Michael Barrow, MS1, Shanglei Liu, MD2, Nelson Ho, BS1, Ryan Kastner, PhD1, Sonia Ramamoorthy, MD2. 1UCSD Department of Computer Science and Engineering, 2UCSD Department of Surgery
Introduction: Augmented reality (AR) is an emerging visualization technology that can enhance intraoperative imaging during robotic surgery by superimposing preoperative studies into the camera display. The biggest challenge of this is the lack of accurate modeling for soft tissue in 3 dimensions. Current operative video recordings only capture image feed without measurements of force or tissue surface location (ground truth 3D data), thus limiting the modeling capabilities. The system resolutions are also too low to explore tradeoffs in system accuracy and speed for practical clinical applications. Our goal is to measure ground truth values for soft tissue manipulation and use it to create a clinically relatable AR model.
Methods and Procedures: A porcine liver was used as a homogenous soft tissue model. Mechanical manipulation and incision on the liver is performed using a custom hydraulic tool attachment to the DaVinci S ? robotic surgical system (+/- 0.01mm accuracy). A laser scanner is used to sequentially scan the porcine liver (0.025 mm accuracy up to 1732 dpcm), while a pressure gauge (+/- 5mpa) measures the applied force. These scans are pre-registered to a color video feed using an Iterative Closest Point method. We use three levels of tissue preparation to stress example AR registration algorithms (table 1).
Results: The scanner and hydraulic actuator were mounted to a camera and endowrist arms of a Davinci system (see fig 1). 1.5kg of porcine liver was scanned at a rate of 1 frame per 45 seconds. The hydraulic actuator was graduated at 1mm per frame. The data sets were tested in pairs of video frames with three registration algorithms (table 2). Talc powder coated tissue was found to give the best average convergence time and the most accurate registration in our experiments.
Conclusions: Benchmark values for soft tissue in robotic surgery were easily and effectively collected through our testing algorithm. This is the first time data of this high resolution is reported in the literature. AR developers can now use these values to explore and validate algorithmic, software, and hardware methods for tissue reconstruction in clinical practice.
Presented at the SAGES 2017 Annual Meeting in Houston, TX.
Abstract ID: 80808
Program Number: P437
Presentation Session: Poster (Non CME)
Presentation Type: Poster
