images. Good separation of the water and fat resonances was obtained, allowing for excellent suppression of the fat signal. C o n c l u s i o n s : Although the SI method used was time consuming, fast spectroscopic imaging methods are available and are currently used in this laboratory to produce high quality images. These methods could be incorporated into conventional MRI sequences to increase sensitivity and specificity for diagnosis of breast cancer. L e a r n i n g Objectives: 1. Spectroscopic MR imaging produces images with decreased artifacts due to susceptibility gradients and superior contrast-to-noise ratio and edge definition. 2. SI may improve sensitivity and specificity for diagnosis of breast cancer.
3-13 Smooth ROC Curve Estimation from Continuous Data: Statistical M e t h o d s for Analyzing t h e Predictive V a l u e of Spiral C T of Ureteral Stones Kelly H. Zou, PhD, Brigham & Women's Hospital, Boston, MA Clare M. Tempany, MD, Julia R. Fielding, MD4St~art G. Silverman, MD, Salim Samuel, MD, Kevin R. ~ l g ) ~ i ~ Purpose: D i a g n o s t ~ c w ~ t ~ l ~ i ~ - a ~ C I T s c a n s in patients with obstructing u r e ' ~ l ~ a l ~ require the analysis of continuous test data (e.g., st6rfe sizes). The accuracy of a test can be summarized using an ROC curve. The authors develop and compare three methods for constructing a smooth ROC curve from continuous diagnostic data. Methods: Nonparametric (N), semiparametric (S), and parametric (P) smooth ROC curve analyses were applied to data from 100 unenhanced spiral CT scans of patients with proven obstructing ureterm stones. The accuracy of stone size for predicting the need for intervention was evaluated using these methods. Characteristics and summary measures of the resulting ROC curves were estimated. Results: All methods fitted the data well. Method N followed the details of the empirical data. Methods S and P yielded similar estimates of the ROC curve parameters. Areas under the ROC curves were 0.807, 0.821 and 0.814 by methods N, S and P, in comparison with 0.811 by the empirical method. C o n c l u s i o n s : Method P is preferred for constructing a smooth ROC curve using the available data derived from spiral CT of stone size. The analyses confirm the predictive value of spiral CT of obstructing ureteral stones for intervention. L e a r n i n g Objectives: 1. Develop statistical methods for assessing imaging accuracy in diagnostic radiology. 2. Demonstrate the preferred method for estimating smooth ROC curves. 3. Show h o w these methods are applied in analyzing the predictive value of spiral CT scans in ureteral stone disease.
3-14 CT Diagnosis of M e s e n t e r i c Panniculitis Vahid Yaghmai, MD, Evanston Hospital, Evanston, IL Richard M. Gore, MD, Gary G. Ghahremani, MD P u r p o s e : Mesenteric panniculitis is characterized by infiltration of the mesenteric fat by lipid laden macrophages leading to a tumor-like mass. Abdominal pain, epigastric tenderness and fever are unusual manifestations of this idiopathic condition. We reviewed the CT manifestations of this entity to establish guidelines for diagnosis. Material a n d M e t h o d s : We retrospectively reviewed medical records of our patients with CT findings of mesenteric panniculitis (N = 16). Six patients (38%) had abdominal pain and tenderness as the presenting symptom. However, ten patients (62%) had neither pain or a palpable mass and had undergone CT for other reasons. The subset of patients with abdominal pain had a mean age of 68 years and consisted of 4 men and 2 women. Results: Clinical features were localized abdominal pain and tenderness associated with mild fever. CT findings included well-defined soft tissue infiltration of mesenteric fat (100%) often with engorged mesenteric vasculature. Mass effect on the small bowel loops without invasion was present in all patients.
C o n c l u s i o n : Mesenteric pannicuiitis is an u n c o m m o n cause of abdominal pain and tenderness. Our illustrative cases highlight the subtle but important CT features of this rare source of abdominal pain. L e a r n i n g Objectives: CT manifestations of mesenteric panniculitis, an unusual cause of abdominal pain and fever, are discussed.
3-15 Reproducibility of M e a s u r e m e n t of Small Lymph Nodes on Smaller and Larger Imaging Format David Rochester, MD, Evanston Hospital-Northwestern University, Evanston, IL, Vahid Yaghmai, MD, Joseph S. Panella, MD, Judy Huang, MD, John R. Iraj, MD P u r p o s e : To assess the reproducibility of measurement of small lymph nodes imaged using 12 on 1 and 20 on 1 formats. Materials a n d Methods: Fifty lymph nodes in patients with lymphoma estimated to measure between 1 and 2 cm were filmed on both 12 on 1 and 20 on 1 formats each on a single x-ray film. These were measured by four radiologists. Each of the two groups of lymph nodes were measured twice. Using a paired t-test for p-values, the reproducibility of measurement by the same radiologist on the two measurements at both 12 on 1 and 20 on 1 was assessed. Using a one way analysis of variance the consistency of measurement of the four radiologists within each format was assessed. Results: Differences in measurement between the four radiologists were not significant at the 12 on 1 format. At 20 on 1, however, differences were significant (p-values < 0.001). There was no statistically significant difference in measurements by the same radiologist within the 12 on 1 and 20 on 1 formats w h e n comparing the first and second measurement. C o n c l u s i o n s : Variation among radiologists is significant on the 20 on 1 format. Individual variation is not significant on this format. L e a r n i n g Objectives: Differences between radiologists in measurement of small objects are more significant on smaller formulated images; however, these differences are not significant w h e n the same radiologists measure objects on these smaller formatted images. When comparing small objects on smaller formatted images, it is particularly important that the same person measures both objects.
3-16 T h e "Sabre" IVC.-A Study of t h e Inferior V e n a C a v a in Fatty C h a n g e of t h e Liver Ashley Davidoff, MD, Joseph Makris, MS, Luis E. Diaz, MD, University o f Massachusetts Medical Center, Worcester, MA, Jay M. Colby, MD, Thomas M. Cummings, MD, Susan A. Afonso, MD The effect of fatty change of the hepatic venous radicles is a well k n o w n diagnostic feature of fatty change of the liver. Because of cell swelling, the increased intrahepatic pressure sometimes supersedes the systemic venous pressure resulting in compression of the veins. This is particularly well seen at the confluence of the hepatic veins near the IVC. We have observed similar shape changes in the intrahepatic portion of the IVC. In health, the medial and posterior aspects of the intrahepatic inferior vena cava are free of surrounding liver tissue. However with enlargement of the liver, and more specificaliy, in diffuse fatty change, borders of the inferior vena cava become surrounded by liver tissue. This results in shape changes of the IVC reminiscent of the "sabre" trachea of COPD. We explored thirty cases of diffuse fatty change of the liver using computed tomography and found that shape changes of the inferior vena cava were characteristic of this disease. The A/P diameter of the IVC became greater than the transverse dimension resulting a "sabre" IVC. We have observed this change in other liver diseases however. C o n c l u s i o n : 1. The inferior vena cava may have characteristic shape changes in diffuse severe fatty change of the liver. 2. This finding is most commonly seen in fatty change but may be seen in other liver diseases. 3. The "sabre" IVC has important implication in MRI of the
745