MR Virtual Endoscopy of the Upper Urinary Tract - AJR

MR Virtual Endoscopy of the Upper Urinary Tract Emanuele Neri 1 Piero Boraschi 2 Davide Caramella 1 Luigi Battolla 1 Roberto Gigoni 2 Nicola Armillotta 1 Giovanni Braccini 2 Carlo Bartolozzi 1

OBJECTIVE. We investigated the feasibility of applying surface-rendered virtual endoscopy to the visualization of the upper urinary tract by processing unenhanced MR urography data sets. SUBJECTS AND METHODS. Twenty-six patients, having neoplastic lesions (n = 9), calculi (n = 8), pelviureteric junction stenoses (n = 4), postoperative fibrotic strictures (n = 3), and extrinsic compressions of the ureter (n = 2), underwent unenhanced MR urography. Virtual endoscopy of the upper urinary tract was obtained using a thresholding technique and surface-rendering MR urography data sets. RESULTS. Virtual endoscopy of the renal pelvis and calices was feasible in all cases on the side of the urinary obstruction. Virtual endoscopy of the ureter was obtained for a diameter of at least larger than 5 mm. The nondilated side could be partially explored in 11 cases (43%). The mean virtual endoscopy threshold required for the visualization of the urinary tract was 157.36–159.94. The mean time for virtual endoscopy was 13.8 min. Endoluminal masses were found in three (12%) of 26 cases on the renal pelvis (corresponding to neoplastic lesions), and occlusions, in 23 (88%) of 26 on the pelviureteric junction and ureter (neoplastic lesions and other abnormalities). CONCLUSION. Virtual endoscopy of MR urography data sets is feasible in patients with urinary tract dilatation. Virtual endoscopy displays the renal pelvis, calices, and ureter and, moreover, can show endoluminal changes caused by abnormalities.

C

Received March 17, 2000; accepted after revision May 31, 2000. 1 Department of Oncology, Transplants and Advanced Technologies in Medicine, Division of Diagnostic and Interventional Radiology, University of Pisa, Via Roma 67, 56100, Pisa, Italy. Address correspondence to E. Neri. 2 Second Department of Radiology, Pisa University Hospital, 56100, Pisa, Italy.

AJR 2000;175:1697–1702 0361–803X/00/1756–1697 © American Roentgen Ray Society

AJR:175, December 2000

ross-sectional urinary tract examinations using rapid three-dimensional (3D) volumetric techniques can be performed with helical CT, electron beam CT, and MR urography. The integration between volumetric acquisition and 3D surfaceand volume-rendering techniques permits the study of the urinary tract. Three-dimensional reconstructions of the bladder can be obtained easily with either external or endoluminal points of view of the organ anatomy [1–6]. MR urography provides a projectional road map of the entire urinary tract. Two methods have been proposed for MR urography, on the basis of unenhanced and gadolinium-enhanced studies. The unenhanced study is based on the acquisition of so-called water images using heavily T2-weighted turbo spin-echo or half-Fourier acquisition singleshot turbo spin-echo sequences and other variations of these techniques. These approaches do not require contrast medium administration and are feasible in patients with urinary tract obstruction [7–20]. The gadolin-

ium-enhanced study, performed using T1weighted gradient-echo sequences, provides morphologic and functional information about the urinary tract [21, 22]. The feasibility of a so-called virtual ureterorenoscopy with gadolinium-enhanced MR urography also has been proposed. This technique permits simulation of endoscopic images from the calices to the ureteral orifices in the bladder and identification of all filling defects that are diagnosed on MR urography [23]. Virtual endoscopy of MR imaging data sets has been successfully applied to the study of vessels, the biliary tract, the colon, and cerebral ventricles [24–31]. To our knowledge, no researchers of virtual endoscopy of unenhanced MR urography have described their experience in the study of the upper urinary tract. Therefore, we aimed to investigate, in a clinical setting, the feasibility of applying surface-rendered virtual endoscopy to the visualization of the upper urinary tract by processing unenhanced MR urography data sets.

1697

Neri et al. Subjects and Methods Patients Twenty-six patients (16 male, 10 female; mean age, 57.08 ± 19.86 years; range, 16–84 years) with signs of urinary obstruction on sonography or excretory urography were examined. The final diagnosis was made on the basis of surgical findings or imaging follow-up of at least 6 months. Institutional review board approval was obtained for the study, and all patients signed consent forms. Our series included neoplastic lesions (n = 9) located in the renal pelvis, calices, and ureter; calculi (n = 8) located in the ureter; pelviureteric junction stenoses (n = 4); postoperative fibrotic strictures (n = 3); and extrinsic compressions (n = 2) located in the lumbar tract of the ureter. Image Acquisition MR imaging was performed with a 1.5-T magnet (Signa; General Electric Medical Systems, Milwaukee, WI) using a body coil for both excitation and signal reception (the surface torso array coil is not available at our institution). To avoid peristaltic artifacts in all patients, 20 mg of scopolamine methylbromide was injected IV before image acquisition. MR urography was performed with a non– breath-hold fat-suppressed respiratory-triggered two-dimensional heavily T2-weighted fast spin-echo sequence in the coronal plane. Imaging parameters were as follows: TR range/TE, 7000–18,000/253; echo train length, 32; section thickness, 3 mm with no interslice gap; field of view, 35–45 cm; matrix size, 256 × 128 pixels; number of excitations, two to four; receive bandwidth, 16 kHz. Anteroposterior spatial presaturation was used for all images. Acquisition time ranged from 5 to 7 min. Virtual Endoscopic Technique To perform virtual endoscopy of the urinary tract, we transferred the source images from the MR imaging unit to a dedicated workstation (Advantage Windows 3.1; General Electric Medical Systems). Virtual endoscopic images were generated with Navigator 2.0 software (General Electric Medical Systems). Navigator reconstructed the MR images in three dimensions and created surface-rendered endoscopic images of the urinary tract. However, for generating virtual endoscopic images, Navigator required the manual segmentation of the urinary tract using a thresholding technique [32]. Virtual endoscopy was performed independently by two observers experienced in image processing and MR urography, who were asked to identify independently the optimal virtual endoscopy threshold; on the source MR images they traced a circular region of interest (maximum size, 1000 mm 2) including the urinary tract and the surrounding tissue, and the relative histogram was obtained. Among the image pixels belonging to the urinary tract, the virtual endoscopy threshold was defined as the cluster of pixels having the minimum signal intensity on the histogram (Fig. 1). Virtual endoscopic images were always displayed in a combined fashion with MR source im-

1698

ages; a dedicated function allowed the observers to select a point on the 3D perspective and to obtain simultaneously the corresponding coronal image crossing through this point [24, 25, 27, 28]. By the help of this function, virtual endoscopic findings could be correlated with the coronal images. Analysis of Data Virtual endoscopy thresholds selected by the observers were recorded for each case. The observers were informed about the presence and location of the abnormalities causing dilatation of the urinary tract and were asked to evaluate the corresponding endoluminal image. The morphologic changes of the urinary tract lumen were ranked by means of a four-point scale as normal lumen (rank 0), stenosis (rank 1), endoluminal mass (rank 2), and occlusion (rank 3). The time required for image processing, or virtual endoscopy time, was divided into a non–operator-dependent phase, including the transfer of the source images from the MR imaging unit to the workstation and the 3D reconstruction of the MR imaging data set (performed by the software itself), and an operator-dependent phase, including the virtual endoscopy threshold selection and the interpretation of virtual endoscopy perspectives. The non–operator-dependent phase was assumed to be equal for both observers because it included hardwaredependent procedures. The non–operator-dependent and operator-dependent phases were recorded. Signal-to-Noise and Contrast-to-Noise Ratios For the quantitative evaluation of mean signalto-noise and contrast-to-noise ratios, a circular region of interest consisting of a minimum size of 5 mm 2 was sampled at the levels of the renal pelvis,

the lumbar muscles, and noise outside of the abdomen. Measurements were taken by the observers for the dilated and nondilated sides. The signal-to-noise ratio was defined as the ratio of mean signal intensity in the renal pelvis to the standard deviation of noise, and contrast-to-noise ratio, as the difference in mean signal intensity between the renal pelvis and the lumbar muscles divided by the standard deviation of noise [33]. Statistical Analysis Data of all patients were processed with Excel 97 software (Microsoft, Redmond, WA). The correlation between the observers for the virtual endoscopy threshold and the operator-dependent phase was determined with the Spearman’s rank correlation coefficient test. A correlation coefficient ( r > 0.67) was considered indicative of good and statistically significant correlation ( p < 0.001) [34]. The agreement between the observers for the assessment of endoluminal perspectives was evaluated using the Cohen kappa statistic. The kappa statistic describes the quality of agreement; kappa values between 0.40 and 0.75 represent fair to good agreement and kappa values greater than 0.75, excellent agreement. The difference in mean signal-to-noise and contrast-to-noise ratios between the dilated and nondilated sides was evaluated with the Student’s t test. A p value of less than 0.05 was considered to indicate a statistically significant difference [33]. Maximum and minimum values, the geometric mean, and the standard deviation were calculated for the virtual endoscopy threshold, non–operatordependent phase, operator-dependent phase, and signal-to-noise and contrast-to-noise ratios.

Fig. 1.—57-year-old woman who underwent colectomy for colorectal cancer. MR urogram shows dilatation of left renal pelvis and ureter, caused by extrinsic compression of ureter by postoperative fibrosis. Histogram was obtained by tracing on MR urogram circular region of interest (size, 1000 mm 2) to include urinary tract (B) and surrounding tissue (A). Among image pixels belonging to urinary tract, virtual endoscopy threshold is identified as cluster of pixels with lower signal intensity.


MR Virtual Endoscopy of the Upper Urinary Tract Results Virtual Endoscopy: Feasibility

Virtual endoscopy of the renal pelvis and calices was feasible in all patients at the side of urinary obstruction (Fig. 2). Virtual endoscopy of the ureter, from the pelviureteric junction to the site of obstruction, was obtained with acceptable quality when its caliber was at least more than 5 mm. The contralateral side could be partially explored in 11 patients (43%), and navigation was feasible in the calices, renal pelvis, and proximal tract of the ureter. The mean virtual endoscopy threshold required for the visualization of the urinary tract was 157.36 signal intensity (minimum, 60; maximum, 260; SD, ±57.43) for one observer and 159.94 (minimum, 69; maximum, 250; SD, ±52.48) for the other. The correlation between the observers for virtual endoscopy threshold was excellent (r = 0.68; p < 0.001). The mean non–operator-dependent phase time was 4.8 min (minimum, 3 min; maximum, 7 min; SD, ± 1.23 min). The mean operator-dependent phase time was 9 min (minimum, 4 min; maximum, 16 min; SD, ± 4.34 min) for one observer and 8 min (minimum, 4 min; maximum, 15 min; SD, ±3.24 min) for the other. The correlation between the observers for operator-dependent phase was excellent ( r = 0.67; p < 0.001). Virtual Endoscopy: Findings

Endoluminal masses were found in three (12%) of 26 patients in the renal pelvis and cor-

responded to neoplastic lesions (Fig. 3). Occlusions were found in 23 (88%) of 26 patients in the pelviureteric junction and ureter and corresponded to neoplastic lesions (n = 6), calculi (n = 8), postoperative strictures (n = 3), ureteroplelvic junction stenosis (n = 4), and extrinsic ureteric compression (n = 2). No stenosis was found. The agreement between the observers for the evaluation of endoluminal patterns was excellent (κ = 0.89; 95% confidence interval). For endoluminal masses virtual endoscopy allowed the study of renal calices, the pelvis, and the proximal portion of the ureter. In three (12%) of 26 patients, the observers reported the occurrence of pierced surface artifacts at the level of the ureter [33]. These appeared as scattered holes of the internal surface of the ureter and were related to the inherent difficulties of the software in separating hyperintense voxels at the periphery of the ureter lumen from the surrounding tissue. An appropriate selection of virtual endoscopy threshold allowed the elimination of these artifacts. Signal-to-Noise and Contrast-to-Noise Ratios

The mean signal-to-noise ratio was 52.22 ± 12.4 and 41.66 ± 12.99 for the dilated and nondilated sides, respectively. The mean contrast-to-noise ratio was 50.11 ± 12.44 and 39.54 ± 12.55 for the dilated and nondilated sides, respectively. The difference in mean signal-to-noise ratio and contrast-tonoise ratio between dilated and nondilated sides was statistically significant ( p < 0.05).

A

B

Discussion

To date, the feasibility of virtual endoscopy of the upper urinary tract has been investigated in patients with and without abnormalities by processing MR urograms obtained with T1weighted 3D gradient-echo sequences after IV administration of gadopentetate dimeglumine [23]. Virtual endoscopy was created using a thresholding technique and was feasible in all patients of the study, even in those with no dilatation of the urinary tract. By contrast, in our study the dilatation of the urinary tract was a prerequisite for generating endoscopic images. In fact, the dilatation of the ureter or renal pelvis increased the availability of bright image pixels and, consequently, the voxels that virtual endoscopy could use to reconstruct the lumen in three dimensions and to create the virtual space for endoscopic navigation. We identified a lumen caliber of at least 5 mm for optimal display of the lumen because at lower diameters it appeared narrowed or occluded. In all patients, virtual endoscopy of the upper urinary tract was also performed on the nondilated side, but it failed to display the lumen with acceptable image quality in 15 (57%) of 26 patients. We considered virtual endoscopic images to be of acceptable quality when they showed a clear distinction between the virtual space for endoscopic navigation and the surface. This limitation was likely inherent to the thresholding method we were using; in fact, this method required a marked MR signal intensity difference between the urinary tract and the

C

Fig. 2.—43-year-old man with recurrent infections of urinary tract. A, MR urogram source image shows dilatation of renal pelvis, suggesting obstruction of pelviureteric junction. B, Maximum-intensity-projection image reveals site of urinary obstruction (arrow ). Note three major calices that fuse to form renal pelvis. C, Virtual endoscopic image, created by simulating position of endoscope at level of renal pelvis, shows internal surface of superior (short solid arrows ), medium (open arrows ), and inferior (long solid arrows ) major calices.


1699

Neri et al.

A

B

D

C

E

H

F

I

G

J

Fig. 3.—78-year-old man with hematuria. A, MR urogram shows filling defect within renal pelvis (arrows ). B, Maximum-intensity-projection image shows extension of lesion to superior major calyx (arrows ). C, Virtual endoscopic image, created from pelviureteric junction to show renal pelvis and calices, reveals well-defined irregularly shaped endoluminal mass. D, MR urogram shows entire dilatation of ureter (maximum diameter, 15 mm) and site of obstruction (arrow ). E–J, Virtual endoscopic images, obtained in ureter, show pelviureteric junction (E), abdominal (F and G) and pelvic (H and I) tracts of ureter, and site of occlusion (arrow , J).

1700


MR Virtual Endoscopy of the Upper Urinary Tract surrounding tissue to obtain the precise reconstruction of the lumen surface, and such condition was present only in patients with urinary obstruction. The statistically significant difference between the dilated and nondilated sides, for signal-to-noise and contrast-to-noise ratios, supports this consideration. In this study, we also tried to analyze whether the virtual endoscopy threshold could affect the interobserver variability for the evaluation of endoluminal images. The excellent correlation between the observers suggests that virtual endoscopy threshold selection did not likely influence the assessment of virtual endoscopy patterns. However, this excellent correlation also can be explained by the observers’ experience in image processing and virtual endoscopy and the use of common criteria (histogram evaluation) for threshold selection. The criteria for the assessment of endoluminal patterns, previously described in the study of the biliary tract [27], were helpful even in the present study because of the similarities between MR cholangiography and MR urography data sets. The excellent correlation between the observers for the attribution of such patterns supports this evaluation method. Most abnormal patterns were described by the observers as occlusion of the lumen, and all abnormalities were causing an obstruction at the level of the ureter or pelviureteric junction. No case of stenosis was identified in our study, and presumably this is related to the ureter caliber. In fact, because virtual endoscopy was feasible only when the ureter was dilated, we believe that the presence of stenosis cannot be easily appreciated in the narrowed tract because the space for navigation is dramatically reduced. Furthermore, dilatation of the ureter is not necessarily associated with stenosis of the lumen but has an obvious association with its obstruction. In cases of neoplastic lesions presenting as endoluminal masses at the level of the renal pelvis, the distal portion of the ureter could not be explored. Virtual endoscopy depicted the endoluminal appearance of the lesions and the morphology, position with respect to calices, and extension of each lesion. However, we believe that the unique advantage of virtual endoscopy with respect to MR urographic source images was the morphologic assessment; position and extension of the lesions could be evaluated with coronal images as well. An important issue related to virtual endoscopy is the time required for image processing. In our experience, the study of


the urinary tract was not time-consuming. The total processing time (non–operator-dependent phase plus the operator-dependent phase) was approximately 12–13 min. If we consider a time range of 30–45 min for the entire examination, including patient preparation and acquisition of various imaging sequences, then unlike other applications (i.e., virtual colonoscopy), virtual endoscopy of the urinary tract did not significantly bias the total examination time. Pierced surface artifacts in virtual endoscopy have been described in CT data sets [34]. In our experience, these artifacts did not influence the study of the urinary tract, but we recommend to future investigators the proper use of the virtual endoscopy threshold to avoid this occurrence. Virtual endoscopy is therefore limited by the degree of dilatation of the ureter and by the occurrence of artifacts; these considerations do not reduce the value of virtual endoscopy in exploring the urinary tract. We believe this technique should be used as a complement to native images in difficult cases in which the degree of a stenosis or the extension of a lesion should be precisely determined in three dimensions. Moreover, we believe this tool will be an important piece of the future imaging-guided surgical tools that are under evaluation and development in different research projects. In summary, virtual endoscopy of MR urography data sets is feasible in patients with urinary dilatation. It allows observers to visualize the renal pelvis, calices, and ureter and to see the effects of abnormalities on the urinary tract lumen (endoluminal masses and occlusions). Virtual endoscopy of the urinary tract can also be easily performed and does not bias the global examination time. However, the aim of the present study was only to show the feasibility and issues related to this technique; we did not investigate whether virtual endoscopy can increase the diagnostic accuracy of MR urography. In our opinion, this issue will require further prospective study.

References 1. Merkle EM, Fleiter T, Wunderlich A, Rilinger N, Gorich J, Sokiranski R. Virtual cystoscopy based on spiral CT data [in German]. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 1996;165:582–585 2. Hussain S, Loeffler JA, Babayan RK, Fenlon HM. Thin-section helical computed tomography of the bladder: initial clinical experience with virtual reality imaging. Urology 1997;50:685–688

3. Fenlon HM, Bell TV, Ahari HK, Hussain S. Virtual cystoscopy: early clinical experience. Radiology 1997;205:272–275 4. Merkle EM, Wunderlich A, Aschoff AJ, et al. Virtual cystoscopy based on helical CT scan data sets: perspectives and limitations. Br J Radiol 1998;71:262–267 5. Frank R, Stenzl A, Frede T, et al. Three-dimensional computed tomography of the reconstructed lower urinary tract: technique and findings. Eur Radiol 1998;8:657–663 6. Stenzl A, Frank R, Eder R, et al. 3-Dimensional computerized tomography and virtual reality endoscopy of the reconstructed lower urinary tract. J Urol 1998;159:741–746 7. Sigmund G, Stover B, Zimmerhackl LB, Frankenschmidt A, Nitzsche E, Leititis JU. RARE-MR urography: a rapid MR tomographic imaging procedure for the diagnosis of urinary tract malformations in childhood [in German]. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 1991;154:535–540 8. Roy C, Saussine C, Jahn C, et al. Evaluation of RARE-MR urography in the assessment of ureterohydronephrosis. J Comput Assist Tomogr 1994;18:601–608 9. Rothpearl A, Frager D, Subramanian A, et al. MR urography: technique and application. Radiology 1995;194:125–130 10. Roy C, Saussine C, Jahn C, et al. Fast imaging MR assessment of ureterohydronephrosis during pregnancy. Magn Reson Imaging 1995;13: 767–772 11. Siewert C, Venz S, Friedrichs R, et al. MR urography with the T2-weighted turbo-spin-echo sequence [in German]. Aktuelle Radiol 1995;5:319–322 12. Aerts P, Van Hoe L, Bosmans H, Oyen R, Marchal G, Baert AL. Breath-hold MR urography using the HASTE technique. AJR 1996;166:543–545 13. Hussain S, O’Malley M, Jara H, Sadeghi-Nejad H, Yucel EK. MR urography. Magn Reson Imaging Clin N Am 1997;5:95–106 14. O’Malley ME, Soto JA, Yucel EK, Hussain S. MR urography: evaluation of a three-dimensional fast spin-echo technique in patients with hydronephrosis. AJR 1997;168:387–392 15. Li W, Chavez D, Edelman RR, Prasad PV. Magnetic resonance urography by breath-hold contrast-enhanced three-dimensional FISP. J Magn Reson Imaging 1997;7:309–311 16. Roy C, Saussine C, Guth S, et al. MR urography in the evaluation of urinary tract obstruction. Abdom Imaging 1998;23:27–34 17. Balci NC, Mueller-Lisse UG, Holzknecht N, et al. Breathhold MR urography: comparison between HASTE and RARE in healthy volunteers. Eur Radiol 1998;8:925–932 18. Klein LT, Frager D, Subramanium A, Lowe FC. Use of magnetic resonance urography. Urology 1998;52:602–608 19. Catalano C, Pavone P, Laghi A, et al. MR pyelography and conventional MR imaging in urinary tract obstruction. Acta Radiol 1999;40:198–202 20. Louca G, Liberopoulos K, Fidas A, Nikolakopoulou Z, Lykourinas M, Strigaris K. MR urography in the diagnosis of urinary tract obstruction. Eur Urol 1999;35:102–108 21. Nolte-Ernsting C, Adam G, Bucker A, Berges S,

1701

Neri et al. Bjornerud A, Gunther RW. Contrast-enhanced magnetic resonance urography: first experimental results with a polymeric gadolinium bloodpool agent. Invest Radiol 1997;32:418–423 22. Nolte-Ernsting CC, Bucker A, Adam GB, et al. Gadolinium-enhanced excretory MR urography after lowdose diuretic injection: comparison with conventional excretory urography. Radiology 1998;209:147–157 23. Nolte-Ernsting CC, Krombach G, Staatz G, Kilbinger M, Adam GB, Gunther RW. Virtual endoscopy of the upper urinary tract based on contrastenhanced MR urography data sets [in German]. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 1999;170:550–556 24. Davis CP, Ladd ME, Romanowski BJ, Wildermuth S, Knoplioch JF, Debatin JF. Human aorta: preliminary results with virtual endoscopy based

25.

26.

27.

28.

29.

on three-dimensional MR imaging data sets. Radiology 1996;199:37–40 Luboldt W, Debatin JF. Virtual endoscopic colonography based on 3D MRI. Abdom Imaging 1998;23:568–572 Shigematsu Y, Korogi Y, Hirai T, et al. Virtual MRI endoscopy of the intracranial cerebrospinal fluid spaces. Neuroradiology 1998;40:644–650 Neri E, Boraschi P, Braccini G, Caramella D, Perri G, Bartolozzi C. MR virtual endoscopy of the pancreaticobiliary tract. Magn Reson Imaging 1999;17:59–67 Neri E, Caramella D, Boraschi P, et al. Magnetic resonance virtual endoscopy of the common bile duct stones. Surg Endosc 1999;13:632–633 Dubno B, Debatin JF, Luboldt W, Schmidt M, Hany TF, Bauerfeind P. Virtual MR cholangiogra-

phy. AJR 1998;171:1547–1550 30. Schmid MR, Hany TF, Knesplova L, Schlumpf R, Debatin JF. 3D MR gastrography: exoscopic and endoscopic analysis of the stomach. Eur Radiol 1999;9:73–77 31. Wildermuth S, Debatin JF. Virtual endoscopy in abdominal MR imaging. Magn Reson Imaging Clin N Am 1999;7:349–364 32. Neri E, Caramella D, Falaschi F, et al. Virtual CT intravascular endoscopy of the aorta: pierced surface and floating shape thresholding artifacts. Radiology 1999;212:276–279 33. Wolff SD, Balaban RS. Assessing contrast on MR images. Radiology 1997;202:25–29 34. Siegel S, Castellan NJ. Non parametric statistics for the behavioral science, 2nd ed. New York: McGraw-Hill, 1988

Notice to Authors Authors of accepted manuscripts can now receive galley proofs via e-mail. The author’s system must be able to read BinHex files and have version 3.0 of Adobe Acrobat Reader. The ARRS Web site (www.arrs.org) has a link to the Adobe Web site, where authors can download free copies of the most recent version of the reader.

1702
