17 medicamundi. Virtual endoscopy is a recent development in ... Virtual endoscopic view from the right nasal cavity after .... virtual colonoscopy (Fig. 7). There is ...
P. Rogalla1
Virtual endoscopy: an application snapshot
Virtual endoscopy is a recent development in postprocessing techniques for cross-sectional images. Unlike conventional 3D reconstructions, the images obtained are given a perspective which, when highly magnified, creates the impression of a true endoscopic image [1]. There are two basic methods that can be used to obtain an artificial surface: surface rendering, in which a wireframe model consisting of small triangles is reconstructed [2], or volume rendering [3]. In both cases it is possible to alter the viewing angle, light source, depth encoding, shading effects and surface characteristics in the reconstructed images. The advantage of using the surface rendering technique lies in the marked reduction of the amount of data required. The surface is determined by means of thresholding. Once the surface has been determined or, in other words, the segmentation process is complete, relatively little computing power is required to calculate images resembling fibre endoscopic views in a real-time fashion, making this technique most suitable when inexpensive hardware has to be used [4]. However, the most prominent disadvantage is the limitation in the 3D spatial resolution, which becomes most apparent in small structures, where the triangles of the wireframe model become visible. In contrast, the volume rendering technique requires considerably more computing power, because the entire volume data set will be calculated for each selected view, according to a classification in which the user has defined which tissues should be transparent and which should be opaque. This has the advantage of allowing the user to modify the classification interactively without needing to resegment the volume. Furthermore, the volume-rendered images provide higher spatial resolution, making this the preferred technique for virtual endoscopy (3D EndoView).
Clinical applications
Nose and paranasal sinuses Functional endoscopic sinus surgery, originally developed in Austria, has become the standard operation procedure for the sinuses all over the world. Preoperative CT scanning of the paranasal sinuses is accepted today as a standard procedure [5, 6, 7, 8, 9]. Some authors prefer an axial slice orientation with coronal reconstruction [10, 11]. This examination technique has the advantage for the patient that an overextended supine position for the coronal slice scanning can be avoided. However, the disadvantage lies in the decreased spatial resolution that can be expected in the coronal reconstruction. The introduction of spiral CT technique has helped to eliminate this problem because, with overlapping reconstructions from primarily very thin slices, high quality secondary reconstructions are attainable [11]. In addition, other reconstructions such as those in the sagittal plane can be calculated from a spiral data set. Nevertheless, the surgeon must combine these images
Virtual endoscopy creates the impression of a true endoscopic image from the CT data. Functional endoscopic sinus surgery has become the standard procedure.
Virtual reconstruction of the CT data saves the mental effort of envisioning the operation site.
Fig. 1. Virtual endoscopic view from the right nasal cavity after functional endoscopic surgery. The artificial entry into the right maxillary sinus can be seen (arrow).
1Department of Radiology, Charité Hospital, HumboldtUniversität zu Berlin, Berlin, Germany.
Volume 43 Issue 1 March 1999
medicamundi
17
Surgeons reported a high degree of similarity between virtual endoscopy and the intraoperative impression. Virtual endoscopy may become the standard means for guiding surgeons.
mentally in order to envision the operation site. The virtual reconstruction from the CT data (Fig. 1) is an optimum means of avoiding this effort. In order to obtain good quality for such images, spiral data with high spatial resolution are just as important here as they are for secondary reconstructions. The thinnest possible beam collimation with overlapping reconstruction is a prerequisite. In a prospective study [12], 45 patients underwent low-dose spiral CT of the sinuses. The data were transferred to a workstation running software for volume rendering (EasyVision, Rel. 4.2, with Endo 3D). Six orthogonal views of the maxillary sinuses and the nasopharynx (Fig. 2) were calculated, as well as a ‘fly-through’ movie of the nose. Two radiologists assessed the coronal reconstructions and virtual endoscopy with respect to the ease of detecting pathology. In 30 patients who underwent subsequent endoscopic surgery, surgeons were asked to rank the degree of assistance of the preoperative virtual endoscopy. In this study virtual endoscopy was possible in all 45 patients. The mean times required for path definition, and movie calculation for virtual endoscopy, were 8 (± 2) minutes and 3 (± 1) minutes, respectively. Overall, more anatomical details were depicted in the coronal reconstructions. For example, in 11 out of 45 patients (24 %) one or both maxillary sinuses were not approachable with virtual endoscopy. In these cases, a full occlusion could be observed in the coronal slices. For the same reason, i.e., occlusion due to mucosal swelling, the upper nasal conchae could not be displayed in the coronal slices in 8 patients (17 %), and in virtual endocopy in 20 patients (44 %). Nonetheless, a high degree of similarity between virtual endoscopy and the intraoperative impression was reported by the surgeons. Based on the results from the study, and the ongoing experience, one can conclude that virtual endoscopy of the nose and paranasal sinuses represents more of a preoperative than a diagnostic assistance for the surgeon, and may develop into a standard means for guiding surgeons during endoscopic interventions. Larynx and tracheobronchial tree As in the area of the nose and paranasal sinuses, the contrast in the larynx and tracheobronchial tree is provided by air against soft tissue, representing the best possible contrast in computed tomography. Being easily accessible
18
medicamundi
Volume 43 Issue 1 March 1999
Fig. 2. 55 year old patient with posterior rhinorrhea. The 3D endoview of the caudal aspects of the conchae demonstrates a mild polypoid swelling of the left inferior nasal concha (inferior turbinate).
by true endoscopy, the larynx and particularly the trachea (Fig. 3) have been investigated by several authors [4, 13]. Aschoff et al. concluded from a series of 12 patients that virtual endoscopy might play a significant role in the noninvasive assessment of the larynx, if the search for tumours is the primary indication. However, the lack of detailed information on the mucosa, and the presence of artifacts due to swallowing, are limitations which prevent the use of virtual endoscopy in the detection of tumours limited to the mucosa, and tumours of the vocal cords. Compared with the nose and paranasal sinuses, Fig. 3. Normal aspect of the tracheal bifurcation on virtual endoscopy. The circular rings in the tracheal wall are a pulsation artifact caused by transmitted heart beats and may be easily misdiagnosed as the tracheal cartilages.
Fig. 4. View into a subsegmental bronchus (normal appearance). The original axial slices were obtained using 1.5 mm slice thickness. Despite the high-power magnification (1 mm marker), morphological details of the bifurcation are still visible.
the tracheobronchial tree imposes a different challenge to the scanning protocol: for optimization of the spatial resolution in virtual endoscopy, the thinnest slice thickness possible would be desirable, which, on the other hand, would decrease the coverage during one breathhold period (Fig. 4). As a consequence, most authors have opted to use a 3 mm beam collimation combined with a pitch factor larger than 1:1, making it possible to obtain a complete scan of the tracheobronchial tree [14]. The disadvantage is the visible limitation in imaging small bronchi that have a lumen smaller than the slice width.
Fig. 5. Virtual endoscopic view into a normal small bowel (jejunum). The irregularities in the folds are caused by noise in the axial images.
A possible solution to this problem would be an imaging modality that allows for rapid scanning during one breathhold at very thin beam collimation. Currently, such application is possible using electron beam tomography [15], with which the entire chest can be examined in 17 s at 1.5 mm slice thickness. The resulting virtual endoscopic pictures exhibit remarkably improved spatial resolution, allowing exceptional views into bronchi of less than 3 mm diameter. It is expected that, with the introduction of multi-slice CT scanners, virtual endoscopy of the bronchi will gain increased clinical acceptance, as it will then be possible to calculate the virtual endoscopic images from virtually any routine chest CT examination, without the need to perform an additional CT scan using a modified scanning protocol.
Virtual endoscopy of the bronchi will gain increased clinical acceptance.
Small intestine Despite advances in fibre optic endoscopy, the greater part of the small bowel has still remained inaccessible. Although the very small endoscopes that have recently been developed allow true endoscopy of the duodenum and proximal jejunum, the conventional, double-contrast small bowel enema [16] currently represents the only reliable technique for assessing the small bowel; however, visualization of the external tissue is lacking. The majority of CTs for abdominal survey are performed after administration of oral contrast material [17, 18]. However, there is no guarantee that the small bowel will be filled homogeneously and, in most cases, there are intermittent areas in which no filling of bowel loops can be observed despite optimal drinking patterns [19]. A possible solution is a small bowel preparation technique as used for the conventional double-contrast enteroclysis: placing a duodenal tube with its tip behind the ligament of Treitz [20], followed by the application of methyl cellulose mixed with oral contrast. In our own series of 21 patients examined using the duodenal intubation technique, complete opacification of the small bowel was achieved in 18 patients with only a few remaining air bubbles of up to 10 mm in diameter (Fig. 5). The resulting intestinal density ranged from 360 – 470 HU, which was found to be the best compromise between obtaining a high contrast against the soft tissue and avoiding beam hardening artifacts. The original axial scans were of diagnostic quality. By reconstructing at Volume 43 Issue 1 March 1999
medicamundi
19
Fig. 6. Pre-terminal ileum in a patient with a long-term history of Crohn’s disease.
Intrinsic pathology (stenoses, fistulae, polyps) was seen on both axial and virtual endoscopic images. Fig. 6 a. Axial slice showing marked wall thickening.
Fig. 6 b. Corresponding virtual endoscopy of the ileum demonstrating the stenosis of the intestinal lumen. Fig. 7. Regular view into the transverse colon after optimal distension with carbon dioxide gas.
Early detection of colorectal polyps is a major goal in healthcare management.
20
medicamundi
Volume 43 Issue 1 March 1999
least 30 views along a manually drawn path through the small bowel, the ‘intrinsic’ pathology (stenoses, fistulae, polyps) was seen on both the axial images and the virtual endoscopic images (Fig. 6 a, b), whereas ‘extrinsic changes’ such as lymph nodes, mesenteric metastases, cystic ovaries, pseudocysts etc., escaped detection by virtual endoscopy. In view of the time that is required for drawing the path through the small bowel (it occasionally takes more than 30 minutes), the clinical value has to be critically evaluated. Due to the small beam collimation of 3 mm, image noise becomes influential and can be recognized in the reconstructed images. Increasing the tube current would improve the image quality, but, at the same time, it would increase the radiation exposure to the patient, which sets a limitation to this approach. Using a softer reconstruction filter would decrease the image noise at the expense of spatial resolution within the scan plane. However, along with the development of improved tools for path definition through the bowel, further improvements of the CT equipment such as rapid scanning capabilities, higher detector sensitivity, sharper slice sensitivity profiles etc., virtual endoscopy of the small bowel might establish itself as a useful adjunct in abdominal imaging techniques. Colon Colorectal cancer represents the second leading cause of cancer death in most industrialized countries. For the United States, it was estimated that 131 600 new cases occurred in 1998 [21]. Early diagnosis and preventive removal of pre-malignant adenomatous polyps are believed to be the most probable reasons for a gradual decrease in both incidence and mortality since 1985 [22]. It is generally accepted that the vast majority of colorectal cancers arise from adenomatous polyps. Bearing in mind that malignant transformation is found in approx. 1 % of polyps of less than 1 cm in size, compared with 10 % of larger polyps [23], early detection of colorectal polyps might be regarded as one of the major goals in health care management. Several studies have focused on the principal characteristics of virtual endoscopy [24, 25, 26]. Dependent on the CT scanner speed, most authors use a primary beam collimation of 3 to 5 mm, at table speeds varying from 5 to 8 mm/s. Such a protocol allows reliable detection of polyps that have a size which is not smaller than the primary beam
collimation. Optimal distension of the colon through rectal insufflation of air or carbon dioxide after bowel cleansing is a prerequisite for virtual colonoscopy (Fig. 7). There is an ongoing debate about how a complete distension of the colon can be achieved: most authors propose scanning the patient in the prone and supine position, accepting that this approach will double the radiation exposure to the patient, the amount of images and the postprocessing time. Remaining fluid levels impose another problem, since polyps may be hidden ‘below the surface’ and would therefore not be visible in virtual endoscopy (Fig. 8). Giving oral contrast material in combination with the bowel cleansing might be a solution, since polyps exhibit a negative contrast when covered by the contrast material. In an initial assessment of the sensitivity and specificity of virtual endoscopy, Hara et al. [27] found a 75 % sensitivity and 90 % specificity for polyps of 10 mm or more in size, and a 66 % sensitivity and 63 % specificity for polyps between 5 and 10 mm in size. Dachman et al. [25] reported 83 % sensitivity for simulated polyps larger than 8 mm. We have examined 63 patients who had a known colorectal cancer or strong suspicion of such, and a total of 53 tumours in all stages were found on virtual endoscopy (Fig. 9 a, b). The sensitivity for the malignant tumours was 100 %. However, the smallest tumour was 8 mm in size (Fig. 10). For the benign polyps (size range: 3–7 mm) including a 10 cm haemangioma, the sensitivity was 91 % and the specificity 85 %. A screening technique is defined as the investigation of individuals at risk for a disease prior to the development of symptoms. Despite the promising results reported in the literature for virtual endoscopy as a screening technique for colorectal cancer, the reliability of this technique still needs to be further investigated. It remains of interest to determine whether the reported detection rates for polyps can be reproduced in a population with a low prevalence of polyps, as would be the case in a screening setting.
Fig. 8. Despite optimal preparation of the patient (bowel cleansing), some remaining fluid levels are often seen in the rectal ampulla. Small polyps may be obscured and not visible on virtual endoscopy. Fig. 9. 63 year-old female with repeated positive faecal occult blood test (FOBT). Fig. 9 a. Axial slice (3 mm thickness) showing a 2 cm polypoid tumour without evidence of transmural growth.
The sensitivity for the malignant tumours was 100 %.
Fig. 9 b. The virtual endoscopy allows additional perception of the tumour shape (final tumour stage after surgery: pT2N0).
Conclusion
Virtual endoscopy is a fundamentally new imaging technique with the potential to alter current clinical approaches to many diseases. Besides applications in the nose and paranasal sinuses, the larynx and the tracheobronchial tree, the Volume 43 Issue 1 March 1999
medicamundi
21
Fig. 10. 8 mm polyp located in the ascending colon. The polyp was endoscopically removed, and histology demonstrated malignancy.
Virtual endoscopy will undoubtedly play a role in the future of whole body imaging.
small bowel and the colon, applications in the vasculature such as the aorta (Fig. 11), the carotids, and even in the coronary arteries are currently being evaluated. Along with improvements in the spatial and temporal resolution of the imaging modalities, further improvements for automated path definitions [28], improved tools for user orientation during the virtual
Fig. 11. Virtual view from the neck of an abdominal aortic aneurysm. Using two different classifications for the vessel lumen containing contrast material and for calcifications, the calcified plaques at the vessel wall (arrow) can be nicely seen on virtual endoscopy.
endoscopy, and improved reconstruction speeds allowing real-time flythroughs at high resolution, virtual endoscopy will undoubtedly play an increasing role in the future of whole body imaging.
References [1] Satava RM. Virtual Endoscopy: Diagnosis using 3-D Visualization and Virtual Representation. Surg Endosc 1996; 10(2): 173 –174.
[8] Babbel R, Optimization of Techniques in Screening of the Sinuses. AJNR 1991; 12: 849 – 854.
[2] Summers RM. Navigational Aids for Real-Time Virtual Bronchoscopy. AJR 1996; 168: 1165 –1170.
[9] Chow JM, Radiologic Assessment Preoperative to Endoscopic Sinus Surgery. Otolaryngol Clin North Am 1989; 22: 691–705.
[3] Rubin GD, Beaulieu CF, Argiro V et al. Perspective Volume Rendering of CT and MR Images: Applications for Endoscopic Imaging. Radiology 1996; 199: 321– 220.
[10] Hosemann W, Endonasal Surgery of the Paranasal Sinuses [Die endonasale Chirurgie der Nasennebenhöhlen]. Eur Archives of Oto Rhino Laryng 1996; 1: 155 – 270.
[4] Aschoff AJ, Seifarth H, Fleiter T et al. HighResolution Virtual Laryngoscopy Based on Helical CT Data Sets [Hochauflösende Virtuelle Laryngoskopie aus Spiral-CT-Datensätzen]. Radiologe 1998; 38: 810 – 815.
[11] Suojanen JN and Regan F. Spiral CT of the Paranasal Sinuses. AJNR 1995; 16: 787–789.
[5] Chakeres D. Computed Tomography of the Ethmoid Sinuses. Otolaryngol Clin North Am 1985; 18: 29 – 42.
22
medicamundi
[12] Rogalla P, Nischwitz A, Huitema A, Kaschke R, Hamm B. Virtual Endoscopy of the Nose and Paranasal Sinuses. Eur Radiol 1998; 8: 946 – 950.
[6] Som PM, CT of the Paranasal Sinuses. Neuroradiology 1985; 27: 189 – 201.
[13] Silverman PM, Zeiberg AS, Sessions RB et al. Helical CT of the Upper Airway: Normal and Abnormal Findings on Three-Dimensional Reconstructed Images. Am J Roentgenol 1995; 165: 541–546.
[7] Beus J, Kauczor HU, Schwikkert HC, Mohr W, Mildenberger P. Coronal Paranasal Sinus CT using the Spiral Technique. Aktuelle Radiol 1995; 5: 189 –191.
[14] Ferretti GR, Vining DJ, Knoploch J, Coulomb M. Tracheobronchial Tree: Three-Dimensional Spiral CT with Bronchoscopic Perspective. J Comput Assist Tomogr 1996; 20: 777–781.
Volume 43 Issue 1 March 1999
[15] Rogalla P, Enzweiler C, Schmidt E et al. Thoracic Diagnostics with Electron Beam Tomography [Thoraxdiagnostik mit der ElektronenstrahlComputertomographie]. Radiologe 1998; 38: 1029 –1035. [16] Sellink JL, Examination of a Small Intestine by Duodenal Intubation. Acta Radiol 1974; 15: 318. [17] Silverman PM, Handbook of Helical (Spiral) Computed Tomography: Techniques & Protocols 1994, Corona Del Mar, CA, USA: Mallinckrodt Medical Inc by Med Write Inc. [18] Rogalla P, Mutze S, Hamm B. Body CT: Stateof-the-Art 1996. Munich: W. Zuckschwerdt, Germany. [19] Thoeni RF and Filson RG. Abdominal and Pelvic CT: Use of Oral Metoclopramide to Enhance Bowel Opacification. Radiology 1988; 169: 391– 393. [20] Rogalla P, Werner-Rustner M, Huitema A et al. Virtual Endoscopy of the Small Bowel: Phantom Study and Preliminary Clinical Results. Eur Radiol 1998; 8: 563 –567. [21] Cancer Facts and Figures. American Cancer Society 1998, 2. American Cancer Society, Inc. Atlanta, USA.
[22] SEER Cancer Statistics Review – 1973 –1992, US Department of Health and Human Services, Public Health Service, NIH, 1995. [23] Muto T, Bussey HJR and Morson BC. The Evolution of Cancer of the Colon and Rectum. Cancer 1975; 36: 2251–2270. [24] Hara AK, Johnson CD, Reed JE, Ehmann RL and Ilstrup DM. Colorectal Polyp Detection using Computed Tomographic Colography: Two-Versus Three-Dimensional Techniques. Radiology 1996; 200: 49 –54. [25] Dachman AH, Lieberman J, Osnis RB, et al. Small Simulated Polyps in Pig Colon: Sensitivity of CT Virtual Colography. Radiology 1997; 203: 427– 430. [26] Beaulieu CF, Napel S, Chin IY, et al. Detection of Colonic Polyps in a Phantom Model: Implications for Virtual Colonoscopy Data Acquisition. JCAT 1998; 22: 656 – 663. [27] Hara AK, Johnson CD, Reed RE, et al. Colorectal Polyp Detection using CT Colography: Initial Assessment of Sensitivity and Specificity. Radiology 1997; 205: 59 – 65. [28] McFarland EG, Wang G , Brink JA, et al. Spiral Computed Tomographic Colonography: Determination of the Central Axis and Digital Unravelling of the Colon. Acad Radiol 1997; 4: 367– 373.
Volume 43 Issue 1 March 1999
medicamundi
23
