There are nearly 700,000 people living with primary brain tumors in the United States, with nearly 80,000 new cases and 11,000 deaths expected in 2018. Surgical removal is the first defense against brain tumors because it relieves symptoms, decreases seizure risks and increases life expectancy. Recent studies have shown that extent of tumor resection is strongly correlated with both freedom from progression and survival. However, tumors are often situated in and around critical brain structures and damaging these structures can cause loss of brain function. Thus, the primary goal in brain surgery is to maximize the extent of tumor resection while minimizing damage to surrounding brain tissue. Commercial neuronavigation systems present pre-operative image data to the surgeon during surgery that can help them visualize the location of their surgical instruments relative to these critical structures. Unfortunately, commercial systems do not compensate for progressive deformation of the brain during surgery, known as brain shift, which can be as large as 1-2 centimeters so the accuracy of neuronavigation systems decreases progressively during surgery. What is needed is a way to measure and compensate for brain shift continuously during surgery so that surgeons have timely, up-to-date information about what tissue has been removed, what tissue remains, and where nearby critical brain structures are. Such a system would enable neurosurgeons to make timely decisions that increase life expectancy and improve quality-of-life. While there is promising research in modeling brain deformation during surgery and a few end-to- end research systems that provide brain shift compensation, current approaches have a number of limitations. In particular, they rely on intraoperative data that is only available at a few time-points during surgery and they require many minutes of computation before model updates can be presented to the surgeon. This proposal addresses the three bottlenecks in state-of-the-art brain shift compensation research that prevent its adoption in clinical practice: 1) current method for modeling brain shift do not directly measure what has been removed during tumor resection so they tend to be inaccurate at the resection boundary, which is precisely where accuracy is most needed; 2) algorithms for modeling brain shift require significant preprocessing, computational power, and are too slow to provide timely feedback to the surgeon; and 3) intraoperative image acquisition is disruptive, time consuming and expensive so updates are infrequent. In this proposal we will address these bottlenecks by applying algorithms developed for real-time computer graphics to model the resection cavity and brain shift, and a new device and surgical workflow that will allow us to collect 3D ultrasound without disrupting surgery so we can monitor brain shift at frequent intervals. This project will address these bottlenecks with the following specific aims:
Aim 1. Investigate the use of Adaptively Sampled Distance Fields for modeling of tumor resection Aim 2. Extend and investigate the use of 3D Chainmail for real-time brain shift modeling from intraoperative ultrasound Aim 3. Demonstrate continuous brain shift monitoring with a new surgical workflow and a prototype intraoperative ultrasound device
During brain tumor surgery, the brain can deform by as much as 1-2 centimeters because of changes in pressure and fluid levels in the head and the removal of tumor tissue. This deformation, known as brain shift, means that advanced imaging acquired before surgery no longer represents the actual shape of the brain during surgery. This project investigates ways to compensate for brain shift in order to help surgeons remove as much tumor as possible while avoiding damage to critical brain tissue, thereby improving both life expectancy and quality of life for brain tumor patients.
