A Computational and AI-Driven Approach to Refraction Correction in Optical Coherence Tomography Imaging.

This thesis tackles the problem of refraction errors in Optical Coherence Tomography (OCT) scans, which is important for getting accurate images during eye surgeries. OCT is a high-tech method that gives clear images of the inner parts of the eye, like the retina and cornea, making it very useful in procedures like cataract surgery. However, as light passes through the different layers of the eye, it is subject to the effects of refraction, which can distort the images and make them less accurate. These distortions make it difficult for Horizon Surgical Systems’ automated surgery system, used in procedures like cataract removal, to fully rely on OCT images, as important details might be missing for precise operations. Current methods can fix these distortions, but they are often slow and complicated, making them less practical for use in real-time surgery. Having to wait for these corrections would slow down the procedure, so speeding up this correction process would make OCT much more useful for surgeons. To solve this, this thesis introduces a two-part solution. The first part involves creating a computational algorithm that uses a 3D model to see how different structures in the eye cause light to bend. This helps spot and understand the specific distortions in OCT images, though it’s a slow process and not ideal for quick corrections during surgery. The second part uses the first part to create a dataset, that is then used to train an AI model to make these corrections much faster. With the AI model, the adjustments happen almost instantly, allowing the robot to know exactly where is the real position of the eye structures such as the anterior and posterior corneas without any delays. Although this system was initially made for Horizon Surgical Systems, a leader startup in the sector of AI-Driven surgeries, it could also help other organizations working with similar imaging technology. Refraction errors in OCT scans are a common issue in many fields that rely on precise imaging. By addressing this problem, this thesis helps improve eye surgeries by making real-time scan corrections possible, leading to safer and more efficient procedures.