Real-time volume rendering of MRCP: clinical applications

E L S E V I E R Magnetic Resonance Materials in Physics, Biology and Medicine 10 (2000) 35-42

MAGMA Magnetic Resonance Materials in Physics, Bh)lo~' and Medicine

www.elsevier.com/locate/magma

Real-time volume rendering of MRCP: clinical applications #

Emanuele Neri a,,, Piero Boraschi b Davide Caramella a, Giovanni Braccini b, Roberto Gigoni b, Mirco Cosottini ", Simone Lodovigi % Carlo Bartolozzi ~

~' Dicision of Diagnostic and hlterventional Radiology, Department of Oncology, University of Pisa, Via Roma 67, 1-56100, Pisa, Italy b 2nd Department of Radiology, Pisa Unicersity Hospital, Pisa, Italy

Received 14 April 1999; received in revised form 4 August 1999; accepted 24 September 1999

Abstract

MR-cholangiopancreatography (Signa Contour 0.5T; GE/Medical Systems, Milwaukee, WI) data sets of 156 patients, obtained with a 2D T2-weighted FSE sequence, in the coronal plane, were volume rendered (Advantage Windows 3.1: GEMS) independently by two radiologists, that were asked to define the range of signal intensities in which the signal of tile pancreaticobiliary system was included and to rank the quality of native images and volume renderings. Patients had biliary stones (n=47), inflammatory ampullary stenoses (n= 18), pancreatic tumors (n= 12), surgical bilio-enteric anastomoses (, = 19), ampullary carcinomas (n = 2), pancreatic duct stone (n = I), cholangiocarcinoma (n = 3) and normal pancreaticobiliary tree (n = 54). Good quality volume renderings of the bile ducts were obtained for at least a maximum diameter of 1.5 ram. The quality rank agreement between volume rendering and native images was excellent (k = 0.94). The correlation between the observers for the setting the signal intensity range was excellent and statistically significant (P < 0.001). The correlation between the observers for the time of volume rendering was not statistically significant. Biliary stones could be displayed in 32/47 (68%) cases. The pancreatic duct stones was displayed as well. Inflammatory ampullary stenoses were detected in all cases (100%). In case of pancreatic tumors, cholangiocarcinomas and ampullary carcinomas volume rendering allowed to identify the site of stenosis. In surgical bilio-enteric anastomoses volume rendering was helpful to display the residual biliary tract, the site of anastomosis and the enteric tract. Volume rendering could be a reliable and efficient tool for the study of the anatomy and pathological changes of the pancreaticobiliary tract. �9 2000 Elsevier Science B.V. All rights reserved.

Keywords: Bile ducts; Stenosis or obstruction; Magnetic resonance (MR): Image processing: Three-dimensional

1. Introduction

Volume rendering (VR) is the latest development in three-dimensional (3D) visualization o f CT and magnetic resonance (MR) data sets. This method allows to display simultaneously different anatomical structures imaged within a single volume. VR provides an image created by simulated rays o f light, arising f rom a virtual source (i.e. the eye o f the user looking through the computer display), traversing the imaging volume that has been at tenuated by its contents, and assigns certain characteristic, such as opacity, color, light and shininess

~" Note: keywords have been selected from the Index to imaging literature (January 1999), of the RSNA (Supplement to Radiology).

* Corresponding author. Tel.: + 39-50-992-509: fax: + 39-50-551- 461.

E-mail a~Mress: [email protected] (E. Neri)

tO specific voxel values [1-13]. By contrast , surt)lce rendering methods are based on the setting of threshold values that allow to select a range of voxel and to display only the surface o f the corresponding volume [14-18].

The use o f VR with segmented echo-planar imaging of the pancreaticobil iary tract has been recently reported by investigators that obtained a clear depiction o f biliary, hepatic, and pancreatic ducts (with lumen diameter o f at least 2 ram), with good signal-to-noise ratio [19]. Other methods o f data collection for MR- cholangiopancrea tography (MRCP), that can be also volume rendered, have been reported [20-25].

At our Institution, a commercial sof tware/hardware system (Advantage Windows 3.1, GE/Medica l Systems, Milwaukee, WI) is used routinely for processing M R data sets. We have applied VR tools to the evaluation of the pancreaticobil iary tract, to assess the visualiza-

1352-8661 00/$ - see front matter 'r 2000 Elsevier Science B.V. All rights reserved. Pll: S 1352-8661(99)0007 I-X

36 E. Neri et al. / Magnetic Resonance Materials in Physics, Biology and Medicine 10 (2000) 35-42

tion advantages of VR in normal and pathologic patients.

2. Patients and methods

2. I. Patients population

We reviewed retrospectively with VR the MRCP (Signa Contour 0.5 T;GE/Medical Systems, Milwau- kee, WI) data sets of 156 patients (72 males and 84 females, age range 8-95) that were obtained with use of a body flexible coil for both excitation and signal reception.

Our series included biliary stones (n = 47), inflammatory ampullary stdnoses (n = 18), pancreatic tumors (n= 12), surgical bilio-enteric anastomoses (n= 19), ampullary carcinomas (n = 2), pancreatic duct stone (n = 1), cholangiocarcinomas (n = 3) and normal pancreaticobiliary tree (n = 54). In all cases MRCP findings were confirmed by ERCP, PTC, surgery, and/or US follow-up for at least 6 months.

2.2, hnages acquis'ition

MRCP was performed with a non-breath-hold, respiratory-triggered (11 slices were collected per trigger), fat-suppressed, two-dimensional (2D), heavily T2- weighted fast spin-echo sequence in the axial and coronal planes. Imaging parameters were: repetition time 7000-18 000 ms; echo time 240 ms; echo train length 16; section thickness 3 mm with no interslice gap; field of view 36-44 cm, matrix size 256 x 192 pixels, four excitations. The total volume covered by the acquisition ranged between 66 and 96 mm with, 22 and 32 images obtained, respectively. Anterior-posterior spa-

Table I

Observer 1 Observer 2 Rs b P

0.657 NS ~' TVR (min) Mean/S.D. 8.1 / _+ 3.06 33.2/8.9 Min~ max/ 3/15/12 20/45/25

range

LT (SII) ~ Mean, S.D. 86.4; + 34.5 88.7/+ 33.5 Min/max~ 23/136/113 26/140/114

range

UT (SII) Mean/S.D. 488.9/+ 56.3 495.4/53.8 Min/max/ 402/580/178 412/575/163

rangc

0.93 P<0.001

0.974 P<0.001

~' NS, correlation not statistically significant. b R s ' Spearman correlation coefficient.

SII, signal intensity interval.

(a)

(b)

Fig. 1. Hepatic duct stone with diameter > 10 mm. The coronal view (a) shows a large area (arrows) of signal void within the hepatic duct, corresponding to a stone with maximum diameter 22 mm. The volume render (posterior view) demonstrates the area of signal void (arrows) within the hepatic duct.

tial presaturation was used for all images in an attempt to reduce respiratory artifacts. Acquisition time ranged between 5 and 7 min.

2.3. bnages processing

VR software was running on UltraSparc 1 workstation (Sun Microsystems, Mountain View, CA) with 256 Mb of random access memory. VR was performed by using coronal images as source data set. Coronal im

E. Neri et al. / Magnetic Resonance MateriaLs h7 Physics, Biology and Medichw 10 (2000) 35-42 37

ages were assumed to be more suitable than axial views since they provided a better depiction of biliary anatomy, allowing an easier analysis of VR, and good

quality 3D reconstructions [17]. MRCP coronal images were stacked by the software to generate a volumetric data set. VR required the setting of the signal intensity interval (SII), represented by the range of MR signal intensities in which the signal of the pancreaticobiliary system is included. The range was defined manually through the selection of the SII by prescription of lower (LT) and upper (UT) signal intensity thresholds. These values were used to assign the maximum transparency to the pixels with intensity values outside the SII, and the maximum opacity to the pixels within the SII by increasing linearly as the pixel value increases.

2.4. Quantitative evaluation

The MR data sets of each patient were processed and evaluated independently on the workstation by two radiologists, each experienced in image processing techniques (E.N. and P.B.). The observers were asked to identify the optimal LT and UT values of SII for VR, and to rank the quality of VR with respect to coronal images.

In principle VR results from the computation of data (voxels) pertaining to native images, as the coronal views used in our experience. Therefore, these images were taken as standard of reference, and in setting the LT and UT the observers aimed to reproduce on VR as much details visible on coronal views. In particular the

(a)

(b) (c) Fig. 2. CBD stone with diameter < 10 mm. The coronal acquisition (a) allows to detect a single stone (arrows), with maximum diameter 8 mm. in the proximal tract of the CBD (diameter 15 mm). The volume renders (anterior view), obtained with low (b) and high (c) degrees of tr~t~sparency of the bile ducts do not show the signal void corresponding to the stone.

38 E. ?v2,ri et al. / Magnetic Resonance Materials in Physics, Biology and Metfichte 10 (2000) 35-42

(a) (b)

(c) (d] Fig. 3, Inflammatory ampullary stenosis. The coronal view (a) demonstrates the ampullary stenosis (arrow) and the dilated pancreatic and common bile ducts. The MIP (b) and volume rendering (c), displayed, respectively, in the anterior and posterior view, show the dilatation of the biliary tree. The stricture length (arrows) can be appreciated easily in the lateral volume rendered perspective (d). CBD, common bile duct: HD, hepatic duct; W, Wirsung; S, stomach; D, duodenum.

setting of the LT was done by looking at the smallest signal possible over the background noise of a small bile duct.

The visualization of the pancreaticobiliary tract with VR was independently scored by the two readers on the basis of a subjective three-point scale. Visualizations were scored as: 1, no visualization of one or more pancreaticobiliary tract components displayed by coronal views: 2, incomplete representation of one or more pancreaticobiliary tract components displayed by c o r o -

nal views; 3, complete representation of all pancreaticobiliary tract components displayed by coronal views. The readers reached a consensus for scoring the visualization of the pancreaticobiliary tract on coronal images.

For the quantitative evaluation of mean signal-to- noise and contrast-to-noise ratios regions of interest, minimum size 10 pixels, were measured from the common bile duct (CBD) and from background. The signal-to-noise ratio was calculated as the ratio of the

E. Neri et al. / Magnetic Resonance MateriaLs" in PIo,sics, Bioh~gy and Medichw 10 (2000) 35-42 39

mean signal intensity within the CBD to the standard deviation of the background signal intensity. The contrast-to-noise ratio was defined as the difference in mean signal intensity between the CBD and the background divided by the standard deviation of noise [26]. The time taken for definition of the VR (TVR) was recorded.

2.5. Statistical analysis

Data series of all patients were processed with the STATVIEW software package (Abacus Concepts, Berke- ley, CA). Geometric mean, max and min, standard

deviation, range between max and min values, were calculated for LT and UT of SII, TVR, signal-to-noise and contrast-to-noise ratios.

The agreement between the observers in the assess- ment of VR perspectives was evaluated by using the Cohen kappa statistic [27]. The kappa statistic describes the quality of agreement between observers, k-values between 0.40 and 0.75 represent fair to good agreement and k-values above 0.75 excellent agreement.

The correlation between observers for attribution of LT, UT and TVR was determined with the Spearman rank test [28,29]. A correlation coefficient (R 0 > 0.765 was considered indicative of good and statistically significant correlation (P < 0.001). Results are presented as means +_ standard deviation.

(a)

(b) Fig. 4. Hepatic duct and CBD obstruction (arrows) caused by a giant chotangiocarcinoma. MIP (a) and volume rendering (b) allow to depict the site and length of stenosis (arrow) and the entire dilatation of the biliary tract. RHD, right hepatic duct: LHD, left hepatic duct: CD, cystic duct; G, gallbladder.

3. Results

VR demonstrated the biliary anatomy in all cases, even if the bile ducts, with a maximum diameter less then 1.5 mm, observed in the majority of normal sub- jects, were incompletely represented on 3D views (scored as 2) and showed very low signal intensity on native images.

The quality rank agreement between VR and coronal views was excellent (k = 0.94). The correlation between the observers t'or the setting of LT and UT of SII was excellent and statistically significant (P<0.001) . The correlation between the observers for TVR was not statistically significant. Results are reported in Table 1.

Forty-five out of 47 (96%) stones were detected at coronal views (la and 2a). Maximum diameter was less then 5 mm in 7/47 (15%) cases, and 2/7 were not detected by coronal views. Biliary stones could be displayed with VR in 32/47 (68%) cases as endoluminal signal voids by increasing the degree of transparency of the CBD surface (Fig. Ib). VR displayed all stones with maximum diameter > 10 mm and missed all stones with maximum diameter < 10 mm. In fact, in the remaining 15 cases in which stones had maximum diameter less then 10 mm, VR did not represent the signal void (clearly visible on coronal images) despite increased transparency settings (Fig. 2b,c). The pancreatic duct stone was displayed as well, because its diameter was more than 10 ram.

Inflammatory ampullary stenoses were detected in all cases (100%), with clear depiction of the stenosis length (Fig. 3). In case of pancreatic tumors, cholangiocarcinomas and ampullary carcinomas VR allowed to identify the site of stenosis (Fig. 4). In surgical bilio-enteric anastomoses VR was helpful to display the residual biliary tract, the site of anastomosis, the enteric tract (water-filled) (Fig. 5).

The mean signal-to-noise ratio was 54.71 (S.D. + 17.106). The mean contrast to noise ratio was 52.32 (S.D. _ 16.784).

40 E. Neri et al. / Magnetic Resonance Mater&ls hi Phy.~'ics, Biology and Me~fichTe 10 (2000) 35-42

(a) (b)

(c) (d) Fig. 5. Stenosis of surgical bilio-enteric anastomosis. The stenosis (arrow) is demonstrated on the coronal view (a). The maximum intensity projection (MIP) shows the stenosis as well. and the dilatation of the biliary tree (b). Volume renderings show the site of stenosis (arrow) and the biliary tree dilation (c,d). HD, hepatic duct: J, jejuno.

4. Discussion

Preliminary experiences on 3D reconstructions of the pancreaticobiliary tract have been carried out by some investigators with surface rendering of CT cholangiopancreatography data sets [30-33]. Other investigators, that studied M R C P data sets, reported the use of maximum intensity projection algorithm as an integrative tool of this technique [20-25].

Our study has shown the capability of VR to demonstrate the biliary tract anatomy. Other authors reported a threshold value of 2 mm for the maximum diameter

of the bile ducts needed to generate good quality volume renderings [19]. We had a slight lower threshold; and this difference could be related to the use of different LT and UT of SII. In fact, by decreasing the LT the anatomical structures with a poor signal can be volume rendered and better represented than native images. This is due to the additive effect and to the better signal to noise ratio of VR. However, the potential risk of this operation is to include the voxels with low signal that belong to the background, creating an artefactual superimposition of 3D elements.

The excellent agreement between VR and coronal images showed that this method does not hide informa-

E. Neri et al. / Magnetk' Resonance MaterktL*" in Physics, Biology and Medicine 10 (2000) 35-42 41

tion contained within the native images during 3D rendering. Such observation has a relevant impact on the clinical acceptance of VR. In fact, one of the major issues related to 3D rendering methods is represented

by the segmentation of the data sets. In case of surface rendering techniques segmentation can determine the lack of some data in the passage from native images to 3D models [17,18]. This issue is not reported for VR methods, that use more data for the computation than other 3D visualizatioil techniques, and allows to control opacity level for showing features that may be hidden under other setting by overprojection.

In our experience, the setting of LT and UT of SII was done manually by the observers that showed an excellent agreement (R < 0.001). The software provided a real-time result of rendering, and volume renders were displayed coupled with the corresponding coronal data set on the same monitor. In this way each observer could change the settings of SII to reach a final optimal result. The excellent agreement between the observers proves that in our study VR was not influenced by the inter-observer variability, although the use of coronal images as standard of reference was helpA,1 to correct the SII settings.

The difference of the TVR showed by the observers can be explained by the difference in image processing experience they had. A further element of difference between the observers was represented by the time of practice with the software interface.

The study of patients with calculi has shown that coronal views were strongly superior than VR in detect- ing stones. This observation suggests the importance of using base images or multiplanar reformations to locate intraluminal defects that may be masked during VR for specific transparency/opacity settings. However, VR has the advantage to provide a roadmap (projection) of the entire biliary tree; therefore, when stones were visible, VR could be coupled to coronal views for demonstrating their position within the biliary tract. We did not investigate whether the ratio between di- ameters of the CBD and the stones could influence their visibility on VR, even if we presume that an increased ratio will reduce the visibility of the stone for the overprojection effects typical of VR.

Typical clinical situations that require a clear roadmap of the biliary tree are represented by bilio-enteric anastomoses. In our experience VR was superior to coronal views to demonstrate biliary anatomy after surgical treatment.

The visualization of stenoses is another important area in which VR can play a significant role. We could detect stenoses of bilio-enteric anastomoses and of the ampulla with coronal view, but their length could be more efficiently assessed using VR.

5. Conclusions

Different authors report that VR is a reliable tool for 3D imaging. They highlight that VR hold into the data set all the information contained in the native images even after the 3D reconstruction. Our experience has shown similar results in case of the pancreaticobiliary tract even if we observed that overlapping features of VR represent a significant issue for the interpretation of reconstructed images. In patients with dilated ducts the review of 3D images can be cumbersome, requiring to analize the native images with multiplanar reformation or using a smaller volume for VR to locate stenoses and intraluminal defects; the rotation of the reconstructed view is beneficial as well.

The capability of VR to reproduce efficiently the signal properties of the bile ducts represented in the native images, suggests that this method could be used as an integrative tool of MRCP, to enhance the visualization of the pancreaticobiliary tract anatomy.

Acknowledgements

The authors would like to thank the radiographers Simona Del Corona and Arianna Garzclla for their support to the MR images acquisition.

References

[1] Rubin GD. Beaulicu CF. Argiro V, et al. Perspective volumc rendering of CT and MR ilnages: applications for endoscopic imaging. Radiology 1996;199:321-30.

[2] Maeder PP, Meuli RA, de Tribolet N. Three-dimensional volume rendering for magnetic resonance angiography in the screening and preoperative workup of intracranial aneurysms. J Neurosurg 1996:85:1050- 5.

[3] Johnson PT. Heath DG, Bliss DF, Cabral B, Fishman EK. Three-dimensional CT: real-time interactive volume rendering. Am J Roentgenol 1996:167:581-3.

[4] Kuszyk BS. Heath DG, Bliss DF, Fishman EK. Skeletal 3-D CT: advantages of volume rendering over surface rendering. Skeletal Radiol 1996:25:207-14.

[5] Beltrame F. Faro M, Raposio E, Sobel I. Recent results in color compositing of three-parameter magnetic resonance scans as a preoperative aid to the management of upper limb sarcomas. MAGMA 1997:5:289-98.

[6] Hesser J, Manner R. Braus DF, Ende G, Henn FA. Real-time direct volume rendering in functional magnetic resonance imaging. MAGMA 1997:5:87 91.

[7] Englmeier KH, Haubner M, Losch A. et al. Hybrid rendering of multidimensional image data. Methods lnf Med t997:36:1-10.

[8] Marcus C, Ladam-Marcus V, Bertini C, Jourdain L, Bonnet- Gausserand F. Menanteau B. Volume rendering technique in 3D vascular imaging. J Radiol 1997:78:481-4.

[9] Jani AB, Pelizzari CA, Chen GT, Grzeszczuk RP. Accuracy of object depiction and opacity transl'er function optimization in CT volume-rendered images. J Comput Assist Tomogr 1998:22:459-- 70.

42 E. Neri et al. / Magnetic Resonance Materials h7 Physics, Biology and Medicble lO (2000) 35-42

[10] Smith PA, Heath DG, Fishman EK. Virtual angioscopy using spiral CT and real-time interactive volume-rendering techniques. J Comput Assist Tomogr 1998;22:212-4.

[11] Johnson PT, Fishman EK, Duckwall JR, Calhoun PS, Heath DG. Interactive three-dimensional volume rendering of spiral CT data: current applications in the thorax. Radiographics 1998;18:165-87.

[12] Sato Y, Shiraga N, Nakajima S, Tamura S, Kikinis R. Local maximum intensity projection (LMIP): a new rendering method for vascular visualization. J Comput Assist Tomogr 1998;22:912-7.

[13] Smith PA, Klein AS, Heath DG, Chavin K, Fishman EK. Dual-phase spiral CT angiography with volumetric 3D rendering for preoperative liver transplant evaluation: preliminary observa- tions. J Comput Assist Tomogr 1998;22:868-74.

[14] Soyer P, Heath D, Bluemke DA, et al. Three-dimensional helical CT of intrahepatic venous structures: comparison of three rendering techniques. J Comput Assist Tomogr 1996;20:122-7.

[15] Rieker O, Mildenberger P, Rudig L, Schweden F, Thelen M. 3D CT of fractures: comparison of volume and surface reconstruction. Rofo 1998;169:490-4.

[16] Bjerner T, Johansson L, Haglund U, Ahlstrom H. 3D surface rendering of images from multiple MR pulse sequences in the pre-operative evaluation of hepatic lesions. Acta Radiol t 998,39:698- 700.

[17] Neri E, Boraschi P, Braccini G, Caramella D, Perri G, Bar- tolozzi C. MR virtual endoscopy of the pancreaticobiliary tract. Magn Reson Imaging 1999;17:59-67.

[18] Neri E, Caramella D, Falaschi F, et al. Virtual CT intravascular endoscopy of the aorta: pierced surface and floating shapes thresholding artifacts. Radiology 1999;212:276-9.

[19] Wielopolski PA, Gaa J, Wielopolski DR, Oudkerk M. Breath- hold MR cholangiopancreatography with three-dimensional, segmented, echo-planar imaging and volume rendering. Radiol- ogy 1999;210:247-52.

[20] Stockberger SM. Wass JL, Shcrman S, et al. Biliary obstruction: evaluation with three-dimensional MR cholangiography. Radiol-

ogy 1994;192:675-80. [21] Soto JA, Barish MA, Yucel EK, Ferrucci JT. MR cholangiopan-

creatography: findings on 3D fast spin-echo imaging. Am J Roentgenol 1995;165:1397-401.

[22] Meakem TJ III, Schnal! MD. Magnetic resonance cholangiography. Gastroenterol Clin N Am 1995;24:221-38.

[23] Reinhold C, Bret PM. Current status of MR cholangiopancreatography. Am J Roentgenol 1996; 166:1285-95.

[24] Boraschi P, Neri E, Braccini G, et al. Choledoco[ithiasis: diagnostic accuracy of MR cholangiopancreatography. Three-year experience. Magn Resort Imaging 1999;17:1245-53.

[25] Becker CD, Grossholz M, Mentha G, de Peyer R, Terrier F. MR cholangiopancreatography: technique, potential indications, and diagnostic features of benign, postoperative, and malignant con- ditions. Eur Radiol 1997;7:865-74.

[26] Wolff SD, Balaban RS. Assessing contrast on MR images. Radiology 1997;202:25-9.

[27] Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Meas 1960;20:37-46.

[28] Fleiss JL. Statistical Methods for Rates and Proportions, 2nd edn. New York: Wiley, 1973:211-36.

[29] Siegel S, Castellan NJ. Non Parametric Statistics for the Behav- ioral Science, 2nd edn. New York: McGraw-Hill, 1988.

[30] Klein HM, Wein B, Truong S, et al. Computed temographic cholangiography using spiral CT scanning and 3D image processing. Br J Radiol 1993;66:762-7,

[31] Fleischmann D, Ringl H, Schofl R, et al. Three-dimensional spiral CT cholangiography in patients with suspected obstructive biliary disease: comparison with endoscopic retrograde cholangiography. Radiology 1996;198:86t-8.

[32] Zeman RK, Berman PM. Silverman PM, et al. Biliary tract: three-dimensional helical CT without cholangiographic contrast material. Radiology 1995;196:865-7.

[33] Hamada Y, Sato M, Sanada T, Tsuji M, Kogata M, Hioki K. Spiral computed tomography for biliary dilatation. J Pediatr Surg 1995;30:694-6.

Documents

Real-time volume rendering of MRCP: clinical applications