TY - JOUR
T1 - Utility-Guided Palpation for Locating Tissue Abnormalities
AU - Ayvali, Elif
AU - Ansari, Alexander
AU - Wang, Long
AU - Simaan, Nabil
AU - Choset, Howie
N1 - Publisher Copyright:
© 2017 IEEE.
PY - 2017/4
Y1 - 2017/4
N2 - Palpation is a key diagnostic aid for physicians when looking for tissue abnormalities. This letter focuses on autonomous robotic palpation for locating regions of interest representing possible tumor locations or underlying anatomy (e.g., a hidden artery). Many approaches direct the robot to exhaustively palpate the entire organ. To reduce exploration time, we define a utility function to guide the search to palpate likely regions of interest and update this function as the new palpation data are collected. The search approach, presented in this letter, incorporates prior information from preoperative images, which can provide an estimate for the location of suspicious sites, thereby reducing unnecessary manipulation and exploration of the organ. To generate search trajectories that encode a coverage goal and locate all regions of interest, two planners are adapted. The first planner is based on ergodic coverage and the second planner is based on Bayesian optimization algorithm (BOA). Both planners are evaluated via simulation and experimentally to elucidate their strengths and weaknesses. The results demonstrate a higher efficacy of the BOA planner in detecting all regions of interest while avoiding exhaustive palpation scan of the organ.
AB - Palpation is a key diagnostic aid for physicians when looking for tissue abnormalities. This letter focuses on autonomous robotic palpation for locating regions of interest representing possible tumor locations or underlying anatomy (e.g., a hidden artery). Many approaches direct the robot to exhaustively palpate the entire organ. To reduce exploration time, we define a utility function to guide the search to palpate likely regions of interest and update this function as the new palpation data are collected. The search approach, presented in this letter, incorporates prior information from preoperative images, which can provide an estimate for the location of suspicious sites, thereby reducing unnecessary manipulation and exploration of the organ. To generate search trajectories that encode a coverage goal and locate all regions of interest, two planners are adapted. The first planner is based on ergodic coverage and the second planner is based on Bayesian optimization algorithm (BOA). Both planners are evaluated via simulation and experimentally to elucidate their strengths and weaknesses. The results demonstrate a higher efficacy of the BOA planner in detecting all regions of interest while avoiding exhaustive palpation scan of the organ.
KW - Probability and statistical methods
KW - reactive and sensor-based planning
KW - surgical robotics: planning
UR - http://www.scopus.com/inward/record.url?scp=85049609216&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85049609216&partnerID=8YFLogxK
U2 - 10.1109/LRA.2017.2655619
DO - 10.1109/LRA.2017.2655619
M3 - Article
AN - SCOPUS:85049609216
VL - 2
SP - 864
EP - 871
JO - IEEE Robotics and Automation Letters
JF - IEEE Robotics and Automation Letters
IS - 2
M1 - 7827116
ER -