| dc.description.abstract | Management of ophthalmic diseases such as diabetic retinopathy, diabetic macular edema, age-related macular degeneration, and glaucoma often require surgical intervention in order to prevent vision-loss and blindness. In general, ophthalmic surgery involves manipulation of ocular tissue under the guidance of a surgical microscope, which provides a two-dimensional view of underlying tissue. However, these devices lack depth-resolved information and therefore preclude full 3D visualization of tissue interactions during surgery.
Optical coherence tomography (OCT) is a rapid volumetric imaging modality that is non-contact, label-free, and provides high-resolution visualization of tissue microstructure. As a result, OCT has emerged as the gold standard for ophthalmic disease diagnosis and has recently been integrated into surgical microscopes for enhanced intraoperative visualization of ophthalmic tissues. However, microscope-integrated OCT technology has several barriers that limit its broad adoption: 1) current commercial systems suffer from slow imaging speeds that limit depth visualization to static cross-sectional imaging, 2) the presence of motion artifacts degrade image quality and prevent quantitative analyses from being performed, and 3) lack of automated-instrument tracking that necessitates manual adjustment of static OCT fields, thus disrupting surgical workflow and increasing operation times.
The work presented in this dissertation addresses the aforementioned limitations by enabling real-time guidance of ophthalmic surgery. In particular, a high-speed multimodal ophthalmic imaging system is optimized and validated for in vivo motion tracking and optimized scanning protocols are developed for video-rate 4D visualization of ophthalmic surgical maneuvers. By leveraging real-time motion tracking and advancements in deep-learning-based object detection, an automated framework is developed for high performance instrument-tracking and 4D imaging of surgical maneuvers in both phantoms and ex vivo porcine eyes. The proposed motion tracking methods enable video-rate 4D visualization of bulk interactions between the surgical instrument and underlying tissue that can potentially be used to improve postoperative patient outcomes and to facilitate surgeon performance assessment and ophthalmic surgical training. | |
