Autonomous Palpation for Tumor Localization: Design of a Palpation Probe and Gaussian Process Adaptive Sampling
Animesh Garg
EECS Department, University of California, Berkeley
Technical Report No. UCB/EECS-2016-140
August 11, 2016
http://www2.eecs.berkeley.edu/Pubs/TechRpts/2016/EECS-2016-140.pdf
In surgical tumor removal, inaccurate localization can lead to the removal of excessive healthy tissue and failure to completely remove cancerous tissue. Automated palpation with a tactile sensor has the potential to precisely estimate the geometry of embedded tumors during robot-assisted minimally invasive surgery (RMIS). This thesis is a step towards enabling autonomous palpation for tumor localization. Specifically, this thesis presents a novel, low-cost design for a palpation probe and a Bayesian algorithm using Gaussian Process Adaptive Sampling for tumor localization.
First, we describe the design and evaluation of the single-use palpation probe, which we call PALP, to localize subcutaneous blood vessels. It measures probe tip deflection using a Hall Effect sensor as the spherical tip is moved tangentially across a surface under automated control. The probe is intended to be single-use and disposable and fits on the end of an 8 mm diameter needle driver in the Intuitive Surgical da Vinci Research Kit (dVRK). We report experiments for quasi-static sliding palpation with silicone based tissue phantoms with subcutaneous blood vessel phantoms. We analyze the signal-to-noise ratios with varying size of blood vessels, subcutaneous depths, indentation depths and sliding speeds. We observe that the probe can detect phantoms of diameter 2.25 mm at a depth of up to 5 mm below the tissue surface.
Secondly, we address the use of our design for autonomous tumor localization. We formulate tumor boundary localization as a Bayesian optimization model along implicit curves overestimated tissue stiffness. We propose a Gaussian Process Adaptive Sampling algorithm called Implicit Level Set Upper Confidence Bound (ILS-UCB), that prioritizes sampling near a level set of the estimate. We compare ILS-UCB to two other palpation algorithms in simulated experiments with varying levels of measurement noise and bias. We find that ILS-UCB significantly outperforms the other two algorithms as measured by the symmetric difference between tumor boundary estimate and ground truth, reducing error by up to 10x. Physical experiments with the PALP in a dVRK show that ILS-UCB can localize the tumor boundary with approximately the same accuracy as a dense raster scan while requiring 10x fewer measurements.
Advisors: Ken Goldberg
BibTeX citation:
@mastersthesis{Garg:EECS-2016-140,
Author= {Garg, Animesh},
Editor= {Goldberg, Ken and Abbeel, Pieter and Atamturk, Alper},
Title= {Autonomous Palpation for Tumor Localization: Design of a Palpation Probe and Gaussian Process Adaptive Sampling},
School= {EECS Department, University of California, Berkeley},
Year= {2016},
Month= {Aug},
Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2016/EECS-2016-140.html},
Number= {UCB/EECS-2016-140},
Abstract= {In surgical tumor removal, inaccurate localization can lead to the removal of excessive healthy tissue and failure to completely remove cancerous tissue. Automated palpation with a tactile sensor has the potential to precisely estimate the geometry of embedded tumors during robot-assisted minimally invasive surgery (RMIS). This thesis is a step towards enabling autonomous palpation for tumor localization. Specifically, this thesis presents a novel, low-cost design for a palpation probe and a Bayesian algorithm using Gaussian Process Adaptive Sampling for tumor localization.
First, we describe the design and evaluation of the single-use palpation probe, which we call PALP, to localize subcutaneous blood vessels. It measures probe tip deflection using a Hall Effect sensor as the spherical tip is moved tangentially across a surface under automated control. The probe is intended to be single-use and disposable and fits on the end of an 8 mm diameter needle driver in the Intuitive Surgical da Vinci Research Kit (dVRK). We report experiments for quasi-static sliding palpation with silicone based tissue phantoms with subcutaneous blood vessel phantoms. We analyze the signal-to-noise ratios with varying size of blood vessels, subcutaneous depths, indentation depths and sliding speeds. We observe that the probe can detect phantoms of diameter 2.25 mm at a depth of up to 5 mm below the tissue surface.
Secondly, we address the use of our design for autonomous tumor localization. We formulate tumor boundary localization as a Bayesian optimization model along implicit curves overestimated tissue stiffness. We propose a Gaussian Process Adaptive Sampling algorithm called Implicit Level Set Upper Confidence Bound (ILS-UCB), that prioritizes sampling near a level set of the estimate. We compare ILS-UCB to two other palpation algorithms in simulated experiments with varying levels of measurement noise and bias. We find that ILS-UCB significantly outperforms the other two algorithms as measured by the symmetric difference between tumor boundary estimate and ground truth, reducing error by up to 10x. Physical experiments with the PALP in a dVRK show that ILS-UCB can localize the tumor boundary with approximately the same accuracy as a dense raster scan while requiring 10x fewer measurements.},
}
EndNote citation:
%0 Thesis %A Garg, Animesh %E Goldberg, Ken %E Abbeel, Pieter %E Atamturk, Alper %T Autonomous Palpation for Tumor Localization: Design of a Palpation Probe and Gaussian Process Adaptive Sampling %I EECS Department, University of California, Berkeley %D 2016 %8 August 11 %@ UCB/EECS-2016-140 %U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2016/EECS-2016-140.html %F Garg:EECS-2016-140