4.1 Summary
4.1.2 Chapter 3: 3D Image-Guided Robotic Needle Positioning System for
This chapter describes the design of a robotic needle positioning systems, its integration with a commercial micro-CT scanner and the characterization of the system’s performance. The robot contains a total of 6 degrees of freedom that consist of 3 linear translational axes, 2 rotational axes and a linear needle driver. The two rotational axes
are created using a kinematic frame based on a spherical linkage design. The rotational axes intersect at a common point in space known as a remote centre of motion (RCM). The entire system mounts onto the bed of the micro-CT scanner and is fully capable of completing interventions within the scanner bore.
A method was developed to calibrate a needle tip to the RCM of the robot. The calibration was accomplished using a calibration fixture. The calibration was validated with photography using a camera equipped with a macro lens. The calibration error was
measured to be ∆x=36 µm, ∆y=70 µm in the roll plane and ∆y=11 µm and ∆z=5 µm in the
pitch plane. The repeatability of the needle driver in positioning the needle tip was
σneedle=±9.1 µm.
A registration process with two different registration methods was developed to register the robot to the micro-CT scanner. The primary registration is the most accurate registration but also takes the most time to perform. The primary registration is also no longer accurate if the robot is removed from the micro-CT scanner bed. To allow a primary registration to be reused after the robot is removed from the micro-CT bed, it can be combined with a secondary registration. The secondary registration can be calculated quickly with only a single image but at the expense of registration quality. The primary
registration errors were FREprimary= 21 ± 6 µm and TREprimary= 31 ± 12 µm. The
secondary registration errors were FREsecondary=70 ± 25 µm and TREsecondary=79 ± 14 µm.
The error of a combined primary and secondary registration was TREcombined=139 ± 63
µm.
The targeting accuracy of the robot was next characterized using tissue- mimicking gelatin phantoms. The first set of targeting experiments consisted of targeting points in the phantom at a fixed needle angle. The accuracy was calculated by measuring the distance of the needle axis to the desired target in micro-CT images. The targeting accuracy using a primary registration was 131 ± 25 µm. The targeting accuracy using a combined registration was 206 ± 20 µm. The second targeting experiment consisted of fixing the translational axes of the robot, inserting a needle into the phantom at angles over the robot’s angular range of motion and imaging the needles using micro-CT. The
distance of the needle axes to a point of best fit was calculated. The mean distance of the needle axes to the point of best fit was 71 µm with a smallest distance of 24 µm and largest distance of 189 µm. These targeting accuracies were combined with one another and the other measured sources of error to approximate the overall targeting accuracy of the system
to be 149 ± 41 um using a primary registration and 218 ± 38 um using a combined
registration.
The chapter finally demonstrates the capability of the robot to complete selected biomedical applications. The robot was able to successfully position a probe under image guidance to perform interstitial tissue pressure measurements in a mouse tumour. The robot was also able to successfully position a needle under image guidance to contact a 1.5 mm bead implanted in dorsal subcutaneous tissue of a mouse.
The principal contribution of chapter 3 was the development of a small-animal robot compact enough to operate within a micro-CT bore and the associated methods to calibrate the robot and register it with the micro-CT scanner. Chapter 3 is the basis of a
paper in preparation for submission to the peer reviewed journal Medical Physics.
4.2 Conclusion
The field of small animal image-guided robotic systems is in its infancy. The previously developed systems in the field consist largely of initial prototypes which have seen limited adoption by their target audience of preclinical researchers. The preceding chapters have attempted to introduce a number of refinements to small animal robots to better facilitate their adoption among preclinical researchers. The focus of these refinements was achievement of a desirable targeting accuracy with minimal variability within a user friendly system.
A phantom was developed that allows for the routine evaluation of micro-CT scanners. The phantom allows for the user of the robot to quickly evaluate the geometric accuracy of a micro-CT scanner to a traceable standard and apply corrections as necessary. No other small animal image-guided robotic system has validated the geometric accuracy of its selected imaging modality. Use of the phantom provides an
important foundation towards the successful completion of image-guided interventions. Although the phantom was initially developed for use with image-guided interventions, it should prove equally useful for a wide range of micro-CT applications such as the characterization of medical devices.
The robotic system introduced in this thesis offers a number of benefits over previous designs. The robot is compact enough to operate entirely within the micro-CT bore. Specimens are not required to be moved between the imaging and robot workspace as in previous designs. This reduces the time required to complete interventions and reduces errors associated with detaching and reattaching beds. A calibration fixture was developed that allows the robot to be calibrated in a fraction of the time required to calibrate other designs. A dual mode registration method was introduced to offer the user greater flexibility between the time requirements of completing registration and the registration accuracy depending on the requirements of the application. Finally, the robot was demonstrated as achieving the desired < 200 µm targeting accuracy with reduced variability then previous designs.