THREE-DIMENSIONAL IMAGING AND DISPLAY OF RENAL TUMORS USING SPIRAL CT: A POTENTIAL AID TO PARTIAL NEPHRECTOMY

Transcription

1 PRELIMINARY COMMUNICATION THREE-DIMENSIONAL IMAGING AND DISPLAY OF RENAL TUMORS USING SPIRAL CT: A POTENTIAL AID TO PARTIAL NEPHRECTOMY DANIEL M. CHERNOFF, M.D., PH.D. STUART G. SILVERMAN, M.D. RON KIKINIS, M.D. DOUGLASS E ADAMS, M.D. STEVEN E. SELTZER, M.D. JEROME P. RICHIE, M.D. KEVIN R. LOUGHLIN, M.D. From the Departments of Radiology and Surgery, Division of Urology, Brigham and Women's Hospital, Boston, Massachusetts ABSTRACT--Objective. A new technique for creating three-dimensional (D) images of renal tumors using contrast-enhanced spiral computed tomography (CT) is described and preliminarily investigated. Methods. D spiral CT was employed in 2 patients before radical nephrectomy and in S patients before partial nephrectomy. Preoperative and postoperative image analyses were conducted to evalu,,te the ability of the images to depict key anatomic relationships in planning partial nephrectomies. Results. D spiral CT defined the tumor's location and relationship to the kidney surface better than the tumor's proximity to renal hilar vessels and collecting system. Negative surgical margins were obtained in all patients with renal cell carcinoma, and postoperative serum creatinine remained less than 2 mg/dl in all S patients after partial nephrectomy. Conclusions. This early experience suggests that D spiral CT can help in the planning of partial nephrectomy and in attaining complete resection of renal cell carcinoma while conserving normal renal tissue. Partial nephrectomy is a treatment option in patients with renal cell carcinoma who have renal insufficiency, solitary kidneys, contralateral renal abnormalities, bilateral tumors, and in patients with indeterminate renal masses. 1,2 Inadequate tumor resection and postoperative renal insufficiency are two potential complications which might be prevented by preoperative surgical planning using three-dimensional (D) imaging. To date, D imaging techniques have been applied mostly to preoperative planning of brain, maxillofacial, and orthopedic surg~ry., These areas are particularly amenable to D imaging because there is minimal motion artifact during imaging. In distinction, D imaging of the abdomen using computed tomography (CT) has been limited by respiratory misregistration (imag- Submitted: July 15, 199, accepted (with revisions): September 1, 199 ing gaps between individual CT sections), because conventional CT techniques are too slow to permit acquisition of a volume of interest during a single breath-hold. A major technical advance has been the development of continuous acquisition spiral CT, which eliminates respiratory misregistration between slices by acquiring the entire CT data set during a single breath-holdp In this brief communication, we demonstrate D surface renderings of renal tumors and surrounding structures by combining contrast-enhanced spiral CT with a multithreshold segmentation technique based on probability and connectivity, 6 and report our preliminary experience with this technique before both radical and partial nephrectomy. MATERIAL AND METHODS Spiral CT (Somatom Plus or Somatom Plus-S; Siemens Medical Systems, Iselin, NJ) was UROLOGY / JANUARY 199 / VOLUME, NUMBER

2 TABLE I. Summary of patients and results of nephrectomies using D imaging Case Age/ Size Stage/ Creatinine (mg/dl) No. Sex Indication (cm) Diagnosis Location Grade Margin Preop. Postop. 1 67/M Indeterminate 2 Cyst R LP.. (Radical).... cyst 2 21/F Indeterminate.5 Angiomyo- R MP., (Radical).... mass lipoma /F Indeterminate.5 Cyst R UP cyst 65/M Solitary kidney 2.5 RCC R LP 11/11 2 mm 1.5 1,7 5 S2/M Contralateral 7 RCC L UP 1/11 1 mm atrophy 6 59/F Solitary kidney 10 RCC L UP Ifl 10 mm /F Bilateral renal 2 RCC R UP I/I mm artery stenosis Key: R = right; L = left; UP = upper pole; MP = mid pole; LP = lower pole; RCC = renal cell carcinoma. performed with 5 mm collimation and with a 5 mm/revolution table feed, first without intravenous contrast media and again after intravenous bolus-injection of contrast media (0-2 g iodine, 2.5 cc/sec injection rate, 100 sec delay until imaging started). All scans were performed at 120 kvp and 165 ma. Each scan was obtained during a single breath-hold of twenty-four or thirty-two seconds at full inspiration, with a 6 cm field of view and a 512 x 512 image matrix. The first and last 5 mm slices were discarded, yielding an 11 cm (2 sec breath-hold) or 15 cm (2 sec breathhold) volume of imaged tissue. Five mm thick sections were reconstructed from the spiral CT data set at mm increments. The spiral CT was used for diagnosis and staging. Only the postcontrast images were used in the D reconstruction. These data were transferred over a local area network to an imaging workstation (Sun -70, Sun Microsystems, Mountain View, CA) for further processing. The original 512 x 512 matrix was reduced to 256 x 256 and low-pass filtered (to reduce quantum mottle) prior to segmentation; the resulting voxel size was approximately 1.5 mm in transverse dimension and 5 mm perpendicular to the imaging plane. The D reconstruction was performed by a radiologist (D.M.C.) using a supervised classification algorithm. 6 In brief, this algorithm consists of manually sampling points from each tissue of interest by "pointing and clicking" with a mouse device. The mean and standard deviations of attenuation for each tissue were calculated from the sampled points assuming a normal distribution. Using a maximum likelihood criterion, each voxel was assigned to a tissue class based on its attenuation. The initial segmentation ("label map") was displayed alongside the original image, with colors representing each tissue class. Misclassified voxels were visible as islands of incorrect color. The resulting label maps were manually edited using a variety of morphometric tools to correct misclassifled voxels. 6 A dividing cubes algorithm 7 was applied to the corrected label maps to create a list of surface points and surface normals for each tissue. The resulting surfaces were displayed on the Sun workstation using a gradient shading technique, representing each tissue by a different color. Rotations of the image can be performed in software in about 5 sec/view, the views stored and played back as a movie loop at a later time. To plan surgery and display the anatomy of overlapping surfaces, the operator can cut, rotate, and translate individual or multiple tissues. We currently use a custom accelerator board (General Electric Co., Milwaukee, WI) to complete each surface rendering in less than 0.1 sec, permitting "real-time" rotation and cutting of surfaces with a mouse or joystick device. The referring urologist and radiologist met preoperatively to analyze the D data set for each patient. The images were manipulated using the aforedescribed tools to display to best advantage the tumor and its relationship to surrounding structures and were also presented as a rotating model on a video display. Selected views were photographed for reference during surgery. In 1 case, a mobile workstation was brought into the operating room to provide on-line display of the D surface anatomy. The operating urologists completed surveys after the surgery in of 5 patients who underwent partial nephrectomy. These surveys assessed the 1 26 UROLOGY / JANUARY 199 / VOLUME, NUMBER i

3 Fifty-two-year-old male with a 7 cm left upper pole renal cell carcinoma and an atrophic right kidney (Case 5). (A) Axial CT image demonstrates a solid renal mass within the left upper pole. (B) Anterior D view shows the liver, spleen, and pancreas (dark brown), aorta (red), vena cava (blue), Kidney (tan), collecting system (yellow), tumor mass (green), and a small anterior cortical cyst (purple). (C) Left lateral view of the kidney shows relationship of the cyst to the inferior margin of tumor. (D) Right posterior oblique view with sectioning of the kidney shows the close proximity of tumor to the renal hilum. At surgery, the cortical cyst was used as a surface landmark to help judge the inferior extent of the tumor mass. FIGURE 1. Results of surveys of urologists' opinions of D imaging following partial nephrectomy TABLE I1. urologist's opinion of the D displays in predicting the l o c a t i o n a n d e x t e n t of the t u m o r at surgery as well as the utility of the imaging in aiding partial nephrectomy. RESULTS Representative images from D surface recons t r u c t i o n s in 2 patients c o n s i d e r e d for partial nephrectomy are shown in Figures 1 and 2. Surgical results and pathologic correlation for all 7 patients are listed in Table I. Resection margins were free of tumor in all cases. Two patients (Cases 1 and ) underwent radical nephrectomy. In Case 1, this was because a preoperative diagnosis of an angiomyolipoma could not be firmly established, and a partial nephrectomy was not technically feasible. In Case, the tumor appeared to be more invasive at surgery than on the CT study. Based on the urologist's evaluation (Tables II and III), D spiral CT was more helpful in depicting the tumor's size and location than its relationship to the collecting system and blood vessels. UROLOGY / JANUARY 199 / VOLUM~, NUMBER 1 How well was Feature/ Relationship depicted? Size of lesion Lesion intrarenal location Correct pole for lesion Correct surface for lesion Correct lesion depth collecting system vessels Case No. S 6 7 Mean other organs Grading scale: 1 = complete failure; 2 = mostly disagree; = partial success; = complete success. The urologist's grading of the D imaging in terms of its utility in surgical planning suggests that it was most helpful in sparing normal renal tissue by guiding an accurate nephrotomy. 1 27

4 FIGURE 2. Fifty-nine-year-old female with a 10 cm upper pole cystic renal cell carcinoma in a solitary kidney (Case 6). (A) On CT section through the lower portion of the carcinoma, two smaller cysts (arrows) are seen anteriorly. 1) image in the frontal projection (B) shows relationship of the carcinoma (green) and smaller cysts (purple), to the collecting system, vasculature, and surrounding organs. Lateral projection (C) demonstrates the view of the tumor and kidney anticipated at surgery. (D) Cut section through the kidney in the coronal plane shows the close proximity of the tumor to the upper pole calix. Results of surveys of urologists' opinions of D imaging following partial nephrectomy TABLE III. Case No D Imaging Helped spare more nephrons Helped minimize blood loss Shortened operative time Allowed operative plan to be carried out Altered operative plan (compared to plans based on 2D imaging) Overall rating as surgical planning tool Grading scale: 1 strongly agree. = Mean strongly disagree; 2 = mostly disagree; = agree somewhat; = COMMENT In this study, the combination of contrast-enhanced spiral CT, image segmentation, and surface-rendering display software creates D surface images of the kidney and other upper abdominal 128 structures that are virtually free of motion artifact and respiratory misregistration. This preliminary experience suggests that'd spiral CT is a useful adjunct to the axial CT images in visualizing the relationship between the tumor, renal hilar vessels, collecting system, and the kidney surface prior to and during surgery. In particular, the D display of the tumor location aided in the performance of an accurate nephrotomy, which in the urologist's opinion facilitated the attainment of negative resection margins and helped preserve nephrons. Although intraoperative ultrasound has been reported to aid partial nephrectomy,8,9 it does not provide an overall view of the anatomy as with D spiral CT, and does not allow the urologist to plan the surgery before the operation. We emphasize that the D images provide no additional diagnostic information over that provided by the original axial CT images. In fact, CT attenuation data and spatial resolution are lost in the multithresholding and surface reconstruction process. An alternative method of display involving no loss of information is 2D reformatting in UROLOGY / JaNu~v 199 / VOLUME, NUMBER1

5 coronal and sagittal planes, a tool available on most CT scanners. However, in our view, the principal advantage of the D display is that it helps the urologist perceive D spatial relationships "at a glance" as they might be viewed at surgery, and avoids the need to mentally reconstruct the relationships from individual CT slices. Several limitations of this D imaging technique require discussion. As with all renal imaging modalities using contrast media, selection of the phase of enhancement (vascular, nephrographic, and pyelographic) determines which anatomic regions are depicted with greatest contrast. Because renal parenchymal tumor detection and characterization was the main purpose of the CT, the nephrographic phase of enhancement was chosen. This resulted in only partial visualization of the collecting system in most cases, and may explain the urologists' lower grades in this regard. However, a second delayed spiral CT sequence, if obtained, could be used to "add" a three-dimensional pyelogram. CT angiography, using a higher rate of contrast media injection and less delay between injection and imaging, could also be performed to optimize contrast between the renal vessels and surrounding tissues. 1 This might allow D visualization of the intrarenal branch vessels, at a cost of increased x-ray exposure and increased D reconstruction time. A second limitation of the described technique is that the imaged volume was restricted to 15 cm in craniocaudal extent, which in some patients did not encompass the entire renal volume. Improved instrumentation has already extended imaging time to forty seconds (19 cm volume with 5 mm slice thickness). Advances in reconstruction algorithms, and faster CT table feed speeds, n can now be used to extend the imaged volume during a single spiral CT scan. These will permit imaging most of the abdomen, including both kidneys, in nearly every patient. A third limitation is the time and level of training required for D segmentation by the multithresholding technique. Segmentation of the CT data set using the described techniques requires the expertise of a radiologist and a minimum of two hours of workstation time, the majority of which is spent "editing" the segmented data. This makes the technique somewhat impractical for frequent clinical use. Further advances in segmentation algorithms, perhaps incorporating a limited knowledge of abdominal anatomy into software, may decrease the amount of manual editing required and possibly enhance the reproducibility of the D reconstructions. CONCLUSION The opportunity to add a new surgical planning tool to the urologist's armamentarium is attractive, since a partial nephrectomy is an increasingly important surgical option in patients with small indeterminate renal masses, particularly when they are likely benign, and in patients with renal cancer who are candidates for renal-sparing surgery. In effect, spiral CT may provide a diagnostic test, a staging examination, and a surgical planning tool in one examination. Complete evaluation of the utility of this D imaging method as a surgical planning tool in patients needing partial nephrectomy will require additional patients and is under further study. Stuart G. Silverman, M.D. Department of Radiology Brigham and Women's Hospital Boston, Massachusetts REFERENCES 1. Licht MR, and Novick AC: Nephron sparing surgery for renal cell carcinoma. J Urol 19: 1-7, Carini M, Selli C, Barbanti G, Lapini A, Turini D, and Constantini A: Conservative surgical treatment of renal cell carcinoma: clinical experience and reappraisal of indications. J Urol 10: , Fishman EK, Magid D, Ney DR, Chaney EL, Pizer SM, Rosenman JG, Levin DN, Vannier MW, Kuhlman JE, and Robertson DD: Three-dimensional imaging. Radiology 181: 21-27, Fishman EK, Magid D, Ney DR, Drebin RA, and Kuhlman JE: Three-dimensional imaging and display of musculoskeletal anatomy. J Comput Assist Tomogr 12: 65-67, Kalender WA, Seissler W, Klotz E, and Vock P: Spiral volumetric CT with single-breath-hold technique, continuous transport, and continuous scanner rotation. Radiology 176: , Cline HE, Lorensen WE, Kikinis R, and Jolesz F: Threedimensional segmentation of MR images of the head using probability and connectivity. J Comput Assist Tomogr 1: , Cline HE, Lorensen WE, Ludke S, Crawford CR, and Teeter BC: Two algorithms for the three-dimensional reconstruction of tomograms. Med Phys 15: 20-27, Gilbert BR, Russo P, Zirinsky K, Kazam E, Fair WR, and Vaughan ED Jr: Intraoperative sonography: application in renal cell carcinoma. J Urol 19: , Assimos DG, Boyce H, Woodruff RD, Harrison LH, Mc- Cullough DL, and Kroovand RL: Intraoperative renal ultrasonography; a useful adjunct to partial nephrectomy. J Urol 16: , Rubin GD, Dake MD, Napel SA, McDonnell CH, and Jeffrey RB Jr: Three-dimensional spiral CT angiography of the abdomen: initial clinical experience. Radiology 186: , BrinkJA, Vannier MW, HeikenJP, Kalender WA, Yoffie RL, and Brunsden BS: Abdominal spiral CT: effect of interpolation algorithm, collimation, and zoom on effective section thickness and noise. Radiology 185(P): 126, UROLOGY / JANUARY 199 / VOLUME, NUMBER