TMI-2006-0769

z�

Fig. 1. MATLAB Simulation of the spatial distribution of the resonant frequencies of 1 kHz, 2.5 kHz, and 5 kHz.

Abstract—A novel MR-tracking method specifically developed

to locate the ferromagnetic core of an untethered microdevice is described. The tracking method relies on the application of RF excitation signals tuned to the equipotential magnetic curves generated by the magnetic signature of the object being tracked. Positive contrast projections are obtained with reference to the position of the magnetic source. A correlation function performed on only one k-space line for each of the three axes and corresponding to three projections, is necessary to obtain a 3D location of the device. In this study, the effects of the sphere size and the RF frequency offset were investigated in order to find the best Contrast Noise Ratio (CNR) for tracking. Resolution and precision were also investigated by proper measurement of the position of a ferromagnetic sphere by MRI acquisition and by comparing them with the real position. This method is also tested for a moving marker where the positions found by MRI projections were compared with the ones taken with a camera. In vitro and in vivo experiments show the operation of the technique in tortuous phantom and in animal models. Although the method was developed in the prospect of new interventional MR-guided endovascular operations based on miniature untethered devices, it could also be used as a passive tracking method using tools such as catheters or guide wires.

Index Terms—Interventional MRI, passive tracking, RF excitation frequencies, magnetic dipole, ferromagnetic core tracking

I. INTRODUCTION

ONSISTENT and reliable tracking of an inserted device is one of the critical requirements for the success of MR-

guided endovascular interventions and as such, several methods have been reported. These methods can be categorized as passive [1]-[6], semi-active [7], or active [8]-[14]. Passive methods exploit the susceptibility difference between tissues and the inserted devices; therefore they are

This work is presently supported by a strategic grant from the Natural

Sciences and Engineering Research Council of Canada (NSERC) and in part by the Canada Research Chair (CRC) in Micro/Nanosystem Development, Fabrication and Validation and grants from the Canada Foundation for Innovation (CFI) and the Government of Québec. Asterisk indicates corresponding author.

S. Martel (phone: 514-340-4711 # 5098; fax: 514-340-3240; e-mail: sylvain.martel@polymtl.ca), J.B. Mathieu, and O. Felfoul are with the NanoRobotics Laboratory, Department of Computer Engineering and Institute of Biomedical Engineering, École Polytechnique de Montréal (EPM), Campus of the University of Montréal, Montréal, Canada.

G. Beaudoin is with the Department of Radiology, Centre Hospitalier de l’Université de Montréal (CHUM), Montréal, Canada.

image based techniques. Typically, they are simpler, regarding both instruments design and MRI integration, but slower compared to other methods. Semi-active tracking techniques are similar to passive methods except that the susceptibility difference necessary to cause a signal loss is controlled through external means. This is done through the induction of a magnetic field that dephases the MR-signal by activating a coil integrated as part of the inserted device. Active methods provide superior spatial and temporal resolutions by using projections instead of images for tracking, but require connecting the antenna to the scanner. A tremendous amount of work was done on passive and active tracking methods. For a more complete description we refer the reader to [15] and [16].

In known passive tracking methods, small paramagnetic rings are typically mounted as markers on catheters and guidewires [1]. These markers produce local field distortions appearing as areas of signal loss in MR imaging in a region surrounding the markers. Furthermore, this positioning technique by signal loss, referred to as negative contrast, is limited in regions of low proton density. More recently, a novel approach to passive tracking of paramagnetic markers has been described [2] where positive contrast of the markers to their background (white marker tracking) is exploited. A dephasing gradient is added in the direction of the slice selection during excitation to enhance the contrast between the markers and the background. This compensation gradient

In Vivo MR-tracking Based on Magnetic Signature Selective Excitation

Ouajdi Felfoul Student Member, IEEE, Jean-Baptiste Mathieu Student Member, IEEE, Gilles Beaudoin Member, IEEE, and Sylvain Martel* Member, IEEE

C

TMI-2006-0769

Fig. 3. Spin echo sequence used for tracking. There is no slice select gradient applied other than the dipole gradient (dashed lines) which is always present. The frequency of the RF excitation pulse is tuned to a given equipotential curve. Three orthogonal projections are acquired with a single RF excitation.

induces a signal loss in the image resulting in a positive contrast (instead of a negative contrast when no compensation gradient is used) where the markers appear bright on a darker background. A positive contrast can also be obtained with 4-6% doped Gd-DTPA on the instrument to generate an increase of the MR signal [3]. The Gd-DTPA is characterized by a very short longitudinal relaxation time and shows more signal than biological tissues if the images are acquired with a very short repetition time. The addition of a compensation gradient improves further the contrast between the instrument and the background. Another method for rapid MRI tracking has been described in [4] where six central k-space lines were usually sufficient to locate a needle in a plane at the cost of an increase in computation.

The tracking method described in this paper has been developed within the constraints imposed by the use of an untethered microdevice. The microdevice being developed in our laboratory contains a ferromagnetic core to allow propulsion by magnetic forces created by magnetic gradients generated by an MRI system. This technique is referred to as Magnetic Resonance Propulsion (MRP) [17]-[20]. Therefore, only a passive tracking method may be implemented for such application. Although trade-offs can be achieved between spatial resolution and refresh rate, existing passive tracking methods are far too slow in achieving acceptable performances to be integrated within the real-time constraints of MRP-based applications.

II. METHODS

A. Magnetic Signature Selective Excitation (MS-SET) Since no passive MR-tracking techniques seemed to be

adequate for MRP-based applications, a new method was developed. The proposed method referred to here as Magnetic Signature Selective Excitation Tracking (MS-SET) is based on the selective excitation of magnetic equipotential curves

caused by a magnetized marker such as the ferromagnetic core in our particular case.

The induced magnetic field from a small object with an arbitrary shape can be approximated by the field of a dipole as described by Eq.1 where µ0 = 4�10-7 H·m-1 is the permeability of free space.

( ) ( )05 3

.' 3

4m r r m

B Pr r

µπ� �

= −� �� �

� � � ��

(1)

For a uniformly magnetized sphere, the dipolar magnetic moment m

�(A·m2) is given by

343 SATm a Mπ=

�� , (2)

where SATM�

(A/m) is the saturation magnetization of the

sphere, and a its radius (m). The new magnetic core visualization method is based on the

application of an RF excitation tuned to the frequency of the desired magnetic equipotential curves as given in Fig. 1. Encoding the positions of the excited spins is done through the application of the readout gradient as it is the case with traditional imaging. A projection image of the magnetic equipotential, as depicted in Fig. 2, is therefore obtained whose shape depends on the excitation frequency as well as in the dipole magnetic signature. The projection of the given image along the read axis, corresponding to the central line of the k-space along the same axis, gives the 1D marker position. Unlike previous passive tracking methods, MS-SET requires only three k-space lines for 3D tracking, resulting into significant gain in speed compared to other mentioned passive techniques.

B. Sequence In this paper, MS-SET employs a modified Spin Echo (SE)

sequence. The slice select gradient was removed as in Fig. 3, and the frequency and the duration of the RF excitation signal were modified through customized dialog boxes in the user interface. A SE rather than a Gradient Echo (GE) implementation was chosen due to its capacity to rephase the excited spins. High readout gradient amplitude of 30 mT/m

Fig. 2. 3D FEMLAB simulation of the excited volume and corresponding projection images with the MS-SET method.

TMI-2006-0769

was used in order to minimize geometric distortions, as discussed later. Three orthogonal projections were acquired with a single RF excitation, where (TE1/TE2/TE3 = 12.5/25.5/38.5 ms).

C. Precision Experiments were done that aimed at characterizing the

tracking method by quantifying its performance criteria in terms of precision. To ensure highly reliable positions and displacements for characterization purposes, the ferromagnetic sphere was mounted on a specially made gauge structure that was fastened on top of a rectangular phantom (200×300 mm). The gauge structure made of garolite offered a positioning reference for the sphere as well as precise position increments of 10 mm ± 15 µm and 15 mm ± 15 µm along the MRI axes x and z respectively. This experimental set-up was previously calibrated and characterized using a Mitutoyo Legex 9106 Coordinate Measuring Machine (CMM). A maximum assembly imprecision of ± 25 µm resulting from differences between the nominal and actual sphere positions were recorded. This imprecision translates to 17 % of the size of a pixel and as such, it should not affect reliability in the characterization of the tracking method.

The entire volume of the phantom was filled with 7 liters of water mixed with 20 g/l gelatin, 1.25 g/l NiSO4.6H2O, and 5 g/l NaCl, providing a solid homogeneous medium with short relaxation times. The ferromagnetic core was made of a chrome steel sphere with a diameter of 1 mm. The saturation magnetization of this alloy is SATM

�= 1.36 × 106 A/m. This

value was measured with a vibrating sample magnetometer (Princeton Applied Research Corp. model 155). The magnetization of the ferromagnetic core reached saturation when placed in a 1.5 T bore of a Siemens Magnetom Avanto MRI system used for experimentation.

For each position, three projections were taken, one for each orthogonal axe. The first position, acquired when no increment was added, was considered as the reference. The measured displacement corresponded to the strongest correlation between each subsequent position and the first one.

D. Contrast Noise Ratio (CNR) The excitation frequency offset and bandwidth affect the

signal, and can be used to adjust the CNR. While low frequencies offset excite a large volume and thus give higher signal intensity, it may also excite region with high susceptibility difference, such as air-tissue interfaces, and hence cause a reduction of the CNR. On the other hand, higher frequencies increase selectivity and confine the excitation to region surrounding the marker while still decreasing the CNR by a decrease in signal intensity. Given that the CNR is important to optimize when choosing the sequence parameters, it was investigated against the excitation frequency offset for the 1 mm chrome steel sphere. The CNR was also investigated against the sphere diameter (1.5, 1.2, 1.0, 0.8, 0.5, and 0.3 mm) for a fixed RF frequency offset and bandwidth.

E. Dynamic tracking The effect of motion on the tracking was also studied. For

that purpose, the sphere was mounted on the tip of a catheter and was moved manually. The position was monitored by both the MRI scanner and a video recorder (PIXELINK CAPTURE SE camera). In order to synchronize the MRI with the recorder, the TTL output of the Siemens AVANTO MRI system was used. The trajectory given by the camera was obtained by manually sketching the pixel coordinate of the sphere in each frame.

We implemented in the Image Calculation Environment (ICE), the dedicated image reconstruction computer of Siemens, an algorithm that computes the correlation between two projections and finds the peak of the correlation [21]. The computational time required was less than 30 ms, which is the minimum feed-back time successfully implemented in the sequence for control purposes.

F. In vitro setup A Shelley vascular phantom (Shelley Industrial Automation

Inc.) placed in a container filled with water doped with a 1.25 g/l of NiSO4.6H2O was used for this experiment. A 5-F GLIDECATH catheter with a ferromagnetic sphere of 1.5 mm wedged on its tip was inserted into the phantom. The position of the catheter was visualized in real-time with custom software receiving the positions in real-time from the image calculation program and displaying them on a 2D grid. Then, these positions were superimposed on the angiogram using Matlab. The angiogram was processed using the software VolView (Kitware Inc. Clifton Park NY) in order to enhance visualization of the vessels of interest.

G. In vivo setup In vivo experiments were performed with a 35-kg domestic

pig. The study was approved by the animal care and use committee of the Centre Hospitalier de l’Université de Montréal (CHUM) as well as the committee of École Polytechnique of Montréal (EPM). During experiments, the pig was under general anesthesia. A 7-F introducer was inserted into the left femoral artery under X-Ray standard interventional procedure. Once in the MRI interventional room, a 3D angiogram with gadoteridol injection was acquired. Then, a 5-F GLIDECATH catheter with a ferromagnetic sphere of 1.5 mm wedged on its tip was inserted into the introducer by an interventional radiologist following a standard procedure. The catheter was moved inside the introducer and the positions were displayed and post-processed as described in Sect. II.F.

III. RESULTS

A. Imaging First, we show some image based on a magnetic signature

selective excitation. As depicted in Fig. 4 where high resolution (1024×1024) projection images in the axial (a) and sagittal plan (b) are shown, the excited spins lie as expected on the equipotential curve that characterizes the dipole field. The frequency offset used was 2 kHz exciting lobes at

TMI-2006-0769

X axis Z axis

Fig. 4. MRI images of the ferromagnetic sphere with MS-SET method in the transversal (a) and in the sagittal plane (b). The offset RF frequency was 2 kHz and its duration was 2560 µs. Imaging parameters was TR/TE = 150/7.5 ms, 1024×1024 and a FOV = 300×300 mm. Projections of the images in (a) and (b) along the readout direction are presented in (c) and (d).

TABLE I REAL VERSUS MEASURED DISPLACEMENTS AND CORRESPONDING ERRORS

Displacement

(mm) x y z

Real 0 15 0 Position 1 Measured 0.15 14.5 0.15

Real 0 0 15 Position 2 Measured 0.15 0 14.79

Real 0 10 0 Position 3 Measured 0.29 9.96 0

Real 0 0 10 Position 4 Measured 0.29 0 9.96

Real 0 10 15 Position 5 Measured 0.29 9.96 15.09

Real 0 15 10 Position 6 Measured 0.29 14.79 10.11

Average Error (mm) 0.24 0.13 0.10

approximately 15 times the sphere diameter. Projection along the read axis is given for the two images in Fig. 4 (c and d). These projections can be directly obtained through acquisition of the central k-space line. For the remaining of the text only projections were used in order to find the position of the marker.

Fig. 5 shows how the selection of the RF excitation frequency affects the MR images in two antagonistic ways. Increasing the offset ensures that only the induced field of the dipole is responsible for the selective excitation of the spins, however, the CNR is decreased. On the other hand, decreasing the offset may excite additional regions of high susceptibility difference such as for air-water interfaces. Fig. 5a shows a

homogeneous volume excited with a frequency of 500 Hz and a bandwidth of 1.111 kHz. The non-specific selection is due to the fact that the frequency band includes the Larmor frequency. When the frequency is increased to 800 Hz, some selectivity is noticeable as depicted in Fig. 5b, although air-water interfaces still degrade the contrast between the marker and the background. On the contrary, Fig. 5 (c and d) show images where the excitation is restricted to a volume surrounding the sphere, although increasing the frequency will decrease the CNR as depicted in Fig. 5d.

B. Precision Table I gives the real and measured displacements as well as

the average error for 6 different positions. An average value of 0.157 mm was obtained from error measurements for the three axes, corresponding approximately to half pixel precision.

Fig. 5. Effect of the RF excitation frequency offset on the excited region extent. The offsets used are: (a) 500 Hz , (b) 800 Hz , (c) 1 KHz and (d) 15 kHz . The FOV = 300×300 mm.

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 20

5

10

15

20

25

30

35

40

45

Frequency (KHz)

CN

R

(a

.u.)

Fig. 6. Contrast Noise Ratio (CNR) versus the frequency for a chrome steel sphere (1 mm diameter). The bandwidth is 250 Hz.

TMI-2006-0769

C. CNR Fig. 6 shows the CNR when varying the excitation

frequency offset from zero to 2 kHz with successive increments of 250 Hz, with each increment having a constant bandwidth of 250 Hz; hence avoiding overlap between successive RF excitations. The optimal excitation offset frequency in term of CNR was found to be 500 Hz corresponding to a frequency bandwidth extending from 375 to 625 Hz.

Fig. 7 shows the CNR for different sphere diameter obtained with an excitation bandwidth frequency extending from 300 Hz to 1.3 kHz. The CNR exhibits a cubic decrease with smaller sphere diameter as predicted by theory since the dipolar field is function of the cube of the radius of the sphere. A bead diameter of 300 µm can be detected with a CNR of

3.46, which correspond to approximately the same signal intensity as air-water interfaces.

D. Dynamic tracking As discussed earlier, the average positioning error for a

static sphere was 0.157 mm. In the dynamic regime, however, this value is found to be approximately 0.6 mm. In fact, the T1/T2 relaxation causes the CNR to decrease. Since the mean velocity of the catheter was 9.2 mm/s and the positioning refresh rate was 10 projections/second, the sphere moves by a mean value of 0.92 mm between two excitations. However, the extent of the excited region is approximately 15 mm. So, between two acquisitions more than 90% of the spins excited by the first excitation, and which do not have to be returned to equilibrium, will be solicited to contribute to the signal in the next excitation which will cause a signal decrease and hence positioning inaccuracy. The trajectories obtained through both MRI projections and by an external camera were superimposed and shown in Fig. 8.

E. In vitro and in vivo experiments Fig. 9 shows the trajectory traveled by the ferromagnetic

sphere in the Shelley phantom. The offset frequency and the bandwidth were both 1 kHz. The resolution and field of view were 256 points and 300 mm respectively. The positions were acquired at a rate of 10 positions/second. The upper bifurcation shows two paths corresponding to the inward and the outward motion of the sphere. Although we acquired simultaneously the three axes as given by the sequence shown in Fig. 3, we intentionally acquired the y axis last. Since the TE for the last axis is the highest among the three axes, the CNR is as well the lowest. This choice is made based on the fact that the vessels for human are oriented along the z axis and no significant change in the y position occurs generally. For an intervention that requires high accuracy along y, the axes order can be changed from an acquisition to the other.

Fig. 10 shows in the first column the projections acquired during the experiment for every axis, in the second, the first projections corresponding to the correlation mask and in the third column the correlation function performed on these two signals. The correlation function is performed in the complex domain. These projections were acquired in the pig with an offset frequency and bandwidth corresponding to 1.4 kHz and 1 kHz respectively. The resolution and field of view were respectively 512 points and 300 mm. The overall trajectory of the catheter is depicted in Fig. 11, where we show, on

0 0.5 1 1.50

10

20

30

40

50

60

70

80

90

(0.3,3.46) (0.5,8.1)

(0.8,26.15)

(1,42.58)

(1.2,56.34)

(1.5,87.13)

sphere diameter (mm)

CN

R (

a.u.

)

Fig. 7. CNR versus the sphere diameter for an RF excitation offset of 800Hz and a bandwidth of 1 kHz. The numbers on the plot (Sphere diameter, CNR) reproduce the exact value of the acquired data.

0 2 4 6 8 10 12 14-14

-12

-10

-8

-6

-4

-2

0

time (s)

dist

ance

(cm

)

Camera dataMRI data

Fig. 8. The trajectory traveled by the ferromagnetic sphere (1 mm diameter) as given by the camera and the MRI.

Fig. 9. Superposition of the trajectory of the marker at the end of the catheter as given by MS-SET projections on the phantom's angiogram using Matlab.

TMI-2006-0769

Fig. 10. In vivo projections: the first column represents the projection obtained after starting the tracking by 8 seconds, the catheter have traveled approximately a distance of 30 mm along the z direction and 10 mm along x. The second column represents the first projection considered as the correlation mask. The third column shows the complex correlation of the mask with the current projection. The peak in the correlation gives the displacement between the two instants of time.

anatomical images, the initial in (a) and final in (b) position of the sphere.

The positions were recorded and superimposed on the angiogram as depicted in Fig. 12. At the entry of the renal artery and due to motion and proximity to the heart and lungs, the signal of the third axis was completely spoiled. Otherwise, a good CNR was seen along the entire path.

IV. DISCUSSION AND CONCLUSION

A new method for rapid MRI passive tracking of magnetic devices has been introduced. The method requires the acquisition of only three k-space lines in order to locate a device in 3D. The resolution, precision and acquisition time achieved already overcome the drawbacks of previous passive MR-tracking methods in which acquisition and processing time as well as poor resolution prevented the use in clinical applications. The method relies on a non-selective excitation with the use of the ferromagnetic core as a marker.

The ferromagnetic marker encompasses an attraction force while in the entry region of the MRI system and while imaging gradients are applied. A 1.5 mm chrome steel sphere weighs approximately 13.5 mg. In the entry region of the MRI bore, when submitted to an extremely strong gradient of 20 T/m such a marker would undergo an attraction force as high as 300 times the gravitational attraction. The pull on the ferromagnetic marker will be directed along the longitudinal direction of the bore and its amplitude will be equivalent to

4.05 grams while being introduced into the MRI system. If the marker is mounted on a catheter, such force should not exceed the stresses generated on vessel walls during usual catheter placements or angioplasty procedures. The first reason is anatomical. The main arteries used for catheter placement in the entry region of the MRI are preferably longitudinally oriented (femoral artery, abdominal aorta). Therefore only a small fraction of the magnetic pull on the marker is oriented along the radial direction of the vessels and applied on their walls. Moreover, marker’s movement will be constrained by the stiffness of the catheter and should not damage vessel walls.

In the case of the imaging gradients applied while tracking. Their amplitude is in the order of 30 mT/m and they are applied for approximately 50% of the time. Hence, the net average force will not exceed one fourth of the gravitational attraction. Therefore, magnetic force will cause a very small movement between two consecutive acquisitions of the marker's position (especially if embedded inside a medical instrument) and its impact on CNR will be negligible.

Geometric distortions occur also in the tracking process due to the presence of the dipole’s magnetic field during the readout encoding step as shown by the sequence diagram in Fig. 3. This background field causes the location of spins to be shifted along the read axis as a function of the field strength [22]. The perturbed encoded position is given by

TMI-2006-0769

xGB

xx'

' += . (3)

In Eq. (3), x' is the resulting translated position along the axis of the readout gradient of the excited spins found at position x, B' is the value of the dipole’s magnetic field given by the RF excitation pulse, and Gx is the readout gradient along the x-axis. From Eq. 3 and for a given offset excitation frequency, the only way to reduce the effect of pixel shifting is by increasing the applied readout gradient intensity.

Tracking the marker relative to the reference axis of the

MRI system is possible if these encoding errors are compensated. Another technique consists on acquiring two projections with opposite polarity in order to locate the sphere with reference to the isocenter of the MRI system [23]. However, tracking by correlating a projection with the first one considered as the reference as we did in this paper overcome the need to correct for geometric distortions caused by magnetic field distortion.

Results of ferromagnetic marker in in vivo experiment showed that this technique translates to more complex surrounding anatomies. Although body motion can limit in some cases the number of axis that have a usable signal, alternating the order of axis on which the projections are taken can overcome the problem at the price of a reduction of positioning refresh rate. The fact that this technique requires a signal emitting tissue made it problematic for artery near the surface of the body and lung. Saline bags can be stuck in region where the artery is near the surface in order to correct in part for this problem. The use of a ferromagnetic device in MRI constitutes a source of susceptibility artifact that degrades the image quality. Until an effective way of susceptibility artifact correction is developed, the navigation of the device is made possible through superposition of the position on a pre-acquired road map. However, ferromagnetic materials are good candidates to achieve smaller marker size [24]. Even if in many situations the interventionist wants to see more than the end of the catheter, we limited this study to a single marker.

From the in vivo experiment, it is seen that the properties of the background define the lowest excitation frequency that can be used without significant background excitation. In fact, an offset excitation frequency of 1.4 kHz was required for the in vivo instead of 1 kHz for the in vitro test. Thus the properties of the background in conjunctions with other parameters like

a b

Fig. 11. In vivo images using the TRUFI_IRTTT real-time sequence provided by Siemens showing the path traveled by the catheter. (a) shows the initial position of the ferromagnetic sphere and (b) shows the final position.

Fig. 12. Superposition of the trajectory as given by MS-SET projections on the pig’s angiogram using Matlab. The points represent the hypothetical path by which the introducer passes.

TMI-2006-0769

the antenna sensitivity and the B0 value play a role in defining the minimum size that can be detected.

In conclusion, a tracking technique for a ferromagnetic core of an untethered microdevice and which can be extended to catheters and guide wires has been developed. In order to still be able to image and locate the instrument with MS-SET it may be possible to use coils embedded in the instrument to generate the dipole rather than a ferromagnetic material. Another design possibility consists of filling a balloon catheter with a highly magnetized liquid such as a ferrofluid, which can be taken out of the balloon whenever imaging is required. Otherwise this technique is well suited for tracking an untethered device propelled by the force generated by the MRI system gradient coils.

ACKNOWLEDGMENT

This project is supported in part by a Canada Research Chair (CRC) in Micro/Nanosystem Development, Fabrication and Validation, the Canada Foundation for Innovation (CFI), the National Sciences and Engineering Research Council of Canada (NSERC), and the Government of Québec. We acknowledge Dr. G. Soulez for conducting the chirurgical procedure and E. Aboussouan for helping with the dynamic tracking with the camera.

REFERENCES [1] R. van der Weide, C.J.G. Bakker, and M.A. Viergever, “Localization of

intravascular devices with paramagnetic markers in MR images,” IEEE Trans. Med. Imag., vol. 20, no. 10, pp. 779-785, Oct. 2001.

[2] J-H. Seppenwoolde, M. A. Viergever, and C. J. G. Bakker, “Passive tracking exploiting local signal conservation: The white marker phenomenon,” Magnetic Resonance in Medicine, vol. 50, no. 4, pp. 784-790, Oct. 2003.

[3] O. Unal, F. R. Korosec, R. Frayne, C. M. Strother, and C. A. Mistretta, “A rapid 2D time-resolved variable-rate k-space sampling MR technique for passive catheter tracking during endovascular procedures,” Magnetic Resonance in Medicine, vol. 40, no. 3, pp. 356-62, Sep. 1998.

[4] E. Kochavi, D. Goldsher, and H. Azhari, “Method for rapid MRI needle tracking,” Magnetic Resonance in medicine, vol. 51, no. 5, pp. 1083-1087, May 2004.

[5] C.J.G. Bakker, R.M. Hoogeveen, J. Weber, J.J. van Vaals, M.A. Viergever, and W.P. Mali, “Visualization of dedicated catheters using fast scanning techniques with potential for MR-guided vascular interventions,” Magnetic Resonance in Medicine, vol. 36, no. 6, pp. 816-820, Dec. 1996.

[6] R. van der Weide, K.J. Zuiderveld, C.J.G. Bakker, T. Hoogenboom, J.J. van Vaals, and M.A. Viergever, “Image guidance of endovascular interventions on a clinical MR scanner,” IEEE Trans. Med. Imag., vol. 17, no. 5, pp. 779-785, Oct. 1998.

[7] A. Glowinski, J. Kursch, G. Adam, A. Bucker, T. G. Noll, and R. W. Gunther, “Device visualization for interventional MRI using local magnetic fields: basic theory and its application to catheter visualization,” IEEE Trans. Med. Imag., vol. 17, no. 5, pp. 786-793, Oct. 1998.

[8] C.L. Dumoulin, S.P. Souza, R.D. Darrow, “Real-time position monitoring of invasive devices using magnetic resonance,” Magnetic Resonance in Medicine, vol. 29, no. 3, pp. 411-15, March 1993.

[9] D.R. Elgort, J.L. Duerk, “A review of technical advances in interventional magnetic resonance imaging,” Academic Radiology, vol. 12, no. 9, pp. 1089-99, Sept. 2005.

[10] D.R. Elgort, E.Y. Wong, C.M. Hillenbrand, F.K. Wacker, J.S. Lewin, and J.L. Duerk, “Real-time catheter tracking and adaptive imaging,” J. Magnetic Resonance Imaging, vol. 18, pp. 621–626, 2003.

[11] Q. Zhang, M. Wendt, A.J. Aschoff, L. Zheng, J.S. Lewin, and J.L. Duerk, “Active MR guidance of interventional devices with target-navigation,” Magnetic Resonance in Medicine, vol. 44, no. 1, pp. 56–65, July 2000.

[12] S. Zuehlsdorff, R. Umathum, S. Volz, P. Hallscheidt, C. Fink, W. Semmler, and M. Bock, “MR coil Design for simultaneous tip tracking and curvature delineation of a catheter,” Magnetic Resonance in Medicine, vol. 52, no. 1 pp. 214-218, July 2004.

[13] C. Armenean, E. Perrin, M. Armenean, O. Beuf, F. Pilleul, and H. Saint-Jalmes, “RF induced temperature elevation along metallic wires in clinical magnetic resonance imaging: Influence of diameter and length,” Magnetic Resonance in Medicine, vol. 52, no. 5, pp. 1200-1206, Nov. 2004.

[14] W.R. Nitz, A. Oppelt, W. Renz, C. Manke, M. Lenhart, and J. Link, “On the heating of linear conductive structures as guide wires and catheters in interventional MR,” J. Magnetic Resonance Imaging, vol. 13, pp. 105–114, 2001.

[15] L.W. Bartels, C.J.G Bakker, “Endovascular interventional magnetic resonance imaging,” Physics in Medicine and Biology, vol. 48, no. 14, pp. R37-R64, Jul 2003.

[16] J.L. Duerk, E.Y. Wong, J.S. Lewin, “A brief review of hardware for catheter tracking in magnetic resonance imaging,” Magnetic Resonance Materials in Physics, Biology and Medicine, vol. 13, no. 3, pp. 199-208, Jan. 2002.

[17] S. Martel, J. B. Mathieu, Y. L’Hocine, G. Beaudoin, and G. Soulez, “Method and system for propelling and controlling displacement of a microrobot in a blood vessel,” U.S. Patent 10/417,475, Apr. 15, 2003.

[18] J. B. Mathieu, S. Martel, Y. L’Hocine, G. Soulez, and G. Beaudoin, “MRI systems as a mean of propulsion for a microdevice in blood vessels,” in Proc. 25th Ann. Int. Conf. IEEE-EMBS, 2003, pp. 3419-3422.

[19] S. Martel, J. B. Mathieu, O. Felfoul, H. Macicior, G. Beaudoin, G. Soulez, and Y. L’Hocine, “Adapting MRI systems to propel and guide microdevices in the human blood circulatory system,” in Proc. 26th Ann. Int. Conf. IEEE-EMBS, 2004, pp. 1044-1047.

[20] J.-B. Mathieu, G. Beaudoin, S. Martel, “Method of propulsion of a ferromagnetic core in the cardiovascular system through magnetic gradients generated by an MRI system” IEEE Trans. Biomedical Engineering, vol. 53, no. 2, pp. 292-299, Feb. 2006.

[21] A. Chanu, S. Martel, G. Beaudoin, “Real-time Magnetic Resonance Gradient-based Propulsion of a Wireless Microdevice Using Pre-Acquired Roadmap and Dedicated Software Architecture” in Proc. 27th Ann. Int. Conf. IEEE-EMBS, 2005, pp. 5190 – 5193.

[22] A. Glowinski, G. Adam, A. Bucker, J. van Vaals, and R.W. Gunther, “A perspective on needle artifacts in MRI: an electromagnetic model for experimentally separating susceptibility effects,” IEEE Trans. Med. Imag., vol. 19, no. 12, pp. 1248-1252, Dec. 2000.

[23] E. Aboussouan, S. Martel, “High-Precision Absolute Positioning of Medical Instruments in MRI Systems,” in Proc. 28th Ann. Int. Conf. IEEE-EMBS, 2006, pp. 743 – 746.

[24] J.M. Peeters, J.-H. Seppenwoolde, L.W. Bartels, C.J.G. Bakker, “Development and testing of passive tracking markers for different field strengths and tracking speeds,” Physics in Medicine and Biology, vol. 51, no. 6, pp. N127-37, March 2006.

Ouajdi Felfoul received the B.Sc. degree in Electrical Engineering in 2003 from Sherbrooke University, Quebec, Canada and the M.S. degree in Biomedical Engineering from Ecole Polytechnique de Montréal (EPM) in 2005. He is currently with the NanoRobotics Laboratory in the Department of Computer Engineering at EPM as a Ph.D. student. His current research interests are in the development of tracking techniques for interventional devices using MRI.

TMI-2006-0769

Jean-Baptiste Mathieu was born in Perpignan, France, in 1979. He received the B.Sc. degree in Mechanical Engineering in 2002 and the M.S. degree in Biomedical Engineering from Ecole Polytechnique de Montréal (EPM) in 2004. He is currently a Ph.D. student with the NanoRobotics Laboratory in the Department of Computer Engineering at EPM. His current research interests are in the development of medical applications of microdevices propelled in the blood vessels by magnetic gradients generated by Magnetic Resonance Imaging (MRI) systems.

Gilles Beaudoin received his PhD in experimental high energy physics from the University of Montréal in 1990. For the first few years, he worked in detector development with the OPAL and ATLAS collaborations at CERN. Since 1996, he is at the Centre Hospitalier de l’Université de Montréal, where he works in the imaging aspect of medical physics. He is head of the master in medical physics program at the university of Montreal, a new program that he helped instigating. Since 2002, he is president of the Association Québécoise des Physicien(ne)s Médicaux Cliniques. His research interests are functional

and interventional MRI and X ray detection systems.

Sylvain Martel (S’95-M’96) received the B.E. degree from the Université du Québec (UQTR), and the M.Eng. and Ph.D. degrees from McGill University, Montréal, Canada in 1989 and 1997 respectively, all in Electrical Engineering. Following postdoctoral studies at MIT, he was Research Scientist at the BioInstrumentation Laboratory, Department of Mechanical Engineering at MIT. He is currently Associate Professor in the Department of Computer Engineering and the Institute of Biomedical Engineering at École Polytechnique de Montréal (EPM), Montréal, Canada, and Director of the NanoRobotics Laboratory at

EPM. Dr. Martel holds the Canada Research Chair (CRC) in Micro/Nanosystem Development, Fabrication and Validation. He has over 130 refereed publications and participates in many international committees and organizations Presently, Dr. Martel leads a multidisciplinary team involved in research and development of new platforms and nano-factories based on a fleet of scientific instruments configured as autonomous miniature robots capable of high throughput autonomous operations at the molecular scale, minimally invasive tools based on microdevices propelled in the blood vessels by magnetic gradients generated by Magnetic Resonance Imaging (MRI) systems, Micro-Electro-Mechanical System (MEMS) and System-on-Chip (SoC)-based miniature instrumented robots, development if new MEMS/NEMS based on the integration of bacteria as biological components, and many other related projects. Dr. Martel’s main expertise is in the field of nanorobotics, micro- and nano-systems, and the development of novel instrumented platforms including advanced micromechatronics systems and a variety of related support technologies. He has also a vast experience in electronics, computer engineering, and also worked extensively in biomedical and mechanical engineering.