. I

Technical Note

Rapid Tip Tracking with MRI by a Limited Projection Reconstruction Technique

Kenji Shimizu, MS, MD Robert V. Mulkern, PhD Koichi Oshio, MD, PhD Lawrence P. Panych, PhD Seung-Schik Yoo, BA Ron Kikinis, MD Ferenc A. Jolesz, MD

To rapidly track invasive de- vices within MRI systems, a novel approach using a lim- ited projection reconstruction technique is presented. Our method exploits the differ- ence between images recon- structed from a limited number of projections and serves to depict the tip of a needle during its advance- ment. This method was imple- mented on a standard MRI system with a radial fast- spin-echo sequence and ex- amined in phantom studies. We demonstrated that the proposed method could track the tip every 300 msec and the tip depicted by the pres- ent technique was consis- tently displaced along the needle by a small distance (5 mm) .

Index terms: Devlce (racking * Kapid imaging - Projection reconstruction * In- tewentioii.il MRI

JMRl 1998; 8'262-264

Abbreviations: = radial last spin echo

FOV = field of view. rFSE

From the Ijepartment of Ratliology, Bngham m d Womrn's Hospital (K.S , K 0.. I< P.P.. S.- S.Y.. R K , F A J.) and the Department of Ra- d~ology. Children's Hospital (R.V.M.). Harvard Mc.dical School. Boston. MA021 15. E-mail: sinliQimagt..med.osaka-u.ac.jp. Re- ceived Junc 3. 1997: rewsion requested Au- gust I , rt'vision received August 29: second rewsion rryuested Septembrr 17: second re- msion reloved October 6 accepted October 15. This paper has been presented at the mnua l scientific meeting of the Intema- tional Society for Magnetic Resonance in Medicine. \'ancouver. 1997. Address reprint requests I n K.S.. Division 01 Functional Di- agnostic Iniaging. Osaka University Medical Srhool. 2-2 Yamadaoka. Suita. Osaka 565. ,Japan

ISMRM. 1998

RECENTLY DEVELOPED open-configura- tion MRJ systems allow direct access to the patient during interventional and surgical procedures [ 1.2). During most intraopera- tive applications, interactive control of scan planes is necessary for localization and targeting. This requires the tracking of various flexible and rigid instruments and the ac,quisition of images that correspond to the position of these devices.

For tracking invasive devices within MRI systems, two approaches have been pro- posed: an active method and a passive method. In the active method, a small re- ceiver coil is built into the tip of the device to obtain three-dimensional spatial coor- dinatcs of the tracked device (3-6). These techniques require continuous use of readout magnetic field gradient pulses and use the specialized hardware. In the pas- sive niethod, direct visualization of the de- vices in MR images is attempted on the basis of the associated signal void and susceptibility effects (7,8). The dedicated devices with paramagnetic components ( 9 , l O ) allow conspicuous and consistent visualization, although they produce large artifacts, depending on orientation and imaging parameters. In this method, de- vice ti-acking is accomplished by updating images and, therefore, fast scanning tech- niques are necessary to achieve sufficient speed for tracking. To address this prob- lem, ihe keyhole technique (1 1) and its variants (12,13) have been exploited for MR-gilided catheter tracking (7.8).

Recently. projection reconstruction techniques have been applied successfully for abdominal and thoracic imaging as fast motion-insensitive imaging sequences (14-16). In this work, we propose a rapid projection reconstruction approach that requires no additional hardware for device tracking. We demonstrate its feasibility on a standard MR imager using a phantom.

0 METHODS Initially, a high resolution static base-

line image is acquired and reconstructed. This iinage may be acquired with any stan- dard sequence. It serves as a baseline im- age upon which dynamic needle placement related changes will be super- imposed. Then, a low resolution second static baseline image is acquired using only a few projections with a radial fast spin echo (rFSE) sequence ( 15). During a dynamic process, for example needle

placement, additional images are acquired using a few projections with the same rFSE sequence. The difference between the images acquired during the dynamic changes and the static baseline image should only contain information about the needle advancement. When the image is reconstructed from a few projections. the difference between these images shows streaking artifacts caused by angular un- dersampling. Using these artifacts as markers, the position of the needle tip can be identified clearly as a strong signal in- tensity spot as shown in Figure 1. Then, this image is superimposed on the high resolution baseline image. Because acqui- sition of only a few projections takes a much shorter time compared with that re- quired for the full resolution image, con- siderably reduced imaging times, applica- ble to rapid imaging of dynamic processes, can be achieved.

This method was implemented on a commercially available MRJ system (1.5-T Signa, General Electric Medical Systems, Milwaukee, WI). For the high resolution image, a spin-echo image was acquired using the following parameters: TR = 150 msec; TE = 20 msec: field of view (FOV) = 180 mrn: slice thickness = 10 rnm: image matrix size = 256 X 256 pixels. For the acquisition of projections, we applied a n rFSE sequence with a n interleaved ap- proach as described by Rasche et a1 (15). The pulse sequence used for the study is illustrated in Figure 2. Imaging param- eters were as follows: TR = 150 msec: TE = 20 msec: number of echoes in one echo train = 4: total number of shots = 24: FOV = 180 mm: slice thickness = 10 mm; number of data points per projection = 256. Only a single slice was acquired. The projections were acquired for the following angular order. For odd shots, (~r/4) * n: for even shots, (71/4) * n + n/8: where n is the echo number. In this way, we exploited the great flexibility of MRI in choosing the ge- ometry for acquiring projections, in con- trast to the mechanical restrictions of modalities such as CT scanner. Using only eight projections, an image update was ob- tained every 300 msec with only two shots of the rFSE sequence.

To examine our device tracking method, phantom studies were performed. We fab- ricated a cage for placement of both a phantom and a long bar attached to a n 18- gauge MR-compatible biopsy needle (E-Z-

!62

Figure 1. The difference between the im- age reconstructed by eight projections during a dynamic process and the static baseline image reconstructed using the same eight projections. Although streak- ing artifacts are observed, the tip of the needle is clearly demonstrated as a strong signal intensity spot.

Em. Westbury, NY). By manuaIIy advanc- ing this bar, the needle was inserted through a watermelon phantom during application of the pulse sequence. We also correlated the actual tip location and the tip depicted by the present technique or a standard spin-echo sequence for the high resolution image.

9ox RF

and signal

RESULTS The conventional spin-echo image ac-

quired as the high resolution static base- line image is shown in Figure 3. Figure 4 shows 3 of 12 frames obtained in our im- aging technique. These images clearly demonstrate that our method is feasible to dynamically monitor and locate the tip of the needle during its advancement in this phantom study. Because only a few pro- jections are acquired for reconstruction, streaking artifacts emanating from the needle tip are observed. These artifacts may provide markers for the needle tip lo- cation.

180y 180y

In the correlation study between the ac- tual tip location and the location mea- sured with the rapid projection recon- struction technique or the standard high resolution spin-echo sequence, we found that the positions depicted by these two methods were almost identical. The im- aged tip location obtained with these two sequences, however, was consistently dis- placed along the axis of the needle, slightly behind the actual tip location, by approx- imately 5 mm.

0 DISCUSSION In theory, the radial streaks caused by

angular undersampling do not have strong signal intensity (17) and, therefore, do not cause serious image degradation to ob- scure the potential target lesion, where at- tention should be paid. In a n interven- tional procedure such as biopsy, the image plane must contain both a n instrument and a target lesion. To accomplish this task, the needle tip should be monitored carefully to avoid maneuvering it out of the plane. Nevertheless, in practice, multi- plane imaging will be necessary to com- pletc. this procedure. These images should cover the entire volume within which the procedure is to be performed. The ex- pected course or needle path can be used to limit the imaging volume to be covered.

The method described here merely spots the tip of the needle, which is then super- imposed on the previously acquired base- line image. This procedure works as long as the baseline image has remained un- changed. Therefore, after any gross mo- tion. the baseline image must be re- freshed. The present technique has achieved the imaging time of 3.6 seconds for 12 frame updates, indicating its poten- tial for abdominal applications in which breath-holding is required.

The spatial accuracy of the tip depicted by the present technique compared to its actual location should be considered care- fully. First, we compared these two loca- tions in the correlation study. We found that the tip depicted by the present tech- nique was consistently displaced along the

180y 180y

..... . . ...._. . . . . . . . . _ . . . . . _ .

G Y ... ..:

... ... Figure 2. Pulse sequence used for the acquisition of projections. RF = radiofrequency; Gx and Gy are readout gradients.

axis of the needle by a small distance (5 mm]. This type of displacement niay be well compensated by a physician, or a sim- ple display technique for compensation is available. The other issue is the difference between the actual tip location and the tip depicted by the present technique during the needle advancement. This accuracy will be determined when our method is in- tegrated with real-time reconstruction, im- age display, and a n independent posi- tion-tracking system (1,2). However, the temporal resolution of 300 msec achieved with the present technique can largely re- move blurring during the dynamic pro- cess.

The active MR-tracking approach, in which a small receiver coil is incorporated into the device, uses a nonselective radio- frequency pulse and a gradient-recalled echo signal to provide three-dimensional coordinates that can be mapped onto the arbitrary MR images of different planes (3-6). This method requires special hard- ware and the installation of a miniature coil a t the tip of catheters and flexible en- doscopes. Therefore, it is a n expensive so- lution to the problem of tip tracking. In the active tracking method, a physician can choose whether to track the device on the previously acquired images of single or multiple planes or to have continuous se- rial image updates to monitor the progress of tip position. This active tracking method requires additional time to obtain localiz- ing information using the pulse sequence. In contrast, our method can rapidly track the location of the tip as image updates. These images are believed to be sufficient for monitoring the needle just before reaching the target lesion, as long as gross motion does not occur.

To date, there are some limitations with passive techniques (7.8), in which xisual- ization of the device is achieved by means of the susceptibility effects of paramag- netic materials (9,IO). The appearance of the device based on susceptibility effects is not sufficiently reliable; the size and shape of device-induced signal void de- pends on device orientation in the mag- netic field. In addition, the position of the device is updated only as fast as new im- ages are acquired and reconstructed, re- sulting in the inability to make real-time adjustment of the imaging plane. In MR-

Figure 3. acquired with a spin-echo sequence.

High resolution baseline image

Volume 8 Number 1 JMRl 263

b. a.

Figure 4. Needle tip tracking by the proposed method. Shown are three frames of a dynamic series of images during the needle advancement: (a), at .9 second: (b), at 1.8 seconds; and (c), a t 3.6 seconds. The needle tip is clearly visualized a s a white spot with streaking artifacts.

C.

guided endovascular interventions (78). keyhole technique (11) and its variants (12,131 have been used to improve a tem- poral resolution. The use of this technique derives from the fact that the paramagnetic ring attached to the device causes a signal loss that is not sharply defined. When the keyhole technique is applied to monitor the needle, several considerations must be taken to meet the requirements of accept- able needle tip location and needle width with minimal blurring artifacts (18). To our knowledge, a dedicated imaging technique for passive tracking of the needle in real time within interventional MR guidance has not been established. Our method has suc- cessfully achieved passive tracking of a standard MR-compatible thin needle. This is accomplished by a simple strategy to de- pict only the tip using a projection recon- struction scheme with a few projections, avoiding serial acquisition of updated im- ages with full spatial resolution.

In this work, we used an rFSE sequence to implement our method. Radial gradient- echo sequences are also applicable (161, with a concomitant increase in the rate of image update. At present, reconstruction in real time has not been implemented: however, it can be achieved because each reconstruction takes only eight filtered back projection steps. Although current implementation of our method does not di- rectly offer the three-dimensional position of the device, dedicated software for cal- culation of positional information using the selected slice and the projections could be developed. Two tip positions can define the needle trajectory. An arbitrary plane containing this line can be displayed after multiplanar reformation of the multislice and/or volumetric images. A new baseline

image can also be acquired along the pre- dicted trajectory. Following these proce- dures, the dynamic imaging protocol can be repeated in that scan plane. For track- ing the tip of flexible instruments, a more elaborate image acquisition and process- ing scheme should be designed. Because our method is implemented on a commer- cially available MFU system with standard hardware and does not require a special tracking device, combined with a n inte- grated scan and display technique, it may provide a practical approach to the track- ing problems.

References 1. Schenck JF. Jolesz FA, Roemer PB. et

al. Superconducting open-configuration MR imaging system for image-guided ther- apy. Radiology 1995; 195:805-814. Silvmman SG, Collick BD. Figueira MR. et al. Interactive MR-guided biopsy in an open-configuration MR imaging system. Ra- diology 1995; 197:175-181.

3. Dumoulin CL, Souza SP. Darrow RD. Real-time position monitoring of invasive devices using magnetic resonance. Magn Reson Med 1993: 29:411-415. Leung DA. Debatin JF. Wildermuth S. et al. lntravascular MR tracking catheter: preliminary experimentation evaluation. AJR 1995: 164:1265-1270. h u n g DA, Debatin JF. Wildermuth S. et al. Real-time biplanar needle tracking for interventional MR imaging procedures. Ra- diology 1995; 197:485-488.

6. Wildennuth S. Debatin JF. Leung DA, et al. MR imaging-guided intravascular pro- cedures: initial demonstration in a pig model. Radiology 1997: 202578-583. Bakker CJ, Hoogeveen RM. Weber J. van Vaals JJ. Viergever MA, Mali WP. Visualization of dedicated catheters using fast scanning techniques with potential for

2.

4.

5.

7.

MR-guided vascular interventions. Magn Reson Med 1996: 36:816-820. Bakker CJ. Hoogeveen RM, Hurtak WF. van Vaals JJ. Viergever MA. Mali WF'TM. MR- guided endovascular interventions: suscep- tibility-based catheter and near-]real-time imaging technique. Radiology 1997: 202: 273-276. Rubin DL. Ratner AV, Young SW. Magnetic susceptibility effects and their application in the development of new ferromagnetic cath- eters for magnetic resonance imaging. Invest Radio1 1990; 25:1325-1332. Kochli VD. McKinnon GC, Hofmann E. von Schulthess GK. Vascular interventions guided by ultrafast MR imaging: evaluation of different materials. Magn Reson Med 1994: 31 :309-314.

1. Van Vaals JJ, Brummer ME, Dixon WT. et al. "Keyhole" method for accelerating im- aging of contrast agent uptake. J Magn Re- son Imaging 1993; 3571-675.

12. Hu X. Parrish T. Reduction of field of view for dynamic imaging. Magn Reson Med 1994: 31 :691-694.

13. Fredrickson JO, Pelc N J . Temporal reso- lution improvement in dynamic imaging. Magn Reson Med 1996: 35:621-625.

14. Glover GH. Pauly J M . Projection recon- struction techniques for reduction of mo- tion effects in MRI. Magn Reson Med 1992: 28:275-289.

15. Rasche V, Holz D, Schepper W. Radial turbo spin echo imaging. Magn Reson Med 1994; 32629-638.

16. Rasche V, de Boer RW, Holtz D. Proksa R. Continuous radial data acquisition for dynamic MRI. Magn Reson Med 1995; 34: 754-761.

17. Bracewell RN, Riddle AC. Inversion of fan-beam scans in radio astronomy. Astro- phys J 1967; 150:427-434.

18. Duerk J L , Lewin JS, Wu DH. Application of keyhole imaging to interventional MRI: a simulation study to predict sequence re- quirements. J Magn Reson Imaging 1996: €2918-924.

8.

9.

0.

264 JMRl JanuarylFebruary 1998