Reflexiones finales

As the technique of MRI-guided prostate interventions became more popular, the need for improvements in guidance technology became evident. While the technique of using a grid template to guide needles in an open-bore MRI had proved feasible, most clinical centers only have access to closed-bore MRI scanners intended for diagnostic use. In addition, closed-bore scanners have the potential for producing much higher quality images due to increased field strengths and higher gradient performance, making them more attractive.[84] However, closed-bore MRI scanners present a challenge in the available workspace for an MRI-guided prostate procedure: the patient, along with all interventional devices, must fit within a bore of ~55 - 60 cm diameter that is generally ~1.5 - 2 m in length, while leaving enough room for physician access.[77]

In spite of these workspace constraints, Susil et al. and Menard et al. presented the results of 10 MRI-guided prostate HDR brachytherapy procedures performed on 5 patients within the bore of a 1.5 T clinical closed bore MRI scanner. Their approach employed the use of a custom-made integrated grid template and endorectal (ER) receive coil device with the patient placed in a lateral decubitus position. The authors reported good dosimetric results with their technique, which they partially attributed to the advantage of having high-field intra-treatment MR images at their disposal for needle guidance and target delineation. However, they reported a long overall procedure time (>

5 hours), and anticipated possible issues with having the patient in the lateral decubitus position, as well as instability of the prostate gland as compared to the standard lithotomy position used for brachytherapy.[84, 85]

In attempts to overcome the issue of limited physician access in a clinical MRI scanner, several researchers developed novel custom needle guidance robots. One of the first reported robotic systems for an MRI-guided prostate procedure was designed for use in an open MRI scanner, and consisted of a 5 axis linear motion module located above the MRI bore that actuated the motions of two rigid arms reaching into the bore of the scanner. The authors suggested their device could be used for the navigation of needles for prostate brachytherapy.[86] Other seminal works included a pneumatic cylinder- actuated robot presented by DiMaio et al., and a unique parallel robot called “MRI

Stealth” employing the use of newly developed pneumatic stepper motors by Muntener et al. in 2006.[87-89] The pneumatic device originally reported by DiMaio et al., which was intended for biopsy and brachytherapy, demonstrated good MRI compatibility and

evaluation in terms of positioning accuracy and repeatability, seed placement accuracy in tissue-mimicking phantoms, and tests in a canine model.[92, 93] Though clinical tests of the MRI Stealth robot (AKA MRBot) have yet to be reported, the same group indicated a possible clinical trial using the device, and development of a commercial system for brachytherapy.[94] Various devices employing novel MRI-compatible actuation

techniques were developed, including a binary robot employing a parallel arrangement of dielectric elastomer actuators, a similar concept employing newly developed “air muscle” actuators, and a concept of a wire-driven manipulator for MRI-guided prostate

cryoablation.[95-97] Fully-actuated robots that have seen use in humans include a device employing hydraulic actuation for positioning and a pneumatic needle-tapping system described by van den Bosch et al., and an ultrasonic motor-driven robot reported by Goldenberg et al.[98-101] Su et al. presented a unique design of a master-slave user- controlled robot featuring a custom-made optical force sensor for haptic feedback.[102] Recently developed systems still in the preclinical phase include a 4 degree-of-freedom (DOF) pneumatically-actuated parallel robot, first presented by Song et al., and a

piezoelectric motor-driven robot designed for the guidance of needles and adapted for the guidance of a concentric tube manipulator, demonstrated by Su et al.[103, 104]

Thesis Hypothesis and Objectives

1.6

The central hypothesis of this thesis is that an MRI-compatible mechatronic needle- guidance system, combined with a treatment planning strategy that recognizes and compensates for the uncertainties in system performance, can provide an accurate and reliable method for completely ablating focal prostate cancer targets identified on imaging. Such a method would allow an accurate appraisal of the clinical efficacy of

focal laser ablation therapy for controlling cancer in men with localized prostate cancer, and the level of treatment-related side effects associated with this technique.

1.6.1

Specific Objectives

The four primary objectives of this thesis, described respectively in each of the four main chapters, are to:

I. Develop and validate a method of accurately registering the coordinate system of an MRI-guided interventional device to that of a clinical MRI scanner under the unfavourable conditions generally found in the interventional MRI environment. II. Develop a mechatronic system for accurately guiding needles within the bore of

an MRI scanner. Verify the system’s safety and MRI-compatibility, and quantify the achievable accuracy to which it can guide needles to the prostate.

III. Use the system to perform focal laser ablation therapy in men who have consented to participate in an ongoing Phase I/II clinical trial. Quantify improvements gained in usability and clinical workflow, and quantify the achievable accuracy in needle placement.

IV. Develop a method of treatment planning for MRI-guided focal laser ablation therapy that compensates for a given level of uncertainty in needle placement error. Combined with results from the previous objective, this will lead to more precise, evidence-based patient selection criteria for focal laser ablation

eligibility, and improved treatment plans to ensure a high probability of complete focal target ablation in each case.

Outline of this Thesis

1.7

The following four chapters form the body of this thesis, and are summarized here:

1.7.1

Chapter 2: The Effects of Magnetic Field Distortion on the Accuracy

of Passive Device Localization Frames in MR

The intra-treatment magnetic resonance (MR) imaging environment presents many challenges for the accurate localization of interventional devices. In particular, geometric distortion of the static magnetic field may be both appreciable and unpredictable. This chapter aims to quantify the sensitivity of localization error of various passive device localization frames to static magnetic field distortion in MR.

Three localization frames were considered based on having distinctly different methods of encoding pose in MR images. For each frame, the effects of static field distortion were modeled, allowing errors in rotational and translational pose estimation to be computed as functions of the level of distortion, which was modeled using a first order approximation. Validation of the model was performed by imaging the localization frames in a 3T clinical MR scanner, and simulating the effects of static field distortion by varying the scanner’s center frequency and gradient shim values.

Plots of both rotational and translational error in localization frame pose estimates are provided for ranges of uniform static field distortions of 1 – 100 μT and static field

distortion gradients of 0.01 – 1 mT/m in all three directions. The theoretical estimates are in good agreement with the results obtained by imaging.

The error in pose estimation of passive localization frames in MR can be sensitive to static magnetic field distortion. The level of sensitivity, the type of error (i.e. rotational

or translational), and the direction of error are dependent on the frame’s design and the method used to image it. If 2D gradient echo imaging is employed, frames with pose estimate sensitivity to slice-select error (such as the z-frame) should be avoided, since this source of error is not easily correctable. Accurate frame pose estimates that are insensitive to static field distortion can be achieved using 2D gradient echo imaging if: a) the method of determining pose only uses in-plane measurements of marker positions, b) the in-plane marker positions in images are not sensitive to slice-select error, and c) methods of correcting in-plane error in the readout direction are employed. Results from the work in this chapter were critical to the development of the needle guidance system described in Chapter 3.