NeuroSpark (AR/VR)

Spring 2024 - Present
Partners: Dr. Andrew Abumoussa, UNC Medical School/MD-PhD Program

NeuroSpark is developing a tool to improve patient experiences and outcomes during awake craniotomies—a common neurosurgical procedure for removing brain lesions such as tumors. In an awake craniotomy, the patient remains conscious, allowing neurosurgeons to test critical functions (e.g., motor control, memory) to minimize the risk of functional loss while removing as much of the lesion as possible.

Currently, these tests involve tasks like playing a musical instrument or recognizing famous figures on a printout, but these methods have limitations and can cause significant stress as patients witness the operating environment. NeuroSpark aims to address this by placing patients in a virtual environment using a VR headset (e.g., Meta Quest) during surgery. While the patient’s head must remain still, they will interact with the virtual environment, which includes an agent designed to assist the neurosurgeons.

The agent will be provided x, y, z coordinates of the incision tool and tasked with identifying the brain region undergoing surgery at the voxel level and providing relevant tasks to assess the patient’s function in real-time. Based on task performance, the agent will “grade” the patient’s abilities, informing the neurosurgeon about the potential impact of further surgery in that area. This approach will leverage the existing neuro-literature on task-based assessments and “translate” these into a format for virtual reality to reduce patient stress and improve surgical outcomes.

Additionally, NeuroSpark is collaborating with the Biomedical Research Imaging Center (BRIC) on a second part of the project, which involves probabilistically mapping brain function. This will not be worked on by a majority of team members but it is still crucial to the success of the project. Using a large dataset of fMRI scans from the Human Connectome Project (HCP), a general model will be trained such that it can be fine-tuned via transfer learning to an individual’s brain function. This is critical, as patients undergoing surgery often present with physical and mental abnormalities that would make their brain differ from any kind of generalized model.

Several of the teams will have publishing goals and the project has been provided lab space on the med school campus for members of the team to work and meet in.