Surgical pathology-enabled augmented reality for head and neck tumor resection with deformation modeling

Vanderbilt University, Nashville TN

Investigators

Abstract

Project Summary The importance of achieving a clear margin in head and neck surgery is emphasized by the fact that a positive margin status increases all-cause mortality by 90%. Yet positive margin rates in oral cavity cancers are among the highest of solid malignancies and have not improved in the past two decades. Of the 65,000 cases each year, up to 78% will undergo surgical treatment. Surgical treatment generally involves lump resection of the tumor with a margin. After initial resection, the resected specimen is sent to pathology to analyze margins through frozen section analysis. Due to the importance of clear margins, the current standard of care is to re-resect if there is a close or positive margin. Unfortunately, studies have found surgeons to be 10.3-20.6 mm off in relocating margin and additional cancer is only found in 20-30% of the cases. Patients with initial positive margins re-resected to negative margins have local recurrence rates similar to those of patients with ?nal positive margins. To improve re-resection success, we propose an augmented reality (AR) guidance system with deformation modeling to aid the relocation task. Once the surgeon completes the initial resection, we 3D scan the resected specimen to create a mesh of the tumor. An RGBD camera mounted above the patient will additionally capture a 3D point cloud of the resection bed. Using the resection bed point cloud as the target, the 3D scanned mesh will be deformed to match. The deformation will be driven by our previously proposed linear iterative boundary reconstruction models. Additionally, we will extend the models to account for additional mechanisms that drive deformation in head and neck specimens and develop novel methods to quantify uncertainty in the deformed mesh. This mesh will be uploaded to an AR system so that a projection of the deformed specimen with pathology annotations appears on the resection bed to aid relocation. To ensure clinical translation, we will conduct co- design studies with surgeons and surgical trainees to discover what features they consider important for such a guidance system. The interviews will be coded and feedback will be used to re?ne the AR interface. This proposal will be evaluated based on two phases. In the ?rst, surgeons with little AR guidance system experience will be asked to resect and relocate margins on phantoms that experience no deformation to test how well the AR interface aids relocation. In the second phase, a surgeon with extensive experience with our AR guidance system will perform resection and margin relocation in cadavers that experience deformation. This phase tests how well the deformation model aids relocalization. Together, these two phases represent the upper bound of how well the proposed system can aid margin relocation. The endpoint of this R01 will be 1) a deformation model for head and neck tumor specimens and 2) a validated AR guidance system to guide re-resection. This study will generate the necessary experimental data to power clinical trials in patients to evaluate margin relocation. As our AR guidance system does not require integration with operating room equipment, it could in principle adapted to bene?t other solid malignancies.