Optical Coherence Tomography (OCT) Image Guidance of Anterior Disease Assessment and Surgical Assistance

Abstract

Most ophthalmic diseases, such as anterior chamber diseases, require a professional ophthalmologist to diagnose. Those disease treatments may need an eye surgery operated by an ophthalmic surgeon. Both for eye disease diagnosis and treatment, physicians and surgeons need specific medical devices such as slit lamps and microscopes. Besides the required medical devices, to gain enough skills and experience for disease diagnosis and surgical operations, an ophthalmologist needs > 10 years of training starting from medical school. Optical coherence tomography (OCT), a laser-based imaging technology, has the ability to generate micron-level, cross-sectional / 4D views of biological tissue. OCT has great potential to improve eye disease diagnosis accuracy and achieve better surgery outcomes. The algorithm development of robotics, computer vision, and image processing have different implications for surgery operation accuracy, image distortion correction, and small particle size estimation. This dissertation targets developing new methods for anterior eye disease assessment and surgical assistance by combining the OCT with different robotics, computer vision, and image processing algorithms. First, I focus on automatic micro suturing such as corneal suturing to mitigate the surgeon's burden. The human cornea thickness is around 500 µm, which is challenging both for visualization and needle insertion. I propose an OCT 3D volume-based calibration algorithm that can calibrate the suturing needle frame and the robot frame at the same time. The wound detection algorithm is fully automatic in finding the needle insertion points in OCT B-scans. With known suturing needle tip pose, robot arm base pose, OCT imaging frame pose, and pre-defined needle insertion points, the path planner automatically generates the needle insertion path, and the system controls the robot arm to implement needle insertions. The measured suturing error is 0.2 mm on tissue phantom and porcine skin. Then, I target distortion-free OCT imaging to further improve anterior eye surgical outcomes. One example is deep anterior lamellar keratoplasty (DALK), a form of corneal transplant surgery that inserts a needle into corneal stroma to approximately 90% of corneal thickness. However, the corneal shape is distorted due to the light refraction and path length distortion. The needle tip is also “broken” when it is inside the cornea due to the distortion. The OCT image distortion creates extra difficulties for human surgeons and robot surgeons to access absolute spatial positioning for determining needle insertion depth. I propose a method to correct both refraction errors and path length distortion errors, and quantify the residual errors after the distortion correction. The results show that the residual errors are lower than the OCT pixel resolution. Finally, I propose small particle size estimation such as might be needed in assessment of cells in the ocular anterior chamber (anterior uveitis). Traditional anterior uveitis diagnosis relies on surgeons using a slit lamp microscope to find white blood cells in patients’ anterior chambers. Different types of uveitis involve different subtypes of white blood cells, but the slit lamp has no ability to distinguish the white blood cell subtypes. Different subtypes have different sizes. Towards a size classification approach, I developed a 3D small particle tracking algorithm to localize small particles in a chamber and then use multi-variable Mie theory for size estimation. The above contributions contain OCT-guided robot calibration and path planning for automatic surgical suturing, computer vision for light distortion correction, and image processing for small particle size estimation. These developments have the potential to improve the outcome of anterior disease assessment and surgical assistance, and increase the role of OCT in clinics and surgery operation rooms.