| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, Case Western Reserve University, 11100 Euclid Ave.,
MRI Bolwel B124, Cleveland, OH 44106.
2 Department of Radiology, Charité–Universitätsmedizin Berlin,
Campus Benjamin Franklin, Berlin 12200, Germany.
3 Present address: Department of Radiology, Johns Hopkins Hospital, 601 N
Caroline St., Rm. 4210, Baltimore, MD 21287-0842.
Received October 2, 2003; accepted after revision February 25, 2004.
Address correspondence to F. K. Wacker
(wackerfrank{at}web.de).
OBJECTIVE. Our aim was to test the feasibility of a hands-free approach to MRI that allows the interventionalist to track an angiographic catheter in real time throughout the procedure and to automatically change imaging parameters by catheter manipulation.
MATERIALS AND METHODS. A tracking method that is based on an active device localization was implemented on a 1.5-T MRI scanner. The system determines the current position and orientation of a catheter in 3D space in an endless feedback loop. Automatic scanning plane–adjustment procedures written in the software of the MRI system ensure image acquisition at the location of the catheter tip. The system calculates the device velocity to automatically adjust parameters such as field of view (FOV) and resolution. To evaluate the feasibility and performance in vivo and ex vivo, we performed experiments in two vessel phantoms and on six pigs.
RESULTS. The system collected the tracking data within 40 msec; an additional 10–20 msec was then required to perform the localization and velocity calculations and to update the image parameters. The system could localize a motionless catheter in the aorta in 100% and a moving catheter in 98% of measured attempts. The system responded in real time to changes in device velocity by dynamically adjusting spatial resolution and FOV in both phantom and porcine trials. Using this technique, we successfully catheterized the renal artery in two pigs.
CONCLUSION. Active tracking, combined with automatic scanning plane and imaging parameter adjustment, provides an intuitive MRI scanner interface for the guidance of the vascular procedure.
MRI is an attractive tool for guiding endovascular interventions. In contrast to X-ray fluoroscopy, MRI offers more detailed anatomic information and is not limited to evaluation of the vascular lumen, as is conventional angiography using contrast agents and radiography. In addition, functional information on the target organ is easy to obtain using MRI.
A number of different approaches are currently used for MRI-guided endovascular interventions. Susceptibility artifact–based tracking uses either paramagnetic markers attached to [1, 2] or included in [3] the catheter or a current-controlled susceptibility artifact [4–7]. Other approaches image a guidewire using the wire as a receiver coil [8–10] and image a catheter by flushing the lumen with a suitable fluid [11, 12]. With active tracking methods, signals are acquired by one or more microcoils incorporated into the interventional device [13]. These signals are sent to one of the MRI system receiver channels, where they are digitized and then used to locate the position of the receiver coil in the magnet. The position of the coil can be displayed to the operator as a cursor superimposed on a previously acquired static image [14]. In previous studies using this technique, the active feedback of the coils and hence the catheter position were used for display on previously acquired static MR angiography road maps [8, 14–16]. The principal limitation of such road maps was that they were acquired long before the actual intervention and might therefore be outdated at the time of the actual procedure; this time lapse limited their accuracy and clinical utility. With other approaches using projection techniques [17] and real-time image fusion [18], either instrument visualization and thus real-time imaging were compromised or they did not offer either interactive scanning-plane adaptation or automatic imaging-parameter adjustments.
The results of MRI-guided vascular procedures presented to date have been mainly restricted to phantom experiments and animal studies [2, 3, 12, 14, 17, 19–22]. The few patient studies already performed show numerous limitations that must be addressed before the introduction of such techniques into clinical practice [1, 23, 24].
One important shortcoming of MRI-guided vascular procedures is that a multitude of steps must be performed to find and follow a catheter during a vascular procedure. In most of the studies mentioned previously, manual techniques were used. However, even if catheter tracking can be accomplished automatically, multiple MRI parameters (e.g., TE, TR, flip angle, and resolution) still must be spontaneously adjusted during the procedure to ensure adequate catheter control. In contrast to this real-time interaction, the user interface of an MRI scanner is based on the fact that MRI has traditionally been a purely diagnostic tool in which the operator would normally sit outside the magnet and have adequate time to adjust the imaging orientation and parameters using a complex graphic user interface. Even with conventional input devices such as keyboard and mouse combined with a monitor positioned next to the scanner, operating an MRI scanner while simultaneously performing a vascular intervention is currently nearly impossible.
This study was intended to test the feasibility of a hands-free approach to an interventional MRI user interface that allows the interventionalist to see and follow the catheter (and the surrounding anatomy) in real time throughout the procedure and to automatically change a specific imaging parameter, such as the field of view (FOV), using only the catheter. This hands-free system is based on active tracking software that not only updates the scanning plane position and orientation according to microcoils mounted on the catheter but also monitors the catheter movement. Changes in catheter velocity are automatically used by the scanner to adjust a predefined imaging parameter such as FOV or resolution.
Materials and Methods
MRI Scanner Hardware and Software
All experiments were performed on a 1.5-T short-bore MRI system (Magnetom
Sonata, Siemens Medical Solutions) equipped with eight radiofrequency receiver
channels and a 40-mT/m maximal gradient amplitude and a 200 mT/m per
millisecond slew rate. All software was implemented on the scanner PC using
pulse-sequence development and image-reconstruction software environment of
the MRI vendor without any additional hardware or software requirements (other
than the tracking catheter coils we describe). The graphic user interface for
tracking and adaptive imaging was fully integrated into the standard scanner
interface using a freely configurable task card. The lack of additional
software applications or user interface helps to avoid any interference that
such applications might cause on a commercial MRI scanner.
For device tracking, an electrically active tracking method in which radiofrequency coils were incorporated into the interventional instruments was used. Customized tracking software was implemented on the scanner, which used a five-stage method [25] to accurately determine the 3D coordinates of one or two microcoils. In phase 1, the algorithm converted the raw MRI projection data into the spatial domain by applying a one-dimensional inverse Fourier transform and found the location of the microcoil in each projection. The algorithm performed marker localization using a predetermined number (N) of projections in each scanning plane. For a projection to be useful to the localization algorithm, all N marker signals should be identified in the projection. Numeric algorithms search for fiducial marker signals in the projections by identifying the largest signal peaks and then verifying that all the peaks meet an experimentally determined minimal signal-to-noise–ratio threshold. Phase 1 of the algorithm continued until enough satisfactory projections were identified. In phase 2, the algorithm analytically determined the location of all the intersections that would be created by backprojecting the marker signal peaks that were found in each projection into the scanning plane. Phase 3 designated a subset of these intersection points as references and formed a closest-point set around each of them. In phase 4, the centroids of the densest closest-point sets were used to represent the 2D locations of the microcoils. In phase 5, the final phase, a linear least-squares routine was used to match the corresponding y coordinates of the fiducial markers from the axial (x–y plane) and sagittal (y–z plane) scans. The resulting 3D position of each microcoil included the average of the two y coordinates and the x and z coordinates that were associated with the 2D position of the marker in each scanning plane [25]. The localization data were used to determine the scanning position of the next MR image. This step ensured that the catheter tip was always in the imaging plane. Furthermore, localization data from multiple time points during an interventional maneuver were collected and evaluated to determine the insertion speed of the devices. The system used the output of the speed calculation to adjust the FOV of the MRI sequence on the basis of the following equation:
|
In this equation, P(V) is the FOV value, Pmin and Pmax are the minimal and maximal values of the parameter, V is the current device speed, and S and V0 are constants adjustable via the user interface that can be used to configure the specifics of the relationship between speed and parameter value. The scanner varied the FOV in an endless feedback loop according to the catheter speed. The upper and lower parameter limits prevented pulse-sequence parameters that would exceed hardware limitations. In addition, the settings of S and V0 could be adjusted before starting the procedure via the user interface, which was integrated into the graphic user interface of the scanner software.
Interventional Device Design
Five-French straight catheters and C1 catheters were available for active
tracking. The devices were equipped with one or two single-loop radiofrequency
microcoils wound from 30-AWG (American Wire Gauge) wire mounted at the tips.
With one microcoil, the image orientation during tracking was always coronal.
When a two-microcoil catheter was used, the image orientation was adjusted so
that both coils were in the imaging plane. The dimensions of the loop elements
were 3–4 mm along the long axis and 1.5–3 mm along the short axis.
If two loops were used, the distance between the centers of the loops ranged
from 20 to 40 mm. A layer of thin transparent shrink tube was used to firmly
attach the wire loops to the catheter. The coils were tuned to the Larmor
frequency. Their tuning capacitors were placed as close to the coils as
possible to minimize resistive losses (e.g., losses in quality factor). The
circuit was matched to 50 
at the tip of the catheter. A
positive-intrinsic-negative (PIN) diode and choke were placed in the circuit
for active detuning of the resonance circuit
(Fig. 1). The purpose of
detuning was to minimize any electric currents induced in the circuit during
the excitation phase of imaging that can lead to B1 field distortions and
possible burning of the patient. The diode and choke accomplished this
detuning by shifting the resonant frequency away from the Larmor frequency
during excitation, which effectively decoupled the resonant circuit from the
transmitting radiofrequency coil. A 50- microcoaxial cable was used to
connect the devices to an individual receive-only coil port of the MRI
scanner.
|
Ex Vivo and In Vivo Experiments
To evaluate the performance of the system, we conducted ex vivo experiments
in a vessel phantom consisting of a saline-filled tube system, surrounded by
bags of saline and oil. To further assess tracking reliability and system
feasibility, we conducted animal experiments in vivo on six pigs weighing
22–43 kg. The experimental protocol was approved by our institutional
animal care and use committee. The animals were anesthetized with an
intramuscular injection of 6–10 mg per kilogram of body weight of
tiletamine hydrogen chloride (HCL) and zolazepam HCL (Telazol, Fort Dodge
Animal Health). For maintenance, ketamine and xylazine were infused, and 1 mg
per kilogram of body weight of zolazepam HCL was added intramuscularly every
45–60 min. The animals were placed in a supine position in the MRI
scanner. For MRI, as many as four phased array coils were used. The
near-real-time imaging sequence for device guidance was a real-time true fast
imaging with steady-state free precession (FISP) sequence (TR/TE, 4.5/3; flip
angle, 70°; matrix, 128 x 128; FOV, 150–300 mm; slice
thickness, 5–12 mm). The images were displayed on an in-room monitor
adjacent to the magnet.
To test the tracking precision and reproducibility of the system in vivo, we positioned a straight catheter in the abdominal aorta of each pig. One hundred image frames were then collected in which the catheter was held in a stable position in the aorta, and 100 images were obtained with the catheter moving between the aortic bifurcation and the renal arteries. An imaging protocol option in which the position of the catheter tip was set to be in the center of every updated image was chosen. Deviations in the actual catheter position in each image, as determined by observation of separation of the bright spot of the tracking coil from the center of the FOV, were used to measure the accuracy of the system.
In two pigs, a C1 catheter equipped with microcoils was used to catheterize the renal artery. The catheter passes were performed to show the tracking rate in more detail and to show the ability of the tracking system to perform a simple interventional task. In these two animals, the successful catheterization of the renal ostium was verified by injecting 5–8 mL of diluted gadopentetate dimeglumine (0.025 mol/L) and then observing the resulting change in contrast in the renal vasculature and parenchyma on the near-real-time images.
Results
Before the intervention, the tracking parameters S and Vo were chosen using the customized part of the standard graphic user interface of the MRI scanner (Fig. 2). FOV for our experiments ranged between 150 and 300 mm; V0, between 5 and 10 mm/sec; and S, between 2 and 6. After this initial setup, all interactions with the scanner were based on the different components of the automatic feedback system. As soon as the catheter was introduced through the sheath, the scanner automatically selected a scanning plane orientation that included the catheter tip. With one microcoil, the image orientation during tracking was kept coronal. When a catheter with two microcoils was used, the image orientation was adjusted so that both coils were in the imaging plane. The system determined both the current position and orientation of a catheter in 3D space within 40 msec; an additional 10–20 msec was then required to perform the localization and velocity calculations and to update the imaging parameter values. On the basis of the catheter velocity and the tracking parameter settings, the interactive feedback system responded to changes in device velocity by dynamically adjusting the FOV. The system was running continuously and was responding in real time.
|
In the phantom study, the tracking system could localize the microcoils mounted on the straight catheters in 100% of the measurements, regardless of whether the catheter was motionless or moving. In the porcine experiments, the system could localize all image frames with a motionless catheter in the abdominal aorta. Therefore, the slice position and orientation of the images based on the tracking data of the catheter remained stable throughout these imaging series. While the catheter was continuously moving in the abdominal aorta, the precise location of the coils could be determined in 98% of the images.
In both the phantom and porcine trials, the slice-plane location and orientation were automatically placed at the catheter tip using information extracted during the localization phase. The FOV was automatically changed on the basis of the calculated velocity parameters (Figs. 3A, 3B, 3C, 3D, and 3E).
|
|
|
|
|
The ostium of the renal artery could be successfully catheterized using a C1 catheter in both pigs in which this catheterization was attempted. The success was verified by injecting diluted contrast agent into the renal artery of each pig. During this procedure, the rate of successful localization dropped slightly to 95%. As the catheter approached the renal ostium, the software was set so that the reduction in catheter velocity led to a reduction in the FOV (zooming in), which helped to improve visualization of the anatomy of the renal artery.
Discussion
The application of MRI-guided endovascular interventions in a clinical setting requires the availability of features such as easy patient access within the scanner, high-speed imaging and reconstruction for real-time tracking and imaging, and intuitive interaction with the scanner. By allowing the MRI system to automatically respond to a moving catheter, we created a natural interface with the MRI scanner. All parameter adjustments were applied to imaging slices that also followed the location and orientation of the catheter tip. This application minimized the need for manual adjustments of the scanning plane position or specific imaging parameters during intravascular procedures. Thus, merely advancing the catheter more slowly automatically decreased the FOV and improved resolution of the images in this study. The feedback system can also be used for other MRI acquisition parameters such as bandwidth, temporal resolution, TE, TR, flip angle, and pulse-sequence choice (e.g., switch between FISP and fast low-angle shot). This system could even effect a total change in tissue contrast and would allow more accurate characterization of vessel wall abnormality if sequence parameters such as TE, TR, or flip angle were adjusted in relation to the catheter speed. This technique offers a completely new approach to MRI-guided intravascular procedures. The system described here uses the resources that are part of a clinical MRI scanner, such as the MRI system PC, flexible imaging sequence programming software, and graphic user interface. In contrast to previous approaches using external workstations for postprocessing [14–16, 18], no additional hardware or software, outside the tracking coils, pulse sequences, and reconstruction, is required.
The catheter localization rate in our experiments dropped from 100% in vitro to 98% in vivo and to 95% when the catheter was guided into the renal arteries. This drop most likely occurred because of the more complex environment and the flowing blood surrounding the catheter in vivo. In the case of renal catheterization, the microcoil configuration might be altered because of mechanical stress. This alteration can cause minimal detuning of the microcoils, which then increases the localization error. To minimize localization errors in a clinical setting, we can set up the system to monitor the distance that the catheter has moved between each image frame and the slice position, and adaptive parameters are not updated if the detected change in catheter position distance is greater than a predetermined value. This safe-guard ensures that a catheter localization failure during a given frame will generally not cause the slice to be placed at an incorrect position and the adaptive image parameters to be set to incorrect values.
The short-bore magnet of the scanner allowed manipulation of the catheter entering through the introducer sheath in the groin along with simultaneous scanning of the pig abdomen. However, given the current magnet length, procedures such as uterine fibroid embolization, in which the access site is usually close to the target organ, are not feasible. Therefore, open magnet designs or cylindric magnets with a shorter bore length are required. In addition to easier access, these new designs would also improve the patient comfort level because they would allow doctors and supporting personnel to approach the patients as they do during conventional angiographic procedures.
Other limitations of our study were the acquisition speed and the image quality of the imaging sequence. The use of a conventional cartesian true FISP or steady-state free precession (SSFP) sequence provided a rate of 3 frames per second, which was sufficient in speed and image quality to catheterize renal arteries. However, for cerebral or coronary artery procedures, a higher temporal resolution is necessary. Techniques using ultrashort TR or radial SSFP [22, 26, 27] are already in use to overcome this limitation. With the introduction of high-speed sequences, care must be taken that the image reconstruction is maintained at the same level to avoid delays between image acquisition and display.
The use of MRI and capacitively coupled microcoils, as described in this study, might raise some safety questions. Electric current can be induced in the wires because of the radiofrequency applied during imaging and tracking, and this current can lead to heating of the wires or local electric fields that, in turn, causes tissue heating [28–31]. Various solutions have been described to overcome this problem [30, 32]. We used safety measures such as biocompatible plastic wire and coil coating and crossed diodes or PIN diodes or both in combination with chokes for decoupling of our prototypes. Before the technique is used in humans, a complete safety evaluation is required.
In contrast to the active tracking techniques described here, passive tracking methods are technically less challenging. They have been used for iliac artery procedures [22] and renal [2, 3, 20] and coronary [17, 21] artery procedures. Although these methods might be much easier to apply in a clinical setting and on short notice, they do not use the full potential of MRI and have severe limitations. In a recent study in which the passive approach was used for coronary artery stenting, the authors concluded that passive visualization was of limited value because of its unreliable device localization [21]. Another major limitation of the passive technique is the need to search for the catheter (and other devices) in the image. Searching can be time-consuming, especially in tortuous or small vessels; it also requires significant attention from the interventionalist, whose time would be better spent concentrating on the interventional task. If a second person operates the scanner and tries to follow the catheter, every single step must be communicated during the procedure; this communication is also time-consuming and might lead to misunderstandings, especially if imaging parameters must be changed without notice. Another important drawback of the passive technique is the fact that the catheter search must be restarted if the anatomy changes (e.g., after stent insertion) or if the patient moves. Such limitations can be avoided if the MRI scanner can detect the catheter without depending on the operator.
In conclusion, the active tracking technique, combined with automatic adjustment of the scanning plane and imaging parameters as we described, provides an intuitive MRI scanner interface for the guidance of vascular procedures. This system provides one important step toward the clinical application of MRI-guided endovascular procedures.
Acknowledgments
We thank Eddy Wong for his expert technical assistance with tracking coils, Bonnie Hami for her invaluable editorial assistance, and Elena DuPont for help with manuscript preparation.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. Nazarian, A. Kolandaivelu, M. M. Zviman, G. R. Meininger, R. Kato, R. C. Susil, A. Roguin, T. L. Dickfeld, H. Ashikaga, H. Calkins, et al. Feasibility of Real-Time Magnetic Resonance Imaging for Catheter Guidance in Electrophysiology Studies Circulation, July 15, 2008; 118(3): 223 - 229. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. J. Lederman Cardiovascular Interventional Magnetic Resonance Imaging Circulation, November 8, 2005; 112(19): 3009 - 3017. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
