Estimating Splenic Volume: Sonographic Measurements Correlated with Helical CT Determination

Transcription

1 Ellen M. Yetter 1,2 Katrina B. Acosta 1 Mary C. Olson 1 Kenneth Blundell 3 Received February 12, 2003; accepted after revision June 10, Department of Radiology, Loyola University Medical Center, Maywood, IL Present address: Department of Radiology, St. Luke s Hospital of Kansas City, 4401 Wornall Rd., Kansas City, MO Address correspondence to E. M. Yetter (eyetter@saint-lukes.org). 3 School of Interdisciplinary Computing and Engineering, University of Missouri Kansas City, Kansas City, MO AJR 2003;181: X/03/ American Roentgen Ray Society Estimating Splenic Volume: Sonographic Measurements Correlated with Helical CT Determination OBJECTIVE. The purpose of this study was to determine which sonographic measurements of the spleen most closely correlate with splenic volume as determined on helical CT. MATERIALS AND METHODS. From October 17, 2000, to April 27, 2001, 142 consecutive patients prospectively underwent abdominal helical CT and sonography as part of an evaluation for liver disease. Calculations of splenic volumes were based on 10-mm unenhanced images. Maximum length (ML) and width (W), thickness (T), and craniocaudal length (CCL) were measured sonographically. Standard ellipsoid volume formulas (with the addition of new ellipsoid coefficients) and linear regression formulas were calculated for 117 patients whose examinations were performed within 30 days of each other. Mean percent differences, standard deviations, and 95% confidence intervals (CI) were calculated. RESULTS. We calculated the average difference between sonography- and CT-measured volume and the 95% CI for each of the four initial sonographic volume estimates with the ellipsoid method using two lengths and linear regression using two lengths and compared them to CT-determined volume. The ellipsoid formulas were then adjusted for bias. Linear regression formulas were derived in which splenic volumes were separately calculated on the basis of each of the two lengths. Mean percent differences and standard deviations for ellipsoid formulas with varying coefficients using the three length measurements were also calculated. CONCLUSION. Sonographic measurements allow accurate determination of splenic volume. Estimating splenic volume with the formula W T (ML + CCL) / 2 provides the greatest overall accuracy. S plenomegaly is a well-known manifestation of several diseases that may involve the liver, immune system, and hematopoietic system. Accurate noninvasive assessment of splenic volume is used in the clinical treatment of patients with these diseases. Previously described techniques for measuring the spleen have relied on nuclear scintigraphy [1 4], sonography [5 10], and CT [11 20]. CT is known to be a reliable and accurate method for assessing the volume of the spleen and other intraabdominal organs [11, 12, 17, 18 20]. To the best of our knowledge, no prior study has compared splenic volume determined using sonographic measurements with helical CT determined splenic volume as the gold standard. Because clinicians at our institution commonly request sonography to evaluate patients for clinically suspected splenomegaly and because sonography is a rapid and relatively inexpensive means of assessment with no radiation exposure, we undertook a reexamination of sonographically determined estimates of splenic volume. Two methods of obtaining longitudinal measurements were proposed, and the same measurements were applied to the two methods of calculation. The purpose of this study was to determine which formula, using sonographic measurements, most closely correlated with splenic volume estimated by helical CT. Materials and Methods From October 17, 2000, to April 27, 2001, the hepatologists at our institution referred 142 consecutive adult patients for abdominal helical CT and sonography as part of an evaluation for liver disease. One hundred twenty-three of the patients had both CT and sonography performed. At the initiation of data collection, we chose a 30-day interval between CT and sonographic studies on the premise that a significant change in the size of the spleen would be unlikely to occur during this time period. Six pa- AJR:181, December

2 Yetter et al. TABLE 1 Underlying Conditions of 117 Patients in Study of Sonographic Measurement of Splenic Volume Patients History No. % Hepatitis A B C Chronic Cirrhosis Alcoholic Amiodarone-induced Cryptogenic Primary biliary Hepatocellular carcinoma Primary sclerosing cholangitis Elevated levels of function tests Steatohepatitis Cholestasis Liver transplantation Amyloidosis Hepatic sarcoidosis History unknown Note. Some patients had more than one underlying condition. tients were eliminated from our study because the interval between CT and sonography was more than 30 days. The final study population consisted of 117 patients. This study was approved by our institutional review board. Retrospective review of the 117 sets of studies revealed that 84 (71.8%) were performed on the same day and that 21 (17.9%) were performed within 1 week of each other. Therefore, in 105 (89.7%) of the 117 studies, CT and sonography were performed within our chosen interval. The ages of the 66 men and 51 women in the study ranged from 32 to 85 years (mean age, 57 years). Clinical histories were obtained by reviewing the electronic patient record; these histories are shown in Table 1. Nine patients had histories of more than one condition: two patients had both hepatitis B and hepatitis C virus and hepatocellular carcinoma; one patient had hepatitis C virus and hepatocellular carcinoma; three patients had undergone liver transplantation and had a history of hepatitis C virus; one patient had a history of alcoholic cirrhosis and hepatitis B virus; one patient had a history of hepatitis B and C virus; and one patient had undergone a liver transplantation and had a history of alcoholic cirrhosis. CT was performed on either a CT/i or LightSpeed multidetector scanner (General Electric Medical Systems, Milwaukee, WI). The information was transferred to a workstation where the contour of the spleen was outlined on 10-mm unenhanced images obtained through the entire spleen by experienced technologists. Volume software on the workstation calculates the total volume by adding the volume estimated from the area to the thickness of the spleen on each image (Fig. 1). This technique has previously been described and was shown to be accurate in previous studies, with an error rate of less than 5% [11, 12, 18, 20]. Sonography was performed on Sequoia, Aspen, or XP-128 (Siemens-Acuson, Mountain View, CA) units. Measurements of the spleen were taken by a single author using electronic calipers. Maximum width and thickness were measured on the transverse images (Fig. 1). Width was measured as the greatest overall dimension, and thickness was measured as the shortest distance between the hilum and the outer convex surface of the spleen. We attempted to make these measurements as perpendicular to each other as possible. Maximum length and craniocaudal length were measured on the longitudinal images (Fig. 1). Maximum length was measured as the greatest overall dimension, and craniocaudal length was measured from the most superior margin to the most inferior margin of the spleen. Initially, two formulas were applied to each measurement set. The first formula is the conventional calculated volume of a prolate ellipsoid (0.524 W T L, in which W = maximum width, T = thickness, and L = length). This formula was used to estimate sonographic spleen volume twice for each patient, once using maximum length (ML) as L and another using craniocaudal length (CCL) as L. Two other estimates of sonographic volume were calculated using linear regression, again with L = ML and L = CCL. The response variable was CT-determined volume, and the explanatory variables were L, W, and T. Two linear regression formulas were generated, one using L = ML and the other using L = CCL. For each sonographic volume, a point estimate and 95% confidence interval (CI) were calculated for the average difference between sonographic volume and CT volume. Data were analyzed a second time, examining true and percent differences between the sonographic and CT volumes for each patient. Initially, four sonographic volumes were calculated for each patient; the ellipsoid formula and the linear regression formula were each calculated twice (once using ML and a second time using CCL as L). After further consideration, we recalculated the volumes using a new length measurement, average length (AVL) that we defined as (ML + CCL) / 2. A third analysis of the data was initiated by the hypothesis that a more accurate ellipsoid coefficient may exist for calculating sonographic splenic volume. A new ellipsoid coefficient was calculated for each of the three length measurements (i.e., ML, CCL, and AVL) by taking the average of CT volume / W T L (where L = ML, CCL, or AVL). Linear regression analysis was also applied to these calculations to generate three more coefficients. These six new coefficients were subsequently substituted into the previously calculated series of formulas. For each patient, the volume difference (delta volume between CT and sonographic measurements) and percent difference (delta volume / CT volume 100) were calculated for each method of measurement. Mean calculations and standard deviations (SDs) were subsequently calculated for each method. Mean volume differences and mean percent differences were also calculated for the subset of patients (n = 84) whose studies were performed on the same day using the conventional ellipsoid formula. Results The CT volumes, means and ranges of calculated volumes, and SDs using ellipsoid methodology for the total study population are listed in Table 2. The CT volumes, means and ranges of calculated volumes, and SDs using linear regression methodology are listed in Table 3. The average difference (between sonographic volume and CT volume) and 95% CI for each of the four initial sonographic volume estimates versus CT volume were calculated and are as follows: ellipsoid method using ML for L: 16.1 cm 3 (CI, 13.9 to 46.1); ellipsoid method using CCL for L: 61.8 cm 3 (CI, 90.0 to 33.6); linear regression using ML for L: 0 cm 3 (CI, 28.2 to 28.2); and linear regression using CCL for L: 0 cm 3 (CI, 30.2 to 30.2). The ellipsoid formulas were then adjusted for bias such that CT volume = (0.524 W T L) 16.1 when L = ML, and CT volume = (0.524 W T L) when L = CCL. Linear regression formulas were derived with each of the 1616 AJR:181, December 2003

3 Splenic Volume on Sonography and Helical CT A B C D E Fig year-old woman with abnormal results on liver function tests. A, CT scan depicts volume software calculation of splenic volume. Shaded area shows area of spleen estimated by computer on given slice. Volume at given level is calculated by CT software using sequential areas and slice thicknesses. Splenic volume for each level is displayed at top of image (arrow). B and C, Transverse sonograms illustrate method of determining width (B) and thickness (C) of spleen. Width is measured as greatest overall dimension (arrows, B). Obtained at same level as B, C shows thickness, defined as shortest distance between hilum and outer convex surface of spleen (arrows, C), being measured as perpendicular to width as possible. D and E, Longitudinal sonograms illustrate method of determining maximum (D) and craniocaudal (E) lengths. Maximum length is measured as greatest overall dimension (arrows, D). Obtained at same level as D, E shows craniocaudal length being measured from most superior margin to most inferior margin of spleen (arrows, E). TABLE 2 CT- and Sonographically Determined Splenic Volumes Using the Prolate Ellipsoid Method Volume (cm3) Methods and Formulas Range Mean SD , ± W T ML , ± W T CCL , ± W T AVL , ± W T ML , ± W T CCL , ± CT Sonography Conventional prolate ellipsoid New coefficient prolate ellipsoid W T AVL , ± W T ML , ± W T CCL , ± W T AVL , ± Note. SD = standard deviation, W = width, T = thickness, ML = maximum length, CCL = craniocaudal length, AVL = average length. AJR:181, December 2003 two length measurements used initially (i.e., ML and CCL). The derived linear regression formula was determined to be CT volume = (79. 3 L) + (19.4 W) + (15.1 T) using ML for L, and CT volume = (52.8 L) + (29.9 W)+ (32.8 T) using CCL for L. Mean volume differences, mean percent differences, and SDs for conventional ellipsoid formulas calculated using ML, CCL, and AVL for L and the ellipsoid formulas using the six newly generated coefficients are listed in Table 4. Mean volume differences and mean percent differences for the subset of patients whose studies were performed on the same day were calculated using the conventional ellipsoid formulas and are as follows: The mean volume difference and mean percent difference when ML = L was 5.3 cm3 and 10.6%, respectively (SD, ± 161.9). When CCL = L, mean volume difference and mean percent difference were 88.8 cm3 and 14.4% (SD, ± 150.4). When AVL = L, the mean volume difference was cal- 1617

4 Yetter et al. TABLE 3 Comparison of CT- and Sonographically Determined Splenic Volumes Using the Linear Regression Method Volume (cm 3 ) Method Range Mean SD CT , ± Sonograpy with linear regression Using ML = L to 1, ± Using CCL = L to 1, ± Note. SD = standard deviation, ML = maximum length, L = length, CCL = craniocaudal length. TABLE 4 Comparisons of Formulas for Sonographically Determined Splenic Volume Volume Means Formulas Difference (cm 3 Percent ) Difference (%) SD (cm 3 ) Conventional prolate ellipsoid W T ML ± W T CCL ± W T AVL ± New coefficient prolate ellipsoid W T ML ± W T CCL ± W T AVL ± W T ML ± W T CCL ± W T AVL ± Note. SD = standard deviation, W = width, T = thickness, ML = maximum length, CCL = craniocaudal length, AVL = average length. culated at 47 cm 3, and mean percent difference was calculated at 1.9% (SD, ± 153.2). Discussion Splenomegaly is an important clinical finding and can be present in a variety of diseases, including liver disease, portal hypertension, splenic vein thrombosis, lymphoma and other primary and metastatic neoplastic processes, and hematologic entities. Assessment of splenic size by physical examination is subjective and known to be inaccurate [1, 6, 7, 10]; therefore, evaluation with radiologic imaging is common. Sonography is a quick, simple, and relatively inexpensive modality that carries no risk of ionizing radiation. The portability of sonography makes it useful in imaging patients in the ICU who may be too unstable for transport to the imaging suite. At our institution, the hepatology and hematology services commonly request sonographic studies to evaluate patients for splenomegaly because these services are regional referral centers for liver disease and hematologic malignancies. After performing a number of these studies, we realized that we were basing our assessments of the size of the spleen on subjective parameters or a longitudinal length alone. This realization prompted us to establish a more objective means to confirm the size of the spleen using sonography. A number of studies have evaluated sonographic measurement of the spleen. Konus et al. [9] and Rosenberg et al. [10] determined normal sonographic dimensions in healthy children. Prassopoulos and Cavouras [13] correlated CT measurements of the spleen with age and vertebral body size in a pediatric population. Schlesinger et al. [11] described a simple way to measure the spleen in children by correlating linear CT measurements with CT volume. Loftus et al. [8], Downey [5], and Rodrigues et al. [6] evaluated sonographic assessment of the size of the spleen using several methods in cadaveric spleens. Ishibashi et al. [7] established a splenic index as the product of two linear sonographic measurements and correlated the index with volumes of resected spleens. Splenic index, as described by Lackner et al. [21], is the product of L W T as seen on CT [16]. Strijk et al. [14] addressed the role of CT in patients with Hodgkin s disease using a variation of CT splenic index with splenic weight measured in grams approximately corresponding to the product of 0.55 and the splenic index. Prassopoulos et al. [15] assessed splenic volume in adults using axial CT volumetric measurements as compared with linear CT measurements and generated a formula for estimating volume. In reviewing these studies, we noted a similarity in describing the volume or weight of the spleen as estimated by variations of a prolate ellipsoid volume formula. Based on geometry, the formula for the volume of an ellipsoid (i.e., a 3D ellipse such as the spleen or other solid organ) is 4 L W T π This formula is algebraically equivalent to 4 π L W T, 24 which is approximately equal to L W Τ. Downey [5] used the formula for the volume of an ellipsoid to approximate the volume of a cadaveric spleen. By linear regression analysis, he established splenic volume as being equal to 0.57 L W Τ, which is nearly identical to the formula for the volume of a prolate ellipsoid. Rodriques et al. [6] also used the formula 0.52 L W Τ to estimate the sonographic volume of the spleen; this formula is also identical to the formula approximating the ellipsoid volume. As mentioned previously, Strijk et al. [14] described the product of 0.55 and CT splenic index (i.e., L W Τ) to approximate splenic weight. This formula is also similar to that for the volume of a prolate ellipsoid. Given its established use in the literature, we therefore chose to use this formula as our basis for comparing different ways to measure the spleen using sonography. Linear regression was also used by Downey [5] and Rodrigues et al. [6]. We used linear regression to obtain prediction equations with CT volume as the response variable. T, W, and L were the explanatory variables. Two formulas with differing length variables were initially generated, using linear regression to 1618 AJR:181, December 2003

5 Splenic Volume on Sonography and Helical CT estimate splenic volume; one formula uses ML for L and the other, CCL. Although cumbersome to use in daily practice, these equations reflect the mathematic relationship between the sonographically measured L, W, and T values and the CT-determined volume. Mean volumes were obtained for each of the four initially applied methods of sonographic measurement (i.e., conventional prolate ellipsoid and initial linear regression using ML and CCL for L). The difference between each mean sonographic volume and CT volume was calculated and expressed using CI methodology. Our preliminary conclusion was that any of the four methods could be used to accurately calculate splenic volume, given the lack of significant variability in CIs. From a practical standpoint, however, the ellipsoid formula is less cumbersome and may be more convenient for everyday use. At this juncture, the ellipsoid formula using CCL for L was favored because it had a slightly narrower CI. Additional analysis and calculations were made in an effort to determine if a new method of measurement would be more accurate. A third length variable, AVL (i.e., ML + CCL) / 2), was used in ellipsoid formulas. Three new ellipsoid coefficients were calculated for each of the three formulas (differing in the L variable) using CT volume / L W Τ, and subsequent linear regression was used to obtain three more coefficients. Sonographic volumes were recalculated for each patient. A different method of data analysis was then performed. Instead of using CI methodology, volume difference (delta volume between CT and sonographic volumes) and volume percent difference (delta volume / CT volume 100) were made for each patient. Average percent difference expresses the volume difference in comparison to the CT volume such that difference expressed in a percentage more accurately reflects its significance over a wide range of volumes. As seen in Table 4, SDs are similar for all methods. The conventional ellipsoid method using AVL for L and the linear regression generated ellipsoid method using as the new coefficient and CCL for L have the lowest average percent difference (0.2% and 0.9% respectively) compared with CT volume. Although we did not apply all the calculations to the subset of patients whose studies were performed on the same day, we did apply the conventional ellipsoid formula three times for each patient (using ML, CCL, and AVL for L) to assess for any substantial Sonographic Volume Using W T AVL (cm ) 1,800 1,600 1,400 1,200 1, differences. The mean percent difference between CT volume and sonographic volume using AVL was only 1.9% (compared with 10.6% using ML and 14.4% using CCL). Of those formulas that express splenic volume based on a prolate spheroid, W T (ML + CCL) / 2 is considered the most accurate method of estimating splenic volume using sonography because it is associated with the smallest mean percentage difference. Our study has limitations. The study population did not consist of healthy patients or volunteers; all patients had either abnormal results on serum liver function studies, a history of liver disease, or suspicion of liver disease based on other clinical parameters. The study, however, was not designed to establish normal volumes because, clinically, deviations from the norm are more significant. The intent was to establish the best method of sonographic measurement. This population may explain the considerable range in the sizes of the spleen in our study (CT volumes of ,448.1 cm 3 ). Normal cadaveric splenic volumes reported by Loftus et al. [8] are cm 3 with a mean volume of 110 cm 3 and an SD of 70 cm 3. Henderson et al. [19] reported a normal splenic volume of 219 cm 3 as calculated from axial CT acquisitions. Clearly, most of the spleens in this study were larger than normal; only 30 of 117 spleens were smaller than or equal to the 300 cm 3 CT volume. At higher calculated volumes, scattering of the data increased, indicating that, for larger spleens, accuracy of any formula may be compromised (Fig. 2). Selective removal of outliers in the data was not undertaken in this study, but an improvement in the ,000 1,200 1,400 CT-Determined Splenic Volume (cm 3 ) Fig. 2. Scatterplot shows that calculated sonographic volumes for large-volume spleens have wider scatter than spleens with average volume. W = width, T = thickness, AVL = average length, y = x, R 2 = ,600 accuracy of any formula might result if some upper limit of volume is established. Another limitation of our study is that a single investigator performed the sonographic measurements of the spleen, creating the potential for suboptimal reproducibility and intraobserver and interobserver variability, a known limitation in visceral sonographic measurement, but one that was beyond the scope and intent of this article. Currently, we have limited the assessment of splenic volume in the manner described to those patients referred by the liver service for both CT and sonographic studies as part of their evaluation for liver disease. After becoming accustomed to receiving splenic volume information during the study data collection period, the hepatology service now routinely requests splenic volume information on all of their patients (an average of patients per week). We found that obtaining the additional measurements and performing the volume calculation adds only 5 10 min to the study time and anticipate using it in daily practice when sonographic studies are requested on patients with gastrointestinal, liver, or hematologic diseases or with suspected splenomegaly. In conclusion, many methods of measuring the spleen with sonography can result in accurate determination of volume. On the basis of the smallest mean percent difference between measured and calculated CT and sonographic splenic volumes, the conventional ellipsoid method using an average length measurement (0.524 W T (ML + CCL) / 2) is the best formula with which to estimate splenic volume using sonography. AJR:181, December

6 Yetter et al. References 1. Silverman S, DeNardo GL, Siegel E. Determination of spleen size by scintigraphy. Cancer Biother Radiopharm 1999;14: Strauss LG, Clorius JH, Frank T, van Kaick G. Single photon emission computerized tomography (SPECT) for estimates of liver and spleen volume. J Nucl Med 1984;25: Zhang B, Lewis SM. Use of radionuclide scanning to estimate size of spleen in vivo. J Clin Pathol 1987;40: Markisz JA, Treves ST, Davis RT. Normal hepatic and splenic size in children: scintigraphic determination. Pediatr Radiol 1987;17: Downey MT. Estimation of splenic weight from ultrasonographic measurements. Can Assoc Radiol J 1992;43: Rodrigues AJ, Rodrigues CJ, Germano MA, Rasera I, Cerri GG. Sonographic assessment of normal spleen volume. Clin Anat 1995;8: Ishibashi H, Higuchi N, Shimamura R, Hirata Y, Kudo J, Niho Y. Sonographic assessment and grading of spleen size. J Clin Ultrasound 1991;19: Loftus WK, Chow LTC, Metreweli C. Sonographic measurement of splenic length: correlation with measurement at autopsy. J Clin Ultrasound 1999; 27: Konus OL, Ozdemir A, Akkaya A, Erbas G, Celik H, Isik S. Normal liver, spleen, and kidney dimensions in neonates, infants, and children: evaluation with sonography. AJR 1998;171: Rosenberg HK, Markowitz RI, Kolberg H, Park C, Hubbard A, Bellah RD. Normal splenic size in infants and children: sonographic measurements. AJR 1991;157: Schlesinger AE, Hildebolt CF, Siegel M, Pilgrim TK. Splenic volume in children: simplified estimation at CT. Radiology 1994;193: Breiman RS, Beck JW, Korobkin M, et al. Volume determinations using computed tomography. AJR 1982;138: Prassopoulos P, Cavouras D. CT assessment of normal splenic size in children. Acta Radiol 1994;35: Strijk SP, Wagener DJT, Bogman MJ, de Pauw BE, Wobbes T. The spleen in Hodgkin disease: diagnostic value of CT. Radiology 1985;154: Prassopoulos P, Daskalogiannaki M, Raissaki M, Hatjidakis A, Gourtsoyiannis N. Determination of normal splenic volume on computed tomography in relation to age, gender and body habitus. Eur Radiol 1997;7: Lee JKT, Sagel SS, Stanley RJ, Heiken JP. Computed body tomography with MRI correlation, 3rd ed. Philadelphia: Lippincott-Raven, 1998: Sandrasegaran K, Kwo PW, DiGirolamo D, Stockberger SM, Cummings OW, Kopecky KK. Measurement of liver volume using spiral CT and the curved line and cubic spline algorithms: reproducibility and interobserver variation. Abdom Imaging 1999;24: Schiano TD, Bodian C, Schwartz ME, Glajchen N, Min AD. Accuracy and significance of computed tomographic scan assessment of hepatic volume in patients undergoing liver transplantation. Transplantation 2000;69: Henderson JM, Heymsfield SB, Horowitz J, Kutner MH. Measurement of liver and spleen volume by computed tomography. Radiology 1981;141: Cools L, Osteaux M, Divano L, Jeanmart L. Prediction of splenic volume by a simple CT measurement: a statistical study. J Comput Assist Tomogr 1983;7: Lackner K, Brecht G, Janson R, Scherholz K, Lutzeler A, Thurn P. The value of computer tomography in the staging of primary lymph node neoplasms (author s transl) [in German with English abstract]. Rofo Fortschr Geb Rontgenstr Nuklearmed 1980; 132: AJR:181, December 2003