Abstract

Freehand ventriculostomy is one of the most common neurological procedures
performed when the cerebrospinal
uid increases in the ventricular system. This
procedure is most often performed in the emergency room or intensive care unit and
thus without a navigation system to help surgeons locate the ventricles. Surgeons
instead use anatomical landmarks on the face and skull to determine the best location
of the burr hole and trajectory for moving catheter through the brain to the ventricles
to drain excess cerebrospinal
uid (CSF) and decrease intracranial pressure (ICP).
Freehand ventriculostomy has an associated catheter misplacement rate of over 30%
which can lead to a number of complications including mortality and morbidity.
In this dissertation, we propose an augmented-reality pipeline for ventriculostomy
using an optical-see-through head-mounted device, the Microsoft HoloLens. Our system,
projects a 3D constructed model of the patient's skull and ventricles directly onto
the patient's head to guide the surgeon to locate a target on the ventricle. As part
of this pipeline, we implemented an API to send real-time tracking information from
the optical tracker to the the HoloLens, provided a manual gesture-based registration
method, as well as a colored-based depth visualization to help users understand the
spatial relationship between the patient's ventricular anatomy and surgical tool.
In a study with 15 subjects, we found that the proposed gesture-based registration
has an accuracy of 10:75 millimeters and target hitting accuracy of 12:28
millimeters. In terms of usability, our developed system received a score of 74.5
on the System usability scale (SUS), indicating that the system is easily usable.
Our preliminary results suggest that augmented-reality systems can be helpful for
neuronavigation procedures that require target localization.