| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Nuclear Medicine |
1 From the School of Medicine, University of Milano-Bicocca, Milan, Italy (S.S., C.M., B.Z., F.F.); Institute for Molecular Imaging and Physiology of the National Research Council of Italy, Milan (S.S., C.M., F.F.); and Departments of Nuclear Medicine (C.M., M.P., F.F.), Gynecology and Obstetrics (G.M., G.A., E.G., R.V.), Pathology (G.T.), and Radiology, University Vita-Salute (A.D.M.), Institute H S.Raffaele, Via Olgettina 60, 20132 Milan, Italy. Received November 7, 2003; revision requested January 15, 2004; revision received February 4; accepted March 2. Address correspondence to F.F. (e-mail: fazio.ferruccio@hsr.it).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Thirty-one women (mean age, 55.9 years) with ovarian carcinoma treated with primary cytoreductive surgery and followed up with platinum regimen chemotherapy were included. All 31 patients were scheduled for surgical second-look. Before surgical second-look, all patients underwent fluorodeoxyglucose (FDG) PET/CT. At PET/CT, three main categories of persistent disease were considered for data analysis: lymph nodal lesion, peritoneal lesion, and pelvic lesion. In all patients, imaging findings were compared with results of histologic examination after surgical second-look to determine the diagnostic accuracy of PET/CT in the evaluation of disease status. The 
statistic (Cohen ) was used for statistical analysis.
RESULTS: Seventeen (55%) of 31 patients had persistent tumor at histologic analysis after surgical second-look, and fourteen (45%) had no histologically proved tumor. The total number of lesions that was positive for tumor cells at histologic analysis was 41 (lymph nodes, n = 16; peritoneal lesions, n = 21; pelvic lesions, n = 4); maximum diameter of these lesions was 0.3–3.2 cm (mean, 1.7 cm). A correlation was found between PET/CT and histologic analysis ( = 0.48). The overall lesion-based sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of PET/CT were 78%, 75%, 77%, 89% and 57%, respectively. In the detection of a tumor, a size threshold could be set at 0.5 cm, as this was the largest diameter of a lesion missed at PET/CT.
CONCLUSION: Integrated PET/CT depicts persistent ovarian carcinoma with a high positive predictive value.
© RSNA, 2004
Index terms: Computed tomography (CT), comparative studies, 852.12111 • Dual-modality imaging, PET/CT, comparative studies • Positron emission tomography (PET), comparative studies, 852.12163 • Ovary, CT, 852.12111 • Ovary, neoplasms, 852.39 • Ovary, PET, 852.12163 • Ovary, radionuclide studies, 852.12163
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
A new imaging technique combining state-of-the-art PET and CT equipment (integrated PET/CT) has been introduced in clinical use. This PET/CT device acquires both PET and CT images, which are contemporaneous and coregistered by means of hardware arrangement, to localize elevated FDG uptake with improved anatomic specificity. Initial results in oncology with this combined anatomic and functional technique have been promising (35–39). Thus, the purpose of our study was to prospectively evaluate the accuracy of integrated PET/CT for depiction of persistent ovarian carcinoma after first-line treatment, with the use of histologic analysis as the standard of reference.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Integrated PET/CT
All imaging and data acquisition were performed with a combined PET/CT in-line system (CTI/CPS Reveal-HD; CTi PET Systems, Knoxville, Tenn) that was used to acquire CT and PET scans from the same patient in one session. A PET scanner (CTI PET HRT; CTi PET Systems) and a multi–detector row helical CT scanner (Somatom-Emotion Duo; Siemens Medical Systems, Erlangen, Germany) were integrated in this dedicated system. The axes of both systems were mechanically aligned so that shifting of the examination table by 60 cm moved the patient from the CT to the PET gantry. The resulting PET/CT images were coregistered on hardware.
Patients fasted for at least 6 hours before intravenous administration of 10 mCi (370 MBq) of FDG. In addition, all patients drank 500 mL of water during the FDG uptake period and were asked to empty their bladder before positioning for the PET/CT examination.
The combined examination started 45 minutes after the injection of FDG. CT data were acquired first. Unenhanced CT scans were obtained from the patient’s head to the pelvic floor with use of a standardized protocol (140 kV, 80 mA, tube rotation time of 0.5 seconds per revolution, pitch of 6, section thickness of 5 mm to match the section thickness of PET images, and acquisition time of 22 seconds). CT scans were acquired during shallow breathing. No oral contrast agent was administered. Immediately after CT, PET was performed and covered the identical transverse field of view. The acquisition time for PET was 4 minutes per table position. Six incremental table positions were used; thus, the entire PET examination lasted 24 minutes. PET images were acquired during quiet breathing.
Attenuation correction was performed by using CT scans. The CT pixel values, which were measured in Hounsfield units, were transformed into linear attenuation coefficients for the 511-KeV energy radiation. The image reconstruction matrix was 128 x 128 with a transverse field of view of 49.7 x 49.7 cm. The PET component of the scanner has an in-plane spatial resolution of 4.7 mm. In our series, 26 of 31 patients underwent PET/CT after they underwent conventional contrast-enhanced CT at follow-up. In these patients, only unenhanced CT scans were obtained at PET/CT, according to the previously reported protocol. In the remaining five patients, in whom contrast-enhanced CT was not performed prior to PET/CT, a dynamic contrast-enhanced CT examination of the abdomen was also performed after PET. All five patients received 120 mL of nonionic contrast material (Ultravist 300; Schering, Berlin, Germany), which was injected intravenously with a power injector at a rate of 3 mL/sec. Arterial and venous phase images were obtained at 40 and 130 seconds, respectively, after the start of injection.
Image Analysis
The images were prospectively analyzed by a review team whose members had no knowledge of the patients’ serum CA-125 levels or the results of other imaging examinations. The review team consisted of a radiologist (S.S.) with 15 years of experience with CT and a nuclear medicine physician (C.M.) with 15 years of experience with PET; these physicians interpreted the PET/CT images in consensus.
Image analysis was performed as follows: attenuation-corrected PET images, CT scans, and coregistered PET/CT images were displayed together on the monitor, and the PET/CT images were analyzed as a single set of images by both reviewers. The PET/CT results were reported, and the presence of normal and abnormal (ie, that which was suspicious for malignancy) FDG uptake was noted. When abnormal FDG uptake was present, its exact anatomic location was indicated on the CT scans. On PET/CT images, all organs and regions within the pelvis, including the uterus (if it was not previously removed), parametrium, cul-de-sac, pelvic side wall, and vagina, were evaluated. In the peritoneum, different sites were analyzed, including the (a) anterior part of the abdomen, (b) paracolic gutters, (c) subdiaphragmatic spaces, (d) mesentery, and (e) omenta. On PET images, the presence of a persistent tumor was suspected when accumulation of FDG was moderately to markedly increased, in comparison with that of comparable normal contralateral structures or surrounding tissues, excluding physiologic bowel and urinary activity.
Persistent tumor was diagnosed when the abnormal focal uptake observed on PET images corresponded to an abnormal soft-tissue mass in the pelvis and peritoneum observed on CT images. Lymph node sites were also evaluated and divided into the categories of abdominal nodes and pelvic nodes. The diagnosis of an abnormal lymph node on PET/CT images was based on the presence of focal increased FDG uptake on PET images at a location that corresponded to lymph node chains on CT scans. This method of image analysis for PET/CT images was derived from previous reports, in which this combined technique was used for tumor staging (32,39).
All data sets were analyzed at a workstation (Systrium Technologies, Minneapolis, Minn) that was capable of providing interactive multiplanar reformations and any appropriate window and level settings. Semiquantitative analysis was also performed with PET images by normalizing the amount of radiotracer uptake in any measurable lesion to the injected dose and patient body weight to obtain a standardized uptake value. According to previous reports (24,32), standardized uptake values greater than 3.0 were considered to be indicative of malignancy.
Surgical Procedure
Second-look surgical laparotomy was performed as described previously (7,8), with knowledge of the PET/CT results. For all patients, the procedure was performed by one surgeon (R.V.), who has more than 20 years of experience with oncologic gynecology surgery. At second-look surgery, in all patients the presence or absence of tumor tissue at 15 specific sites was recorded; these sites included the abdominal and pelvic lymph nodes, diaphragm, omentum, mesentery, gastric surface, splenic hilum, hepatic surface, serosa of the large and small bowels, paracolic gutter, peritoneum of the anterior abdomen and pelvis, rectal serosa, and vaginal stump. Such a list of specific sites was derived from the surgical gynecology literature (4,5). To better compare surgical and imaging findings, three main categories of persistent disease were considered for data analysis: lymph nodal lesion, peritoneal lesion, and pelvic lesion (7–9). The tumor masses, if present, were excised with debulking surgery, and blinded biopsies were performed at each site—even when no apparent gross mass was present. Only lesions that were examined histologically were considered for analysis. Second-look surgical findings were then classified as (a) negative; (b) macroscopically positive, when a biopsy of a suspicious area was positive at histologic analysis; or (c) microscopically positive, when a nonsuspicious area randomly sampled was histologically positive. Histologic findings were interpreted by a pathologist (G.T.), who has 14 years of experience with gynecologic oncology.
Statistical Analysis
Calculation of sensitivity, specificity, accuracy, and positive and negative predictive values for tumor detection with PET/CT was performed on the basis of a lesion-by-lesion analysis, in relation to results at histologic analysis after second-look surgery. For the purposes of statistical analysis, a true-positive lesion was a lesion seen on PET/CT images and found to be positive for tumor tissue at histologic analysis. A false-positive lesion was a lesion seen on PET/CT images but found to be negative for tumor tissue at histologic analysis. A true-negative lesion was indicated when no lesion was seen on PET/CT images and results of the histologic analysis were negative for tumor tissue; therefore, the number of true-negative lesions corresponded to the number of patients with no persistent disease both at imaging and at histologic analysis. A false-negative lesion was a lesion that was missed at image analysis but was found to be positive for neoplastic tissue at histologic analysis. Sensitivity, specificity, accuracy, and positive and negative predictive values of PET/CT were also calculated on the basis of per-patient analysis. statistics (Cohen ) were used as a measure of agreement between PET/CT and histologic findings. values of up to 0.40 were considered to indicate poor agreement; values between 0.41 and 0.75, moderate to good agreement; and values greater than 0.75, excellent agreement (40). Statistical analysis was performed with SAS software (version 8; SAS Institute, Cary, NC).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
PET/CT Imaging Findings
PET/CT was used to correctly identify 32 of the 41 lesions that were positive for tumor tissue at histologic analysis after the second-look procedure (13 of 16 lymph nodes, 18 of 21 peritoneal lesions, and one of four pelvic lesions) (Figs 1–3). All nine of the 41 lesions missed with PET/CT were equal to or smaller than 0.5 cm in maximum diameter (mean, 0.35 cm). Depiction of tumor lesions was therefore size dependent. As in our series, the largest diameter of a lesion missed at imaging was 0.5 cm; this was considered as a size threshold for the detection of tumor nodules with PET/CT. As for the location of persistent disease, detection at imaging of pathologically proved tumor lesions was good for those located in the lymph nodal chains (13 of 16) and in the peritoneum (18 of 21), but detection was limited for lesions in the pelvic region (one of four).
|
|
|
|
|
|
|
|
|
In total, 32 lesions were true-positive, nine were false-negative, 12 were true-negative, and four were false-positive.
CA-125 Measurements
Of the 17 patients with histologically proved persistent ovarian cancer after second-look procedure, CA-125 values were indicative of active disease (>35 U/mL) in 13 patients; in the other four, CA-125 values were normal (<35 U/mL). These 17 patients have been subsequently treated with second-line chemotherapy or surgical cytoreduction and then included in a strict follow-up protocol. Of the 14 patients with no histologically proved persistent ovarian cancer, CA-125 values were normal (<35 U/mL) in 12 and elevated in two. All 14 patients were included in a follow-up protocol; at present, 11 are disease free.
PET/CT Diagnostic Accuracy
The overall lesion-based sensitivity, specificity, accuracy, and positive and negative predictive values of PET/CT in the detection of persistent ovarian carcinoma were 78% (32 of 41 patients), 75% (12 of 16 patients), 77% (44 of 57 patients), 89% (32 of 36 patients), and 57% (12 of 21 patients), respectively (Table). The accuracy rate for detection of lesions larger than 1 cm in diameter was 90% (38 of 42 patients). The value was 0.48, which meant that in view of the limited number of cases included in the study, the strength of agreement between PET/CT results determined on a lesion-by-lesion basis and with histologic analysis could be considered moderately good. As in our series, almost all primary tumors (25 of 31) were papillary serous adenocarcinomas, and no conclusion could be drawn regarding the relative detectability of persistent tumor lesions of different histologic types. The overall patient-based sensitivity, specificity, accuracy, and positive and negative predictive values of PET/CT were as follows: 53% (nine of 17 patients), 86% (12 of 14 patients), 68% (21 of 31 patients), 82% (nine of 11 patients), and 60% (12 of 20 patients), respectively.
|
Metastases
In our series, PET/CT findings were correlated with histologic findings, which were used as the standard of reference. PET/CT is a whole-body technique that enables, in a single session, the detection of distant metastases. In this regard, in two of 31 patients, five distant metastases were found (one liver metastasis, one lung metastasis, and three retroclavicular lymph-nodal metastases). In these patients, the use of PET/CT to assess the presence of distant metastases modified the stage of disease and allowed appropriate changes in therapeutic care.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Serial measurement of tumor-related antigen CA-125 level is the most common method used to monitor clinical response after treatment. Its limited reliability, however, is well known, as elevation of the CA-125 level may indicate tumor persistence or recurrence, but a negative value does not provide absolute assurance of absence of disease; this is also shown by the results reported in the present study (14–16).
Progress in imaging techniques, particularly CT and magnetic resonance (MR) imaging, has contributed to improvements in the diagnosis and staging of ovarian cancer. Two recent reports of the Radiology Diagnostic Oncology Group (17,18) stated that CT and MR imaging are equally accurate and that either modality can be used in the staging of ovarian cancer. However, correctly assessing persistent or recurrent disease with only these conventional imaging modalities can be difficult. MR imaging was found to be a useful adjunct to the clinical examination to aid in the identification of patients with recurrent disease; however, MR imaging is limited in its ability to depict small calcified peritoneal implants, which are common in patients with serous carcinoma. In a work by Forstner et al (20), in which findings at MR imaging and surgery were correlated, MR imaging showed an accuracy of only 35% in lesions smaller than 2 cm; accuracy increased to 82% when lesion diameter was more than 2 cm, giving a mediocre overall accuracy rate of 59% for the evaluation of recurrent disease. Due to its widespread availability, CT is presently the most common noninvasive imaging modality used to monitor patients with ovarian cancer after first-line treatment (19). In this regard, however, the effectiveness of CT scanning, even with the use of state-of-the-art dynamic techniques and acquisition of thin contiguous sections, remains at least debatable. This is reflected by the wide variety of accuracy rates, which range from 38% to 88%, reported in the literature (11,17–19). This is mainly because CT proved to have low sensitivity for small lesions, especially those with a contrast enhancement pattern similar to that of adjacent normal structures. Furthermore, CT hardly allows reliable differentiation between persistent disease and postoperative changes. In this setting, the need for more sensitive and accurate imaging techniques is clear.
In contrast to CT and MR imaging, diagnoses with FDG PET are made on the basis of functional rather than morphologic criteria. In fact, PET with radiolabeled glucose analogue FDG is a method that is based on the increased glucose metabolism of malignant tumors (21). FDG PET has been shown to be effective in the identification of different primary and metastatic tumor types. It can reveal the biochemical differences between normal and malignant tissues, and it has been used as a functional method of determining tumor viability in several types of cancer, including those of the head and neck, colon, and lung (21). Previous works (22,23) suggested that FDG PET performed after treatment can depict viable residual or recurrent lesions otherwise missed or misinterpreted at conventional morphologic examinations, including CT. On the other hand, FDG PET has some limitations, as it does not provide precise anatomic information about the site of the lesion detected.
Presently, there are several reports in which the role of FDG PET in the evaluation of ovarian cancer response after primary surgery and chemotherapy is discussed (24–34). To our knowledge, the diagnosis of recurrent ovarian cancer with FDG PET was initially reported by Karlan et al (25) in a small series of 12 patients. FDG PET enabled a true-positive diagnosis to be made in six patients, and it showed no recurrence in the other six; in five of these patients, however, the second-look procedure demonstrated microscopic tumor deposits. Torizuka et al (31) recently assessed the value of FDG PET in the diagnosis of recurrent tumor in a series of 25 patients who had previously undergone surgery for ovarian cancer. FDG PET showed a sensitivity of 80% and an accuracy of 84%. In the same series, conventional imaging had lower sensitivity and accuracy rates (55% and 64%, respectively). The findings of Torizuka et al (31) confirmed that viable lesions in patients with treated ovarian cancer might become detectable because of metabolic changes before any morphologic correlate. Nakamoto et al (28) looked at the clinical value of FDG PET in 24 patients in whom ovarian tumor recurrence was suspected. In their work, FDG PET alone had a fairly good rate (79%) of diagnostic accuracy. Interestingly, by adding information of conventional imaging modalities comprising CT, the accuracy rate of FDG PET improved to 94%. Similar results were also obtained by Cho et al (30) and Picchio et al (34) in their series in which FDG PET/CT images were coregistered with software.
Recently, a new type of scanner, in which a full-ring-detector clinical PET scanner and a multi–detector row helical CT scanner are combined in one machine, has been introduced in clinical practice. Both scanners are aligned so that patients can undergo imaging in either of the two gantries by moving one system table. In this way, coverage of anatomically coregistered images from the head to the pelvic floor is obtained with hardware, rather than with postacquisition software. One of the advantages of integrated PET/CT over PET alone is the capability to localize foci of elevated radiotracer uptake with improved anatomic specificity. Initial results in oncology with this integrated imaging modality have been encouraging, and its paramount clinical potential in staging and monitoring the response to treatment in patients with cancer—including those with gynecologic tumors—has been recognized (35–39).
The results of the present study indicate that integrated PET/CT may be an effective means of detecting persistent ovarian carcinoma. However, the performance of PET/CT reported in this study is worse than that previously reported by other authors with the use of either PET alone or combined PET and CT. This discrepancy may be due to the fact that—different from many previous works, in which imaging findings were mainly correlated with clinical follow-up—in our protocol, all PET/CT findings were compared with histologic results after the second-look procedure, which was performed in all patients. Rose et al (27) studied the assessment of ovarian cancer recurrence with FDG PET alone; in their work, the study design and number of patients enrolled were similar to those in our study. It is interesting to note that when their data are compared with ours, an improvement in lesion detection is found with integrated PET/CT. This could be explained by the precise anatomic correlation of the radionuclide uptake provided by the integrated technique, which made possible a more reliable interpretation of PET imaging findings. In a recent preliminary study undertaken by Makhija et al (32), PET/CT and histologic findings were compared in six patients with suspected ovarian cancer recurrence. These authors concluded that PET/CT may be an effective means of depicting recurrent tumors, as it was used to correctly identify lesions in five of six patients. Although their results can hardly be compared with ours because of the limited number of patients and the retrospective nature of the study, the findings of Makhija et al (32) corroborate our findings.
The results of the present study also indicate that integrated PET/CT has a low negative predictive value and a high positive predictive value in the detection of residual neoplastic lesions after treatment. This is in line with previous results obtained with PET alone (29,31). Low negative predictive value seems to depend on the limited capability of PET/CT to depict microscopic or small-volume lesions, as all false-negative results in our series occurred in patients with microscopic disease. Such limitation of PET/CT, which is encountered with other imaging techniques (ie, MR imaging or CT), may make it difficult to identify patients with minimal tumor deposits, in whom assessment of disease may still require surgical second-look. On the other hand, integrated PET/CT showed high positive predictive value in revealing persistent disease. In our opinion, the high positive predictive value of this imaging modality may represent its real strength. In fact, it could be used to reliably identify patients with macroscopic disease who are candidates for salvage treatment, avoiding the morbidity and expense of invasive surgical assessment.
A disadvantage of our study was that the surgeon had knowledge of PET/CT results before the second-look procedure was performed. However, it is important to emphasize that at second-look surgery, the presence or absence of tumor tissue was assessed with systematic evaluation of specific anatomic sites, independently of positive or negative results at PET/CT imaging. Also, a certain limitation of the present study was that in most of the patients, unenhanced CT was used for integrated PET/CT imaging. However, as researchers argued previously (39), we could not justify the use of vascular contrast material in patients who were referred after they underwent conventional contrast-enhanced CT at follow-up. Thus, we are presently unable to establish the additional diagnostic value in integrated PET/CT due to the use of vascular contrast material.
In summary, this study has shown PET/CT to be a valuable diagnostic tool in the follow-up of patients with ovarian cancer after first-line treatment, mainly because PET/CT is able to depict macroscopic residual disease. If the results of forthcoming studies with larger series of patients confirm our results, integrated PET/CT may be proved to have a clear clinical effect on the therapeutic management of ovarian cancer. In fact, after primary cytoreductive surgery and chemotherapy, we believe a patient could be evaluated for persistence of disease with this imaging modality, and then physicians could proceed to use the most appropriate second-line treatment.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Authors stated no financial relationship to disclose.
Author contributions: Guarantor of integrity of entire study, S.S.; study concepts, C.M., S.S.; study design, B.Z.; literature research, M.P.; clinical studies, G.M., G.A.; data acquisition, G.T., M.P.; data analysis/interpretation, R.V., B.Z.; statistical analysis, B.Z.; manuscript preparation and definition of intellectual content, S.S.; manuscript editing, E.G., C.M.; manuscript revision/review, A.D.M., F.F.; manuscript final version approval, F.F.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. K. Schwarz, P. W. Grigsby, F. Dehdashti, and D. Delbeke The Role of 18F-FDG PET in Assessing Therapy Response in Cancer of the Cervix and Ovaries J. Nucl. Med., May 1, 2009; 50(Suppl_1): 64S - 73S. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Czernin, M. Allen-Auerbach, and H. R. Schelbert Improvements in Cancer Staging with PET/CT: Literature-Based Evidence as of September 2006 J. Nucl. Med., January 1, 2007; 48(1_suppl): 78S - 88S. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. K. von Schulthess, H. C. Steinert, and T. F. Hany Integrated PET/CT: Current Applications and Future Directions Radiology, February 1, 2006; 238(2): 405 - 422. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Pandit-Taskar Oncologic Imaging in Gynecologic Malignancies J. Nucl. Med., November 1, 2005; 46(11): 1842 - 1850. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. M. Blodgett, B. Casagranda, D. W. Townsend, and C. C. Meltzer Issues, Controversies, and Clinical Utility of Combined PET/CT Imaging: What Is the Interpreting Physician Facing? Am. J. Roentgenol., May 1, 2005; 184(5_supp): S138 - S145. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
