Intraoperative stereo imaging for oral specimen orientation and margin visualization

ABSTRACT Management of positive surgical margins in the resection of oropharyngeal squamous cell carcinoma (OPSCC) is an imprecise science, yet its results drastically affect patient outcomes. New cancers of the oral cavity and pharynx are diagnosed at 11.4 cases per 100,000 Americans annually, with incidence rising steadily due to HPV- mediated disease. Most patients with early-stage OPSCC are treated surgically; however, positive surgical margins (PSMs) complicate approximately 30% of traditional and 15% of robot-assisted procedures. Re- localization of positive margins is important for both intraoperative management with frozen section guidance, and for postoperative planning of adjuvant radiation therapy. In both cases, inaccurate orientation and localization can lead to incomplete treatment and excessive iatrogenic damage to healthy tissue, resulting in reduced local control and overall survival. We propose a novel, robot-integrated system for intraoperative creation and annotation of digital 3D models of resection specimens and the residual tumor bed. These models will employ visual feature correspondences to warp between the shape of the flattened specimen and the in situ oropharyngeal anatomy. As such, positive margins identified intraoperatively by frozen section or postoperatively by permanent section can be accurately mapped to the specimen and also to the in situ anatomy. We hypothesize that this digital pathology solution will integrate seamlessly into the transoral-robotic surgery (TORS) workflow and provide the care team with qualitative and quantitative margin localization data that will 1) improve clinical communication around specimen orientation and margin context, and 2) enable more precise, patient- specific adjuvant therapies, ultimately reducing recurrence and improving overall survival. In Aim 1, we will develop 1) the necessary workflow from stereovision to 3D specimen models, 2) in-console annotation functionality, and 3) model deformation capabilities. In Aim 2, we will conduct a single-site human subjects feasibility and effectiveness evaluation. At the end of the performance period, we will deliver a robust, robot- integrated specimen orientation and margin visualization system ready for implementation in a multi-site prospective study designed to quantify the effects of the proposed system on local control and survival in TORS.