrected with denoised neural networks. Moreover, the model trained on one type of camera (global shutter) can be used to improve signal in another camera (rolling shutter).
|