Objectives: Before liver transplant, it is necessary to know the size of the organ in advance of the procedure. We studied the correlation between liver volumetric computed tomography results and liver weight.
Materials and Methods: Postmortem volumetric computed tomography was conducted on cadavers before autopsy, and 3-dimensional liver volume was estimated with semiautomated software. Liver weight was then determined at autopsy. Linear regression and univariate analysis of variance results were used to determine the accuracy of volumetric 3-dimensional computed tomography in estimating liver weight. We also used 2-dimensional liver sizes to design a 2-dimensional formula to estimate liver volume.
Results: We found that 3-dimensional volumetric computed tomography was able to accurately estimate liver weight (standard error = 157 g) with a liver density of 0.99 g/mL. Intraobserver and interobserver variabilities were small. The 2-dimensional formula estimated liver weight slightly less accurately (standard error = 212 g).
Conclusions: We conclude that liver weight can be estimated accurately with 3-dimensional volumetric computed tomography; estimates were more precise than with the 2-dimensional formula-based liver volume estimation. Volumetric computed tomography can be an important tool during preoperative workup before transplant surgery.
Key words : Density, Formula, Volume
Liver transplant requires a precise workup of both the donor’s and recipient’s liver, including evaluation of liver anatomy, parenchyma, vascular supply, and volume. The preoperative assessment of liver volume can be crucial for transplant success. The recipient needs a future liver remnant volume of at least about 30% of the standard liver size.1,2 In addition, a liver size of about 30% must be preserved in a healthy living donor for survival. Several studies have investigated the estimation of liver size using mathematical formulas to estimate liver volume.3-6 However, these formulas were based on the patient’s body surface area, body weight, body length, sex, age, and/or race. Standard errors (SE) in the range of 275 to 328 mL were found. Because liver size and shape vary widely between individuals, it may be difficult, if not impossible, to apply estimation formulas with parameters that are based on the body versus the liver.
State-of-the-art transplant surgery necessitates imaging of the liver and its vascular supply, which is usually performed with contrast-enhanced computed tomography (CT) or magnetic resonance imaging. Therefore, most donors and recipients of liver transplant have a preoperative CT or magnetic resonance imaging scan. During the past decade, hepatic volumetry and 3-dimensional imaging techniques with CT have been introduced to increase the precision and safety of liver surgery.7 Furthermore, it is now possible to perform a preoperative virtual liver resection to aid in planning the surgical procedure.8 A volumetric CT scan can be used to estimate precisely liver volume before transplant.9 There are several software programs commercially available to aid in volumetric analyses, which include manual, semiautomated (interactive), or automated segmentation analyses. Manual segmentation is labor intensive and time consuming, whereas automated segmentation may provide overestimates by including more tissue than just the liver (for example, stomach wall, ascites, or gall bladder). However, semiautomated programs allow the possibility of manual correction and are less time consuming than manual segmentation.
In this study, we calculated liver volumes from CT scans using a semiautomated program and then compared the calculated volumes with weights of livers obtained in a postmortem setting. In addition, with 2-dimensional sizes of the liver measured on CT images, we constructed a 2-dimensional formula to estimate the liver volume and compared this to the 3-dimensional liver volumetric CT scan results.
Materials and Methods
Volumetric computed tomography
In this single-center study, we included deceased adults who had received a postmortem total body CT scan before autopsy between August 2012 and February 2014 at our hospital. With the use of a 16-slice CT scanner (Somatom Sensation 16, Siemens Medical Solution AG, Erlangen, Germany), thorax-abdomen-pelvic images were acquired from 2 cm above the clavicles to the trochanter minor. Scan parameters were 120 kV, per the Automatic Exposure Control (Care Dose 4D, Siemens Medical Solution), with rotation time of 1 second. Contiguous 5-mm coronal and axial images were reconstructed with a B30 convolution filter in a standard soft tissue setting (window level 400, width 40). We estimated liver volume from CT images using the semiautomated software Pinnacle (version 8.0d, Royal Philips Healthcare, Amsterdam, The Netherlands). One researcher (LS) outlined the liver contour manually in 5-mm axial images every 10 mm (Figure 1). The contours in the intermediate slices were obtained by automated linear interpolation and were subsequently inspected and corrected manually if necessary. Liver volume and SE results was then calculated with the use of Pinnacle software. For 10 random cadavers, liver volumetric CT was performed twice by the first researcher (LS) and once by a second researcher (JH) to determine intra- and interobserver variabilities. Both researchers were blinded for liver weight.
The maximum craniocaudal height, left-right width, and anteroposterior depth of the livers were measured at coronal and axial CT images, which comprised the 2-dimensional liver measurements (Figure 1). These 2-dimensional measurements were used to design a 2-dimensional formula to estimate liver volume.
Liver weight and clinical evaluation
Upon arrival at our mortuary, cadavers were preserved at 4°C until autopsy. Liver weights were measured during autopsy on a weight scale (type DE12K1N, Kern & Sohn, Balingen, Germany). Patient data (sex, body mass index, date of birth, date of death, time of death, time of CT scan, time of autopsy, cause of death, and presence of liver disorders) were obtained from the electronic patient medical records. In addition, clinical and autopsy medical records were reviewed for the presence of liver steatosis and cirrhosis and for indications of heart failure. These data were used to test for differences in liver volume and liver density as a result of presence or absence of liver steatosis, liver cirrhosis, heart failure, and overweight status and for differences as a result of age and sex. A cardiologist with 10 years of experience interpreted the data to establish the presence of heart failure according to the guidelines of the European Society of Cardiology.10 Patients were classified as overweight if body mass index was ≥ 25 kg/m2.
A univariate generalized linear regression analysis evaluated the correlation between 3-dimensional liver volume determined by CT scan and liver weight from autopsy. Univariate analysis of variance determined the liver density (slope of the best-fit regression line through the origin) and the accuracy to estimate liver weight with 3-dimensional volumetric CT analyses. The intra- and interobserver variabilities were assessed by Bland-Altman plots and intraclass correlation coefficients using the 2-way mixed method.11
With the use of a simple linear regression, a 2-dimensional formula to estimate liver volume was designed using the 2-dimensional liver measurements. Again, univariate linear regression was used to determine the correlation between 2-dimensional liver volume and liver weight. Univariate analysis of variance determined the liver density and the accuracy to estimate liver weight using the formula. Differences in liver density and liver volume for liver steatosis, cirrhosis, heart failure, overweight status, age (< 50 or ≥ 50 y), and sex were tested using univariate ANOVA analysis and independent sample t test. Furthermore, a Spearman correlation coefficient evaluated whether the time interval between imaging and autopsy affected the results.
Our study included the results of 39 human cadavers who had received both a postmortem CT examination and an autopsy in our hospital (23 men, 16 women; mean age [standard deviation (SD)] of 64.3 years (15.1 y) (Table 1). The mean (SD) postmortem interval until imaging and until autopsy was 15.8 hours (9.8 h) and 25.6 hours (17.7 h). The mean (SD) liver volume on CT scan was 1785 mL (515 mL). The mean (SD) liver weight at autopsy was 1779 grams (501 g). Liver volume measured by volumetric CT had an excellent correlation (Pearson r = 0.95, P = .00) with liver weight at autopsy at a density of 0.99 g/mL (95% confidence interval [CI], 0.96-1.02) (Figure 2). With the use of this density, liver weight could be estimated accurately (SE = 157 g) based on its volume at CT.
In the intraobserver agreement, the mean difference in liver volume measured by the first researcher was -0.9 mL (limits of agreement, -36.0 to 34.2 mL). The interobserver reliability showed a mean difference of -31.6 mL (limits of agreement,-98.2 to 35.0 mL). Intraclass correlation coefficients of 0.998 (95% CI, 0.993-1.000; P = .00) and 0.994 (95% CI, 0.975-0.998; P = .00) were found, indicating good agreement. These are the proportions of the variance due to subject-to-subject variability in error-free scores.
Liver measurements from CT scan showed mean (SD) height of 168.9 mm (25.3 mm), mean (SD) width of 201.1 mm (24.0 mm), and mean (SD) depth of 180.6 mm (26.2 mm). From the 2-dimensional liver size results, the following formula was designed to estimate liver volume: Liver volume (in mL) = [(height in mm)(width in mm)(depth in mm)]/3850.
A Pearson correlation coefficient of r = 0.90 (P = .00) was found between liver volume estimated by this 2-dimensional formula and liver weight at autopsy (Figure 2). Univariate ANOVA showed an accuracy of SE = 212 grams when liver weight was estimated using this formula, with density of best fit of 1.09 g/mL (95% confidence interval, 1.05-1.13 g/mL) (Figure 2).
According to the European Society of Cardiology guidelines,10 we identified 9 patients with heart failure, 11 with liver steatosis, and 1 with liver cirrhosis. Thirty patients were considered to be overweight, and 6 were under the age of 50 years. Univariate analysis of variance showed no differences in liver density between the subgroups (P = .68 for sex; P = .17 for age; P = .94 for overweight status; P = .22 for presence of heart failure; and P = .40 for presence of steatosis). Men (P = .09) and patients with heart failure (P = .06) were likely to have significantly larger liver volumes. No differences in liver volume were found between the age subgroups (P = .81), between those who were or were not overweight (P = .49), and between those who had or did not have liver steatosis (P = .19). Cirrhosis was diagnosed in only 1 of the patients and therefore could not be tested (Table 2).
The interval between CT imaging and autopsy (median interval of 1.6 h, p25 = 1.3 h, and p75 = 15.5 h) had no significant effect on the results. The mean difference (SD) of 10.2 grams (156.3 g) between autopsy liver weight and calculated liver weight from volumetric CT analysis (mean [SD] of 10.2 [156.3] g) did not change when the interval increased (Spearman correlation coefficient = 0.02; P = .92). In addition, no correlation was found between calculated liver density and postmortem interval between imaging and autopsy (Spearman correlation coefficient = -0.02; P = .90).
This study demonstrates that liver 3-dimensional volumetric computed tomography had good results in estimating liver weight, with an SE = 157 grams and a liver density of 0.99 g/mL. Volumetric CT is a reliable method and seems appropriate for use during preoperative planning of liver transplant. The 2-dimensional formula, based on 2-dimensional liver measurements, also showed good correlation to liver weight; however, it was less accurate than the 3-dimensional volumetric CT, with SE = 207 grams and a relative overestimation of the density of 1.09 g/mL, corresponding to a slight underestimation of liver volume.
Several other noninvasive methods of estimating liver volume have been proposed.9 The 2-dimensional liver span is traditionally used as measurement for liver size at ultrasonography and physical examination. However, considering the complexity of liver shape, liver span alone cannot appropriately represent liver volume or mass.12,13 Despite liver volume analyses by 2-dimensional ultrasonography being based on geometrical assumptions, significant errors still occur.14 Furthermore, several formulas with parameters based on body sizes have been proposed.3,4,6 These studies used linear or nonlinear models to estimate liver volume based on body surface area, body weight, and body height. The presented formulas are different for different races. Our study is based on direct CT liver measurements and therefore avoids all factors that may influence liver size such as race, age, and sex. Accordingly, this seems the more appropriate method of preoperative liver size estimation. All of the previous studies were postmortem studies with long postmortem intervals up to 7 days or even with formalin-preserved livers, which could have influenced the liver density as formalin density is about 1.09 g/mL at room temperature.15 In addition, the liver parenchyma changes during decomposition. In our study, the postmortem interval until CT imaging and autopsy was short, with means of 15.8 and 25.6 hours. These are short periods from the perspective of postmortem studies. However, from the point of view of liver transplant, with either living donors or deceased donors with short postmortem intervals, the postmortem interval in our study is still a limitation for comparisons to the transplant population. The livers were weighed directly after they were retrieved by the autopsy pathologist; there was no use of formalin or other preservation material. However, the density may still differ after this short postmortem interval from the density shown in a living donor situation.
Volumetric liver CT analysis showed a good correlation with liver weight, although the possibility of slight overestimation or underestimation exists. Overestimation can be explained by presence of perihepatic ascites on several CT scans, which made liver contours difficult to determine. Perihepatic ascites can be a normal finding on a postmortem CT. Perihepatic ascites also can occur in many patients with liver disease who need transplant, although this condition is not found in healthy living donors. Semiautomated segmentation is useful to manually correct the automated inclusion of ascites in liver volume. Furthermore, overestimation can occur when the remaining intrahepatic blood volume is included in the CT image, whereas this is lost during retrieval at autopsy and liver weight assessment. However, this difference in liver blood volume at volumetry and when the liver is weighed is smaller in the postmortem setting than during liver retrieval from a living donor, as there is no intact blood flow and after death the fluids sink out of the organs and into the dorsal parts of the body (known as livor mortis).16 Underestimation of liver volume on CT could be caused by differences in delineation of the included tissue at the liver hilus on CT versus that shown at autopsy.
In our small study group, no differences in liver volume could be found between the subgroups of male versus female sex, age, overweight status, or presence of heart failure or liver steatosis (Table 3). Larger study groups may show significant volume differences between male and female subgroups.
The excellent correlation (r = 0.950) between CT liver volume and liver weight at autopsy was found at a density of 0.99 g/mL. Jackowski and associates17 found a strong correlation between estimated and autopsy liver weight at a density of 1.05 g/mL. Overestimation of liver volume will give underestimation of the density and vice versa. However, because of the shorter mean postmortem interval until imaging (15.8 h) versus that shown in the study of Jackowski and associates (33.5 h), postmortem volume changes will have occurred to a lesser extent at the moment of imaging in our study. The mean interval between imaging and autopsy was 9.8 hours in our study (median 1.6 h) and approximately 12 hours in the study by Jackowski and associates. Maximum lividity takes place during the first day after death; therefore, volume changes could be more prominent in the interval between imaging and autopsy. As a consequence, liver volume at CT could slightly be overestimated compared with the moment of liver weighing at autopsy, resulting in a lower calculated density. We tested whether liver measurements at CT and autopsy were influenced by the time interval between imaging and autopsy and found no correlation. Because the postmortem interval is even shorter during living-donor or deceased-donor transplant situations, it is very unlikely that this time period will cause clinical significant changes in liver volume.
Liver densities between 0.86 and 1.13 g/mL were found in the literature.18,19 It should be noted that a density of 1.13 g/mL is found in a normal liver, whereas pathologic livers with edema or severe fatty degeneration had significantly lower densities. Overall, the density is affected by factors such as time of death and condition of the preserved cadaver such as temperature.18 In our study population, we also tested for differences in liver density between several subgroups (Table 2) but could not find any significant differences. Further study in larger groups could be of interest.
Patients with heart failure were expected to have a higher liver density because of hepatic congestion. When prolonged, hepatic congestion can lead to fibrosis (cardiac cirrhosis), which also increases liver density.20 The assessment of whether a patient had heart failure was not straightforward. An experienced cardiologist made the assumptions based on the available data, according to the European Society of Cardiology guidelines.10 Steatotic livers were expected to have a lower density because the mass of fat is about 0.9 g/mL.18,19 However, the steatotic livers in our study showed no significant difference in density compared to nonsteatotic livers. Because it is about subtle differences, the number of livers included in this study is probably too small to demonstrate this heterogeneity.
The 2-dimensional formula for liver volume, based on 2-dimensional measurements of length, width, and depth, is a simple formula that can be used quickly when assessing the results of a liver CT scan. It is less accurate than the 3-dimensional volumetric liver CT scan results but still quite accurate at estimating the liver volume, with SE of 212 grams. The 3-dimensional volumetric liver CT has the disadvantage of being rather labor intensive, and we needed approximately 30 minutes per case, compared with 2 minutes for the 2-dimensional measurements and formula.
We conclude that volumetric liver CT with semiautomated software has an excellent correlation to liver weight, with a liver density of 0.99 g/mL. Therefore, liver volumetry can be a useful tool for the preoperative planning of a liver transplant. The 2-dimensional formula, based on 2-dimensional liver sizes, was slightly less accurate but quick in use.
Volume : 14
Issue : 1
Pages : 72 - 78
DOI : 10.6002/ect.2015.0142
From the 1Department of Radiology and Nuclear Medicine, Radboud
University Medical Centre, Nijmegen, The Netherlands; the 2Faculty of
Medicine, Radboud University, Nijmegen, The Netherlands; and the 3Department
of Radiation Oncology, Radboud University Medical Centre, Nijmegen, The
Acknowledgements: The authors declare that they have no sources of funding for this study, and they have no conflicts of interest to declare. We thank Dr M van der Vlugt (cardiologist) for excellent help in heart failure assessment.
Corresponding author: Willemijn M. Klein, Department of Radiology and Nuclear Medicine, Radboud University Medical Centre, PO Box 9101, 6500 HB Nijmegen, The Netherlands
Phone: +31 24 3614011
Figure 1. Computed Tomography Liver Volume Measurements
Figure 2. Correlation Between Computed Tomography Liver Volume and Liver Weight at Autopsy
Table 1. Characteristics of the 39 Patients, Including Baseline Patient Data and Liver Data
Table 2. Liver Densities for Sex, Age, Weight, Heart Failure, Liver Steatosis, and Liver Cirrhosis Subgroups
Table 3. Liver Volumes for Sex, Age, Weight, Heart Failure, Liver Steatosis, and Liver Cirrhosis Subgroups