| dc.description.abstract |
<p>Ophthalmology is the only surgical specialty that routinely employs both microsurgical
techniques and live intraoperative imaging, especially optical coherence tomography
(OCT). Consequently, ophthalmic surgeons routinely face challenges in visualization
and manipulation of small ocular tissues or in interpreting intraoperative imaging
to guide their actions. This dissertation therefore seeks to advance the state-of-the-art
in ophthalmic surgery with regards to manipulation, visualization, and interpretation.</p><p>First,
we address the interpretation challenge in ophthalmic surgery where live volumetric
imaging from OCT systems recently incorporated into surgical microscopes has freed
surgeons from the otherwise universal top-down viewpoint. These new viewpoints, however,
disorient surgeons when directions of their hand motions and viewed tool motions do
not align. Thus, we introduce a robotic surgery paradigm to decouple surgeons' hands
from their tools and ensure that viewed tool motions align in arbitrary viewpoints.
We implement this concept in a physical testbed system for performing macroscopic
tasks and evaluate this system through a user study with mock surgical procedures.</p><p>Next,
we consider immersive virtual reality (VR) as a technique for displaying complex images
and thereby overcome the visualization challenge of conveying intraoperative OCT to
surgeons. Far from "blinded" to the outside world, an VR-immersed surgeon potentially
has access to much more information than they could see when obligated to direct their
attention through the microscope oculars alone. To provide a compelling visual experience,
however, immersive VR systems require complete control over users' visual inputs and
thus frequently cause motion sickness with framerates lower than 90 fps per eye. By
contrast, modern volumetric OCT visualization techniques typically render at no more
than 30 fps. Therefore, we introduce GPU approaches and data organization techniques
for high-frame rate ray casting at 180 fps. We conduct performance analyses of these
techniques, develop an interactive VR-OCT viewer, and demonstrate guidance of mock
surgical procedures exclusively by live OCT and video feedthrough from within immersive
VR.</p><p>Then, we focus on deep anterior lamellar keratoplasty (DALK), a promising
technique for corneal transplantation, that poses such significant manipulation and
visualization challenges that 59% of procedures fail. In DALK, surgeons must insert
a needle 90% through the 500 μm cornea without penetrating its underlying membrane
using a surgical microscope with poor depth perception. We propose a robot-assisted
solution to jointly solve the manipulation and visualization challenges using a cooperatively-controlled,
precise robot arm and live OCT imaging, respectively. We develop this DALK workstation
with a commercial robot arm and a custom OCT scanner, evaluate its effectiveness for
cooperative needle insertions in a study with corneal fellows, and assess its ability
to perform fully automatic needle insertions.</p><p>Finally, we mitigate the visualization
challenge surgeons face when obtaining OCT images of incompletely stabilized eyes,
as happens frequently during procedures with only conscious sedation.</p><p>We introduce
a robotically-aligned OCT scanner capable of automatic eye imaging without chinrests
using a hybrid macro-mini approach. This same approach also enables an expanded ability
to image non-surgical patients when chinrest stabilization is infeasible or when a
trained ophthalmic photographer is unavailable. We validate the concept for anterior
imaging in model eyes and perform both anterior and retinal fully autonomous imaging
in human subjects.</p><p>Overall, these contributions have the potential to change
ophthalmic and other surgeries with intraoperative 3D imaging in fundamental ways.
By breaking down manipulation, visualization, and interpretation challenges, robotics
and VR promise procedures that are more efficient for patients and more ergonomic
for surgeons.</p>
|
|
