BACKGROUND:

Endoscopic ultrasound (EUS) can reliably diagnose and stage pancreatic cancer but is less competent for the differentiation between vascular compression (VC) and invasion (VI).

AIM:

Prospective comparison of linear EUS with/without three-dimensional (3D) EUS for vessel involvement in pancreatic cancer to evaluate the feasibility of linear 3D ultrasound.

MATERIAL AND METHODS:

Linear echoendoscopy was used to identify the pancreatic tumor, the tumor-vessel relation and for EUS-FNA to obtain tissue diagnosis. Immediately afterwards, 3D image acquisition was performed using a magnetic tracked 3D sensor. The acquisition time was 10–20 s.

RESULTS:

EUS results of 22 patients with solid pancreatic lesions were compared to surgical histology. This proved adenocarcinoma in 17 patients and chronic pancreatitis in 5. EUS showed VI in 10 patients, VC in 6, and no vascular involvement (NVI) in 6. Additional 3D evaluation showed VI in 6 patients, VC in 10, and NVI in 6. Surgery proved VI in 7 patients, VC in 9, and NVI in 6. EUS showed VI in 3/5 patients with chronic pancreatitis, 3D showed VC only, while surgery found two patients to have VC and with NVI. In two patients with pancreatic cancer, VI was diagnosed on two dimensional (2D), but VC on 3D evaluation. Surgery showed VC and VI in one each. In the 2D, one patient with NVI had VI on surgery; and on 3D one VC proved to have NVI at surgery. In 1/22 patients the result of 3D was false negative, while 4/22 were false positives and one false negative in conventional EUS.

CONCLUSIONS:

Linear 3D EUS seems feasible for pancreatic evaluation. In addition, linear EUS enhanced the evaluation of vascular involvement of pancreatic lesions, especially in chronic pancreatitis.

INTRODUCTION

Endoscopic ultrasound (EUS) has emerged as a reliable tool for the detection of pancreatic lesions (^1,2,3) and the addition of fine needle aspiration (FNA) for tissue diagnosis has helped this method to evolve into an important diagnostic technique with an accuracy of <95% (^4,5,6). EUS can compete with other imaging techniques such as helical CT and MRI (^7,8). All of these modalities, however, have limitations in the accurate evaluation of vessel involvement by malignant pancreatic tumors (⁹), although this information may be particularly important for further management. This imaging limitation is the rationale for the exploration of new methods for better evaluation of vessel invasion (VI).

The addition of a three-dimensional (3D) capability to EUS has been described for radial probes especially for staging of rectal tumors (^10,11,12,13). Although no significant advantages were seen using radial 3D as an additive to EUS for rectal tumor staging, the stereoscopic visualization appeared to provide better understanding of the spread of the lesion (^10,11).

The generation of 3D images in radial EUS but even more in CT or MRI is relatively easy because the two-dimensional (2D) scans, that are to be converted into the 3D images, are taken with the machine in a fixed position and can therefore be clearly related to each other.

If a method is to be developed with which an improvement of VI can be evaluated during EUS and combined with FNA, it has to be with linear echoendoscopy. To date, the potential of linear array 3D image acquisition has been barely explored.

The main difficulty with developing a 3D capability for linear array EUS is that any angulation or torque of the transducer will produce a variety of scans from different planes and positions, which are neither parallel nor at a fixed angle to each other. Technical solutions have allowed the development of 3D imaging using transabdominal ultrasound (US) by placing a reference point within the US field and using transducers on the US probe. This allows the coordinates of the changes in angulation to be tracked. Amongst others, these developments have produced some sensational US images of fetus moving in the uterus.

We modified a commercially available 3D acquisition system for transabdominal US and attached it to the shaft of a linear array echoendoscope in order to obtain 3D images of the pancreas and associated vessels during routine EUS and FNA.

The aim of this prospectively conducted study was to evaluate the feasibility to create and examine readable 3D images using the difficult area of imaging the vascular involvement of pancreatic lesions to compare the analysis of 3D images to images acquired during standard linear EUS.

MATERIALS AND METHODS

Linear 3D

The linear 3D EUS examinations were performed with an electromagnetic positioning/tracking system ("Echotec" Hitachi Ultrasound, Tokyo, Japan) which is commercially available for transabdominal US. One part of the positioning system, an electromagnetic 2 kg block-shaped transmitter measuring 81010 cm³, was placed next to the abdominal wall of the patient on the examination bed (Fig. 1) and connected to a computer creating the images. This component generates a reference spherical magnetic field with an effective range of about 70 cm. Within this range, the position and movements of the probe on the endoscope tip can at any time be determined with the help of a 3 cm light weight sensor. This was fitted onto the shaft of a conventional echoendoscope without limiting the freedom of mobility of the examiner (Fig. 2).

Figure 1.

Prior to echoendoscopy, an electromagnetic tracking system (2 kg block-shaped transmitter of 81010 cm³) is placed next to the abdominal wall of the patient on the examination bed. This generates a reference spherical magnetic field of about 70 cm. Within this range, the position and movements of the probe can at any time be determined with the help of a 3 cm light-weight sensor, mounted on the shaft of the scope (Fig. 2).

Full figure and legend (244K)

Figure 2.

3 cm light-weight sensor, which is fitted onto the shaft of the echoendoscope. It can determine positioning and movements of the probe within the electromagnetic field created by the block shaped device (Fig. 1).

Full figure and legend (217K)

The 3D volume was scanned by a vertical shift of the entire echoendoscope. The sensor is able to differentiate between the following six movements: Shifts of the echoendoscope towards the directions X, Y, and Z and the rotation around the axis X, Y, and Z. Thus, the positioning system (the combination of the sensor and transmitter) ensures that 2D images can be spatially related to specific positions and that the single scan planes can then be transformed into a complete 3D volume image (Fig. 3). Metal objects placed near the structure to be examined may disturb the precision of the magnetic-tracked-procedure.

Figure 3.

3D volumetric image of the head of the pancreas in relation to the portal vein. The common bile duct is seen as echo-poor longitudinal structure on the top of the image.

Full figure and legend (130K)

In order to reduce artifacts induced during the generation of the 3D images, care has to be taken during the manual shifts and rotation of the echoendoscope since rapid movements cannot be transformed effectively into a 3D volumetric image and may result in black areas within the 3D image. This is easier to prevent while the echoendoscope is in the esophagus but essentially more difficult if the pancreas has to be scanned. After all the relevant planes of the anatomical structure of interest have been scanned, the system can start to calculate the spatial relationships of individual pixels.

EUS Examination

The EUS examination was performed by a single experienced examiner, who was blinded to the results of prior imaging studies but was informed about the possible presence of a pancreatic mass lesion and its suspected location within the pancreas. Pentax linear echoendoscopes (FG32UA and FG34UX, Pentax, Hamburg, Germany) and Hitachi Ultrasound consoles (EUB 8000 and EUB 6000; Hitachi Ultrasound, Zug, Switzerland) were used for the examination. All patients were referred for pancreatic surgery and underwent routine EUS and EUS-FNA. All patients gave specific informed consent for the procedure, the additional 3D evaluation, clinical follow-up and anonymous use of their data.

During the procedure, the findings and results of the 2D examination and the FNA were entered into the computer and "locked," so that they could not be altered at a later stage. Subsequent to the 2D EUS examination, FNA and writing the 2D report, the 3D evaluation took place during the same examination without removing the endoscope from the patient. After the sensor was attached to the shaft of the same endoscope, the reference transmitter block positioned near to the patient at the abdominal wall the EUS examination was repeated using the 3D equipment. Care had to be taken to ensure that the movements of the tip of the echoendoscope were paralleled by the movements of the shaft of the endoscope. The computer recorded the 2D images and transformed them into 3D. Two to four 10–20 s segments were taken to image the pancreatic lesion and the relevant vascular anatomy. The newly generated 2D sequences were inspected on the computer to assess quality before the procedure was terminated. The volume data reconstruction did not require more than a few seconds and the volume figures could have been analyzed immediately after acquisition while the patient was still on the examination bed. In order to optimize the interpretation of the 3D images of the pancreatic lesion in relation to its vascular anatomy, the image analysis, which took 15–20 min to complete, was performed after the EUS examination had been finished and the patient had been moved into the recovery room. A second report by the same examiner this time including the 3D evaluation was written and again "locked" on the computer.

Computer Analysis

The dataset size of the area scanned was 15–30 MB. The size was not dependent on the time but on the region of interest = volume of the head of pancreas; i.e., it was lower in normal size pancreas and higher in patients with large tumors, as the volume of the pancreas scanned increased.

Immediately after the scans had been taken the reconstructed 3D volumetric image was converted into a form that could be displayed on a 2D flat screen monitor using perspective "distortions" and the edges of a cube to increase 3D perceptions.

The postprocedure imaging analysis included multiplanar reconstruction out of the volume dataset, rotation and translation in three axis, photorealistic visualization or transparency, brightness, contrast and color modes as well as zoom mode.

Using Main Plane Imaging, all three scan planes—sagittal, transverse, and coronal plane, which are perpendicular to one another, can be displayed simultaneously. The physician can rotate the single planes on the monitor and scroll through the entire volume while all the acquired 2D images can be displayed on the screen.

As there are not sufficient orientation markers to match the volume sets, the acquired 3D dataset cannot be manipulated by reconstruction software used for MRI or CT.

Fundamentals of 3D Ultrasound Imaging

A prerequisite for the generation of 3D ultrasound images is that the geometric relationships occurring during the acquisition of series of scans and their reconstruction into 3D real-time images are recognized and solved. For scans performed with MRI systems or CT scanners, the regeneration and reconstruction of 3D images causes few problems since the 2D scans to be converted into the 3D images are taken of structures with known position and can be clearly related to each other.

With transabdominal US, the probe needs to be moved freely in various axes, which does not allow the generation of coordinated series of scans that can afterwards be reconstructed into 3D. In order to obtain high-quality 3D images, it is essential that the position of each individual scan, and of each pixel in an image is known or that the geometric relationships of the scans can be clearly related to each other. For the generation of a 3D volume image, the various different scan planes have to be converted and positioned accurately into a gate of specific structure.

By using the fixed external electromagnetic sensors as a reference, the position and orientation of the moving transducer can be located by appropriate measurements and the positions of conventional US images can be determined with an accuracy of about 1 mm.

RESULTS

Twenty-two consecutive patients (14 males, 48–77 yr, mean: 63 yr), who had been found by other imaging methods to have solid lesions in the head of the pancreas, were referred for preoperative staging with EUS and FNA to obtain evidence of the nature of the lesion. In four cases, these lesions had been recently detected in patients with known chronic pancreatitis, while the other patients presented with pain, weight loss, or jaundice but did not have a prior history of pancreatic abnormality. Endoscopic ultrasound and FNA results were compared with surgical histology in all of the cases.

All 22 patients were found to have pancreatic lesions both on 2D and 3D EUS imaging. The size of the lesion varied between 1.2 and 3.8 cm with a mean size of 3.0 cm. In 15 patients, the lesions seen were echo-poor, but were surrounded by normal pancreatic tissue, in 7 patients the echo-poor lesions were imbedded in pancreatic tissue that showed features suggestive of chronic pancreatitis.

The staging included EUS inspection of the splenic, portal, and mesenteric veins. Using 2D linear EUS VI was suspected in 10 patients, compression of a vessel due to the tumor mass (VC) in a further 6, and no vessel involvement in 6 patients (Table 1).

Table 1 - Results of Vessel Involvement of 22 Pancreatic Lesions using Conventional and 3D EUS Compared to Surgical Outcome.

Full table

In the additional 3D evaluation, VI was suspected in 6 patients, tumor compression of a vessel in 10, and no involvement in 6 patients (Table 1). The differentiation of VI and VC was based on images of the blood flow within the vessels evaluated in multiple spatial views but mostly from below the tumor (Fig. 4).

Figure 4.

3D image of a tumor in the head of the pancreas. It was thought that the tumor invaded the portal vein. On multiple 3D reconstructions, the images demonstrated that the blood appeared to flow freely around the tumor without local turbulences, indicating vessel compression. This was proven by surgery.

Full figure and legend (148K)

The EUS-guided cytology showed pancreatic cancer in 17 patients and focal chronic pancreatitis in 5. These results were confirmed by surgical histology in all. In addition, surgical histology and/or intraoperative evaluation demonstrated VI in 7 patients, VC in 9, and no vessel involvement in 6 patients.

Comparison of the Results of Conventional versus 3D EUS

In 3/5 of the patients with chronic pancreatitis, 2D EUS provided images, which gave rise to a suspicion of vessel infiltration, since shadowing due to calcifications made evaluation difficult. In these cases, 3D evaluation demonstrated VC only. Subsequent surgery found two of these patients to have VC and one to have no venous involvement at all (Table 1).

In two patients with pancreatic cancer, VI was diagnosed on 2D EUS, but VC on additional 3D evaluation. Surgery showed VC and VI in one each. In the 2D group, one of the cancer patients without any vessel involvement shown had invasion of the mesenteric vein on surgical exploration (Table 1). In the 3D group, one patient with VC at surgery was missed and evaluated as "no vessel involvement."

In only 1/22 patients was the result of 3D evaluation a false negative for VI (there were no false positives), while there were 4/22 false positives and one false negative in conventional EUS. Since false positive results might potentially exclude patients from surgery with curative intent, this fact seems important.

To evaluate whether the 3D interpretations of vascular anatomy could be reproduced by another examiner, recordings of all the 2D and 3D data from 22 patients, were passed to a radiologist with substantial experience in transabdominal US and moderate experience in EUS interpretation. She evaluated both the 2D and the 3D reconstructions of the 22 patients blinded to the outcome of the first investigator and without access to the history or other image findings of the patients. The only clinical information provided was that pancreatic malignancies were suspected in the head of the organ and that VI or compression was also suspected. Her results differed from those of the principle investigator only in two patients with chronic pancreatitis. In all patients without EUS evidence of chronic pancreatitis, there was concordance in the identification of VI.

DISCUSSION

There is a need to improve the imaging of vascular involvement of pancreatic masses with EUS. Roesch et al. analyzed EUS images of the vascular anatomy of 75 patients with pancreatic cancer. Sensitivity and specificity of EUS in the diagnosis of venous invasion were 43–62% and 79–91%, respectively (⁸).

The striking advances in 3D transabdominal US were the inspiration to apply this technique using a linear endoscopic approach. A commercially available 3D acquisition system was modified and used in patients to provide information about the relationship between pancreatic lesions and their surrounding vessels. Because the 3D system was constructed as an add-on, the 3D examination was performed as an optional extra after conventional EUS and FNA. The image generation time was quick and added only a maximum of 10 min to the patient examination time. No additional equipment was used inside the patient, so the add-on procedure appeared harmless.

In this study, the 3D images offered additional clinically valuable information, although images produced were sometimes blurred and of poor quality. This was usually due to unintentionally rapid movements of the tip of the echoendoscope being out of synchrony with slower movements of the shaft of the endoscope. There was a learning curve for the examiner using this new technology. To get the best out of the system three or four different 3D readings were taken for each of the patients in order to obtain at least one set of good quality images for interpretation.

Vascular invasion tended to produce local turbulence (Fig. 5). Vessel compression was diagnosed, when blood appeared to flow freely around the tumor without local turbulence. Peaks or nodules of tumor were also detectable within the vessel on examination of the multiple spatial views. These distinct but subtle changes in vessel behavior could not be detected in the 2D images, since views from below the vessels are not possible. These advantages made it possible to suggest VC rather than invasion in 3 out of 10 patients with VI detected on 2D (Table 1). 3D image interpretation gave a false negative result for VI in only one patient. Conventional EUS image interpretation gave four false positive patients for VI. This improvement in ability to differentiate VI from compression seems important, since in contrast to a more aggressive approach many surgeons would not attempt curative resection if imaging demonstrates the presence of vascular involvement (^14,15).

Figure 5.

3D image of a tumor in the head of the pancreas. There is local turbulence of the blood flow within the portal vein demonstrated in a view from below. These turbulences appear as white waves within the dark vessel, indicating vessel invasion.

Full figure and legend (160K)

Sumiyama et al. also reported using a 3D system in combination with a linear echoendoscope, which was limited to the use in the esophagus and stomach (¹⁶). This system also allowed simultaneous addition of the images to the constructed volume and the production of good quality 3D images. The sensor in their system was fitted at the tip of the echoendoscope. This arrangement has several advantages, giving better images because the movements of the tip can be directly transferred to the tracking system requiring less skill during the examination. A possible disadvantage of this system is that the attachment of the sensor to the tip of the scope increased the diameter of the endoscope shaft and altered the tip profile. In all three patients examined, an overtube was used to pass this system into the upper gastrointestinal tract and to provide smooth movements (¹⁶).

In our study, the sensor was fitted to the distal end of the echoendoscope outside the patient. This arrangement required precise handling in order to transmit the movements to the tip of the scope.

Technical improvements combining 3D imaging with the constraints of linear EUS image acquisition would be valuable. A possible technical solution for the difficulties caused by the sensor attachment to the outer shaft might be to construct a sensor that could be positioned at the tip of the echoendoscope via the accessory channel. This would avoid the problems of movement transfer, which arose in our study and might make the use of an overtube unnecessary as described in the 3D system developed by Sumiyama et al. (¹⁵). It might make the procedure easier and enable a larger community of EUS users to perform such examinations, especially since 3D acquisition is possible using some of the newer US consoles without requiring an additional computer system.

Microsensor-systems, which can determine the position of the transducer with an accuracy of <1 mm and could be inserted through the working channel, are in development. With the use of these or other improvements, it might be of value to assess the potential of 3D linear EUS in further and larger studies.

CONCLUSION

Better 3D view of the vessels near the pancreas might make staging of pancreatic tumors more reliable. In this pilot study of 22 patients, the additional 3D reconstructions provided appeared to improve the evaluation of vessel-tumor relationships in pancreatic cancer. 3D imaging using linear EUS might have other applications but the acquisition system needs to be improved.

References

1. Roesch, T, Lorenz, R, Braig, G, et al. Endoscopic ultrasound for pancreatic tumor diagnosis. Gastrointest Endosc 1991;31:(3 7):347–352.
2. Palazzo, L, Roseau, G, Gayet, B, et al. Endoscopic ultrasonography in the diagnosis and staging of pancreatic adenocarcinoma: Results of a prospective study with comparison to ultrasonography and CT scan. Endoscopy 1993;25: 143–150. | PubMed | ChemPort |
3. Gress, FG, Hawes, RH, Savides, RJ, et al. Role of EUS in the preoperative staging of pancreatic cancer. A large single center experience. Gastrointest Endosc 1999;50: 786–791. | PubMed | ChemPort |
4. Suits, J, Frazee, R, Erickson, RA. Endoscopic ultrasound and fine needle aspiration for the evaluation of pancreatic masses. Arch Surg 1999;134: 639–643.
5. Gress, FG, Gottlieb, K, Sherman, S, et al. Endoscopic ultrasonography-guided fine needle aspiration biopsy in the evaluation of pancreatic cancer. Ann Intern Med 2001;134: 459–464. | PubMed | ChemPort |
6. Harewood, GC, Wiesema, MJ. Endosonography guided fine needle aspiration biopsy in the evaluation of pancreatic masses. Am J Gastroenterol 2002;97: 1386–1391.
7. Dewitt, J, Ciaccia, D, Leblanc, J, et al. Prospective comparison of endoscopic ultrasound and dual phase helical computer tomography for the preoperative evaluation of known or suspected pancreatic malignancy: final results. Gastrointest Endosc 2003;57: AB 99 (abstract).
8. Howard, T, Chin, AC, Steib, EW, et al. Value of helical computed tomography, angiography, and endoscopic ultrasound in determining resectability of periampullary carcinoma. Am J Surg 1997;174: 237–241.
9. Rosch, T, Dittler, HJ, Strobel, K, et al. Endoscopic ultrasound criteria for vascular invasion in the staging of cancer of the head of the pancreas: a blind reevaluation of videotapes. Gastrointest Endosc 2000;52: 469–477. | PubMed | ChemPort |
10. Kim, JC, Cho, YK, Kim, SY, et al. Comparative study of three-dimensional and conventional endorectal ultrasonography used in rectal cancer staging. Surg Endosc 2002;16: 1280–1285.
11. Hünerbein, M, Schlag, M. Three-dimensional endosonography for staging of rectal tumor. Ann Surg 1997;225: 432–438.
12. Hünerbein, M, Pegios, W, Rau, B, et al. Prospective comparison of endorectal ultrasound, three-dimensional endorectal ultrasound, and endorectal MRI in the preoperative evaluation of rectal tumors. Preliminary results. Surg Endosc 2000;14: 1005–1009.
13. Szilvas, A, Szekely, G, Tarjan, Z, et al. Three-dimensional ultrasonography, virtual colonoscopy and endorectal magnetic resonance imaging in the diagnosis of complicated inflammatory bowel disease. Endoscopy 2002;34: 94–99.
14. Nakao, A, Takeda, S, Sakai, M, et al. Extended radical resection versus standard resection for pancreatic cancer: The rationale for extended radical resection. Pancreas 2004;28: 289–292. | PubMed |
15. Park, DI, Lee, JK, Kim, JE, et al. The analysis of resectability and survival in pancreatic cancer patients with vascular invasion. J Clin Gastroenterol 2001;32: 231–234.
16. Sumiyama, K, Suzuki, N, Tajiri, H. A linear-array freehand 3-D endoscopic ultrasound. Ultrasound Med Biol 2003;29: 1001–1006.