Premium
Poster — Thurs Eve‐12: A needle‐positioning robot co‐registered with volumetric x‐ray micro‐computed tomography images for minimally‐invasive small‐animal interventions
Author(s) -
Waspe AC,
Holdsworth DW,
Lacefield JC,
Fenster A
Publication year - 2008
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.2965931
Subject(s) - fiducial marker , imaging phantom , computer vision , artificial intelligence , robot , computer science , image registration , segmentation , centroid , robotic arm , nuclear medicine , medicine , image (mathematics)
Preclinical research protocols often require the delivery of biological substances to specific targets in small animal disease models. To target biologically relevant locations in mice accurately, the needle positioning error needs to be < 200 μm. If targeting is inaccurate, experimental results can be inconclusive or misleading. We have developed a robotic manipulator that is capable of positioning a needle with a mean error < 100 μm. An apparatus and method were developed for integrating the needle‐positioning robot with volumetric micro‐computed tomography image guidance for interventions in small animals. Accurate image‐to‐robot registration is critical for integration as it enables targets identified in the image to be mapped to physical coordinates inside the animal. Registration is accomplished by injecting barium sulphate into needle tracks as the robot withdraws the needle from target points in a tissue‐mimicking phantom. Registration accuracy is therefore affected by the positioning error of the robot and is assessed by measuring the point‐to‐line fiducial and target registration errors (FRE, TRE). Centroid points along cross‐sectional slices of the track are determined using region growing segmentation followed by application of a center‐of‐mass algorithm. The centerline points are registered to needle trajectories in robot coordinates by applying an iterative closest point algorithm between points and lines. Implementing this procedure with four fiducial needle tracks produced a point‐to‐line FRE and TRE of 246 ± 58 μm and 194 ± 18 μm, respectively. The proposed registration technique produced a TRE < 200 μm, in the presence of robot positioning error, meeting design specification.