Original Research
Gastrointestinal Imaging
June 2011

Decreased Detection of Hypovascular Liver Tumors With MDCT in Obese Patients: A Phantom Study

Abstract

OBJECTIVE. The purpose of this article is to assess the impact of large patient size on the detection of hypovascular liver tumors with MDCT and the effect of a noise filter on image quality and lesion detection in obese patients.
MATERIALS AND METHODS. A liver phantom with 45 hypovascular tumors (diameters of 5, 10, and 15 mm) was placed into two water containers mimicking intermediate and large patients. The containers were scanned with a 64-MDCT scanner. The CT dataset from the large phantom was postprocessed using a noise filter. The image noise was measured and the contrast-to-noise ratio (CNR) of the tumors was calculated. Tumor detection was independently performed by three radiologists in a blinded fashion.
RESULTS. The application of the noise filter in the large phantom yielded a reduction of image noise by 42% (p < 0.0001). The CNR values of the tumors in the nonfiltered and filtered large phantom were lower than that in the intermediate phantom (p < 0.05). In the non-filtered and filtered large phantom, 25% and 19% fewer tumors, respectively, were detected on average compared with the intermediate phantom (p < 0.01).
CONCLUSION. The risk of missing hypovascular liver tumors with CT is substantially increased in large patients. A noise filter improves image quality in obese patients.
Obesity is an increasingly common major public health problem in many first-world countries around the globe. According to a recent article [1], 33.8% of adults in the United States—over 90 million people—are obese (defined as a body mass index of ≥ 30). Obesity is associated with various medical conditions, including cardiovascular disease, diabetes mellitus, cholelithiasis, abdominal hernias, and different types of cancers (e.g., breast and colorectal cancer) [2]. In many of these conditions, CT plays an important role in the diagnostic workup and therapeutic follow-up. However, insufficient image quality due to increased image noise is a frequent problem in obese patients.
On abdominal CT examinations, high image noise levels are a critical issue because the noise may obscure subtle low-contrast lesions in parenchymal organs, such as the liver, pancreas, or spleen [35]. Missing a single low-contrast metastasis in an obese patient undergoing hepatic CT for colorectal metastases might have a substantial impact on therapeutic management. To date, to the best of our knowledge, no scientific publication has evaluated the risk of missing hypovascular liver tumors (e.g., metastases from colorectal cancer) in obese patients undergoing MDCT. Furthermore, it is of major interest to determine whether the use of a noise reduction filter can improve the diagnostic efficacy of detection of hypovascular liver tumors in obese patients. Thus, the purpose of our phantom study was to assess the impact of large patient size on the detection of hypovascular liver tumors with MDCT, as well as the effect of a noise reduction filter on image quality and lesion detection in obese patients.

Materials and Methods

Phantom

The custom liver phantom was manufactured to mimic the liver during the portal venous phase and to simulate hypovascular liver tumors of various sizes and contrast enhancements (Fig. 1). The custom attenuation specification of the liver and simulated tumors were obtained from a preliminary unpublished study performed at our institution with 10 consecutive routine clinical patients of intermediate weight who underwent hepatic CT during the portal venous phase at 120 kVp (mean field of view, 380 mm) to evaluate metastases (colorectal cancer, n = 5 patients; breast cancer, n = 2 patients; pancreatic cancer, n = 1 patient; lung cancer, n = 1 patient; and cervical cancer, n = 1 patient). There were a total of 24 metastases (mean diameter, 19.2 mm; range, 8–41 mm) in which the tumor-to-liver contrast values (mean, 49 HU; range, 33–65 HU) and the attenuation difference between normal liver parenchyma and hypovascular tumor were obtained. The mean attenuation value of the liver parenchyma measured 116 HU at 120 kVp.
A cylindric liver phantom (length, 25.5 cm; diameter, 15 cm) with a CT attenuation of 116 HU at 120 kVp was constructed. A total of 45 hypodense spherical lesions, simulating hypovascular liver tumors, were embedded in the liver phantom. The tumors were randomly distributed throughout the phantom. The tumors had three different diameters (5, 10, and 15 mm), with a total of 15 tumors at each diameter. The tumors also had three different tumor-to-liver contrast values, measuring 10, 25, and 50 HU at 120 kVp. Thus, there were a total of five simulated tumors with the same size and tumor-to-liver contrast value. Before the phantom was manufactured, the distributions of the simulated tumors within the phantom were specifically selected to achieve transverse CT images with multiple, one, or no tumors. The construction plan of the phantom served as the reference standard for this study. Although in our preliminary study no lesion had a diameter of 5 mm and a tumor-to-liver contrast value of 10 HU, we chose these specifications to simulate a worst-case scenario for lesion detection.
To simulate intermediate-weight and obese patients, the liver phantom was placed within two water-filled plastic cylindric containers with diameters of 30 and 40 cm, respectively (Fig. 1). The sizes of the two containers were selected to match the abdominal cross-sectional dimensions of the two different patient sizes [6]. The estimated body weight of the simulated intermediate patient ranged between 72 and 85 kg, and that of the obese patient was between 118 and 142 kg.

CT

The two water containers containing the liver phantom were scanned with a 64-MDCT scanner (SOMATOM Sensation Cardiac 64, Siemens Healthcare) using tube current modulation software with an average adaptation strength (CARE Dose 4D, Siemens Healthcare). The scanner was equipped with an 80-kW x-ray tube. Both containers were scanned with the abdominal-pelvic CT protocol that we routinely use for liver lesion detection (120-kVp tube energy, gantry rotation speed of 0.5 s, 160-mAs quality reference tube current–time product, 24 × 1.2-mm collimation, pitch of 1.15, and reconstructed section thickness, 1.5 and 5 mm). The water containers with the two different phantoms were positioned within the isocenter of the CT scanner with their cross-sections perpendicular to the scanner's z-axis. The average tube current value and the volume CT dose index provided by the CT scanner were recorded for the CT scans.

Image Postprocessing With Noise Filter

To reduce the image noise within the simulated obese patient, the CT slice data were postprocessed with a nonlinear 3D optimized reconstruction algorithm filter (3D ORA filter, prototype software, Siemens Healthcare). The details of this noise reduction filter have been described elsewhere [7]. The 1.5-mm transverse CT images of the large container were filtered using the default setting. The filtered series were then reconstructed on a CT workstation (MultiModality Workplace, Siemens Healthcare) as 5-mm transverse CT images, applying the same template that was used for the nonfiltered images. This was done so that the same image position could be determined for each series.

Quantitative Image Analysis

Quantitative image analysis was performed by one author who is a board-certified radiologist with 7 years of experience in CT and who did not act as a reader. The analysis was performed on 5-mm-thick transverse CT images on a high-definition LCD monitor (ME355i2, Totoku Electric). Attenuation measurements of the liver parenchyma and hypodense tumors were made in each of the three datasets (intermediate phantom, the nonfiltered large phantom, and the filtered large phantom). Only the 15-mm tumors were used for attenuation measurements to have enough space for adequate sampling of circular regions of interest (ROIs). The ROIs measured approximately 80 mm2 for the tumors and 1000 mm2 for the liver parenchyma. Attenuation measurements for nine 15-mm tumors were obtained, or three per liver-to-lesion contrast value. A total of three attenuation measurements were obtained for each tumor and the adjacent liver parenchyma. Image noise measurements were also collected in the three datasets. Image noise was defined as the SD of the attenuation value measured in the liver parenchyma. The liver-to-lesion contrast-to-noise ratio (CNR) was calculated as follows: (mean attenuation value of the liver parenchyma–mean attenuation value of the simulated hypovascular lesion) / mean image noise.

Lesion Detection and Qualitative Image Analysis

To assess lesion detection, the 5-mm-thick transverse CT images of the intermediate phantom, the nonfiltered large phantom, and the filtered large phantom were used. Three radiologists with 2, 5, and 9 years of experience in abdominal CT (two of whom are board-certified) evaluated the three CT datasets independently on high-definition LCD monitors (ME355i2). The readers were blinded to the location, size, and number of simulated tumors. Before starting the assessment, each reader was instructed in the criteria for image grading. Readers were asked to mark the location of the tumor along with the grade of conspicuity (1 = very poor, almost not visible; 2 = poor; 3 = intermediate; 4 = good; and 5 = excellent) on a schematic drawing specially designed for the study. The readers noted the slice number (z-indicator) and the precise location on the transverse slice (x- and y-indicator). Readers rated image noise (1 = major, unacceptable; 2 = substantial, above average; 3 = moderate, average; 4 = minor, below average; and 5 = absent) and overall image quality (1 = bad, no diagnosis possible; 2 = poor, diagnostic confidence substantially reduced; 3 = moderate, but sufficient for diagnosis; 4 = good; and 5 = excellent) for the entire CT dataset. The three CT datasets were reviewed in three separate reading sessions in the following order: first, nonfiltered large phantom; second, filtered large phantom; and third, intermediate phantom. The reading sessions were separated by a minimum of 1 week. By starting with the image that represented the most difficult reading task, we attempted to minimize memory bias for detection of the simulated tumors. To further minimize recall bias, the second and third CT datasets were rotated by 90° and 180°, respectively, compared with the first CT dataset. Although images were initially presented with a preset soft-tissue window (window width, 450 HU; window level, 50 HU), readers were allowed to modify window width and level at their own discretion. Ambient room lighting was maintained at a low and constant level for the period of review.
Fig. 1 Set up of liver phantom. Plastic water container on bottom simulated obese patient, and water container on top simulated average-sized patient. Custom liver phantom is placed in center of large container.
TABLE 1: Descriptive Statistics of Image Noise, Contrast-to-Noise Ratio (CNR), and Volume CT Dose Index (CTDIvol) at Various Phantom Dimensions
TABLE 2: Detection of 45 Simulated Hypovascular Liver Tumors in Various Phantoms, With or Without a Noise Reduction Filter

Statistical Analysis

The readers' marks on the evaluation sheets were compared with the localization of the lesions in the phantom. The reading sessions did not involve single CT images; reviewers were provided with the entire CT series with multiple simulated tumors. Thus, true-positive and false-positive findings and image quality parameters at various phantom settings were compared using the analysis of variance for repeated measurements with posthoc tests and Friedman analysis of variance. Multivariate regression analysis was used to assess the effect of size and attenuation of simulated liver tumors, phantom size, and noise filter on the number of true-positive findings. To rule out any possible influence of clustering effects, the probability of detecting a lesion was analyzed with respect to the number of lesions in the same scan position. Interobserver agreement was assessed by calculating the weighted kappa value. Statistical tests were performed using statistical software (StatSoft) and MedCalc software (MedCalc). Values of p less than 0.05 were considered statistically significant.

Results

Quantitative Image Analysis

There was an almost twofold increase in image noise in the nonfiltered large phantom compared with the intermediate phantom (p < 0.0001) (Table 1). The application of the noise reduction filter in the large phantom yielded a reduction of image noise by 42% (p < 0.0001). The image noise of the filtered large phantom was only 13% higher compared with the intermediate phantom (p = 0.15).
The highest CNR values for the simulated hypovascular tumors with the three different tumor-to-liver contrast values were seen in the intermediate phantom (Table 1). When the intermediate was compared with the nonfiltered large phantom, the increase in phantom size resulted in a significant decrease in CNR values (range, 51–55%; p < 0.001). Within the large phantom, the noise reduction filter significantly improved the CNR values (range, 47–69%; p = 0.03). However, the CNR values of the filtered large phantom were still lower than the CNR values of the intermediate phantom (p < 0.05). Because automatic tube current modulation was used, the radiation dose increased substantially at the larger phantom size (Table 1).
Fig. 2 Transverse CT images of phantoms acquired at same location.
A–C, Images show intermediate phantom (A), nonfiltered large phantom (B), and filtered large phantom (C). All three readers detected 15-mm (large arrow, A) and 10-mm hypovascular liver tumor (small arrow, A) in intermediate phantom. However, in nonfiltered and filtered large phantoms, only 15-mm tumor was detected. Note substantially higher image noise in nonfiltered large phantom compared with intermediate and filtered large phantoms.

Assessment of Lesion Detection and Qualitative Image Analysis

The three readers detected, on average, 25% fewer tumors in the nonfiltered large phantom compared with the intermediate phantom (p < 0.001) (Table 2) and (Fig. 2). In the nonfiltered large phantom, no 5-mm tumors were identified (Table 3). The application of a noise filter improved detection of 5-mm tumors in the large phantom, although the difference in lesion detection between the filtered and nonfiltered CT datasets in the large phantom just failed to reach statistical significance (p = 0.054). Significantly fewer 10-mm tumors were detected in the filtered and nonfiltered large phantom than in the intermediate phantom (p < 0.01). No difference was seen in detection of 15-mm tumors within the three CT datasets (p = 0.873). One reader did not detect a 15-mm tumor with a tumor-to-liver contrast of 10 HU in the filtered large phantom. We do not have any other explanation for this finding than inobservance by the reader. Significantly more lesions with a tumor-to-liver contrast of 10 and 50 HU were delineated in the intermediate phantom compared with the filtered and nonfiltered large phantom (p < 0.01). The tumor diameter, the tumor-to-liver contrast, and simulated patient size (p < 0.01) but not the number of simulated tumors per CT image (p = 0.083) had a significant effect on the probability of lesion detection. Differences in the false-positive findings in various phantom settings were not significant (p = 0.384). The conspicuity of the tumors was graded significantly higher in the intermediate phantom compared with the large phantom, both with and without a noise reduction filter (p < 0.0001) (Table 2).
TABLE 3: Number of True-Positive Findings, by Diameter and Contrast Value of Simulated Hypovascular Liver Tumors
Subjective image noise was rated, on average, to be worse in the nonfiltered large phantom compared with the filtered large phantom and the intermediate phantom (2.67, 3.67, and 4.0, respectively). The grading of subjective image quality revealed that readers' rankings were lower for the nonfiltered large phantom than for the filtered large and the intermediate phantom (2.33, 3.0, and 4.67, respectively). The interobserver agreement among the three readers was very good (mean weighted κ, 0.813; range, 0.76–0.86).

Discussion

Imaging obese patients with MDCT poses multiple challenges for the radiologist. Besides image cropping due to a limited field of view and potential exposure to high radiation dose, poor image quality is a key problem when obese patients undergo abdominal CT examination [8, 9]. In obese patients, the image quality is degraded because of an increased attenuation of the x-ray beam by the subcutaneous and intraabdominal fat tissue. Because of inadequate beam penetration, CT images of large patients often have increased image noise. To date, to our knowledge, no studies have determined whether increased image noise results in decreased detectability of hypovascular liver tumors.
Our phantom study revealed a considerable risk of missing hypovascular liver tumors in obese patients undergoing MDCT. When the nonfiltered large phantom was compared with the intermediate phantom, on average, one fourth of the tumors seen in the intermediate phantom were missed in the large phantom by the three readers. The risk of missing hypovascular liver tumors increased with smaller tumor sizes. Significantly fewer 10- and 5-mm tumors were detected in the filtered and nonfiltered large phantoms than in the intermediate phantom. In addition, the CNR values and the conspicuity of the hypovascular tumors were significantly lower in the filtered and nonfiltered large phantoms compared with the intermediate phantom.
Early and correct diagnosis of hepatic lesions is crucial in patients undergoing screening CT examination for liver metastasis to specify the therapeutic management with the best long-term outcome. Therefore, it is important to achieve nearly the same tumor detection rate in large patients as in average-size patients. To reach this goal, if possible, substantial adjustments to abdominal CT protocols are required. In our study, we applied two strategies to combat the greater image noise in obese patients. First, automatic tube current modulation was applied to increase the tube current relative to patient habitus. Although the tube current was automatically increased in the large phantom, image noise almost doubled in the nonfiltered large phantom compared with the intermediate phantom. The x-ray tube of our 64-MDCT scanner did not have enough power to maintain image quality in the large phantom. Second, a nonlinear 3D optimized reconstruction algorithm filter was applied to the CT dataset of the large phantom. The noise reduction filter resulted in improved objective and subjective image quality. Furthermore, a clear trend toward improved detection of 5-mm tumors was seen with the noise reduction filter in the simulated obese patients. Despite the use of automatic tube current modulation and a noise reduction filter, substantially fewer hypovascular liver tumors (19%) were detected in the filtered large phantom compared with the intermediate phantom. To further improve detection of hypovascular liver tumors in obese patients, additional optimization of CT is necessary.
Factors that might improve the detectability of hypovascular liver tumors with MDCT include reduced image noise and increased tumor-to-liver contrast. The reduction of image noise can be achieved by increasing the tube current–time product. With the advent of single high-output x-ray tubes (up to 100 kW) and dual-source CT scanners (up to 200 kW), it is technically feasible to increase the tube current to very high values. In addition, the gantry rotation speed can also be decreased (e.g., from 0.5 s to 1 s) to increase the tube current–time product. Unfortunately, both of these solutions come at the cost of a substantially greater radiation dose to the patient [8, 10]. Future phantom studies are still needed to show the improvements in spatial resolution with an iterative reconstruction algorithm compared with standard convolution filtered back projection. The second factor, the increase in tumor-to-liver contrast, can be achieved by increasing the amount of contrast medium because the degree of maximum enhancement of liver parenchyma during the portal venous phase is directly proportional to the total amount of iodine administered. Because hepatic parenchymal enhancement decreases with increasing patient habitus, it is important to adjust the amount of administered iodine mass according to body size. Different authorities recommend tailoring the amount of contrast medium either to the total body weight or lean body weight [1113]. Future clinical investigations are required to analyze the impact of these three factors described on lesion detection in obese patients.
There are some limitations of our study. First, our liver phantom was simplified to mimic conditions of hepatic enhancement during the portal venous phase using multiple hypovascular tumors of different sizes and tumor-to-liver contrast values. Our liver phantom did not model heterogeneous parenchymal enhancement, distorted hepatic anatomy, or fatty liver disease, any of which may influence the conspicuity of liver tumors. Therefore, it is not clear whether 5-mm tumors with a tumor-to-liver contrast of 10 HU can be detected in intermediate patients. Further in vivo studies are required to investigate the minimum cutoff value with regard to size and contrast for liver lesion detection. Second, our study did not test the diagnostic capability of a dedicated abdominal protocol for obese patients (e.g., increased quality reference tube current–time product, strong adaptation strength, slower rotation time, and so forth). With the availability of automatic tube current modulation, we no longer adjust CT parameters for obese patients at our institution. Third, our phantom study did not take into account the latest CT equipment (e.g., dual-source CT scanner or iterative image reconstruction algorithm). However, to date, only very few institutions are working with this cutting-edge CT technology, whereas the majority of institutions continue to work with single-source CT scanners equipped with 64 slices or less. Fourth, instead of fat-equivalent material, two water-filled containers were used to simulate intermediate-weight and obese patients. Because water attenuates the polychromatic x-ray beam more than fat tissue does and therefore also increases the image noise, our phantom setting represents even larger patient sizes than the actual measured sizes of the containers.
In conclusion, the risk of missing hypovascular liver tumors is substantially higher in obese versus average-sized patients undergoing MDCT. A noise reduction filter improves image quality in obese patients and may improve detection of 5-mm tumors. Future studies are mandatory to evaluate techniques that increase the detection of hypovascular liver tumors in obese patients.

Acknowledgments

We thank Rainer Raupach and Bernhard Schmidt, employees of Siemens Healthcare, for their support.

Footnotes

S. T. Schindera has received a research grant from Siemens Healthcare. R. C. Nelson is a consultant to GE Healthcare.
WEB
This is a Web exclusive article.

References

1.
Flegal KM, Carroll MD, Ogden CL, Curtin LR. Prevalence and trends in obesity among US adults, 1999-2008. JAMA 2010; 303:235–241
2.
Mun EC, Blackburn GL, Matthews JB. Current status of medical and surgical therapy for obesity. Gastroenterology 2001; 120:669–681
3.
Huda W. Dose and image quality in CT. Pediatr Radiol 2002; 32:709–713; discussion, 751–754
4.
Kalra MK, Maher MM, Blake MA, et al. Detection and characterization of lesions on low-radiation-dose abdominal CT images postprocessed with noise reduction filters. Radiology 2004; 232:791–797
5.
Funama Y, Awai K, Miyazaki O, et al. Improvement of low-contrast detectability in low-dose hepatic multidetector computed tomography using a novel adaptive filter: evaluation with a computer-simulated liver including tumors. Invest Radiol 2006; 41:1–7
6.
Menke J. Comparison of different body size parameters for individual dose adaptation in body CT of adults. Radiology 2005; 236:565–571
7.
Bai M, Chen J, Raupach R, Suess C, Tao Y, Peng M. Effect of nonlinear three-dimensional optimized reconstruction algorithm filter on image quality and radiation dose: validation on phantoms. Med Phys 2009; 36:95–97
8.
Schindera ST, Nelson RC, Toth TL, et al. Effect of patient size on radiation dose for abdominal MDCT with automatic tube current modulation: phantom study. AJR 2008; 190:W100–W105
9.
Uppot RN, Sahani DV, Hahn PF, Kalra MK, Saini SS, Mueller PR. Effect of obesity on image quality: fifteen-year longitudinal study for evaluation of dictated radiology reports. Radiology 2006; 240:435–439
10.
Uppot RN, Sahani DV, Hahn PF, Gervais D, Mueller PR. Impact of obesity on medical imaging and image-guided intervention. AJR 2007; 188:433–440
11.
Ho LM, Nelson RC, Delong DM. Determining contrast medium dose and rate on basis of lean body weight: does this strategy improve patient-to-patient uniformity of hepatic enhancement during multi-detector row CT? Radiology 2007; 243:431–437
12.
Kondo H, Kanematsu M, Goshima S, et al. Body size indexes for optimizing iodine dose for aortic and hepatic enhancement at multidetector CT: comparison of total body weight, lean body weight, and blood volume. Radiology 2010; 254:163–169
13.
Yamashita Y, Komohara Y, Takahashi M, et al. Abdominal helical CT: evaluation of optimal doses of intravenous contrast material—a prospective randomized study. Radiology 2000; 216:718–723

Information & Authors

Information

Published In

American Journal of Roentgenology
Pages: W772 - W776
PubMed: 21606267

History

Submitted: July 19, 2010
Accepted: November 22, 2010

Keywords

  1. CT
  2. image quality
  3. liver tumors
  4. noise filter
  5. obese patients

Authors

Affiliations

Sebastian T. Schindera
Institute of Diagnostic, Interventional, and Pediatric Radiology, University Hospital Berne, University of Berne, Freiburgstrasse, CH-3010 Berne, Switzerland.
Jaled Charimo Torrente
Institute of Diagnostic, Interventional, and Pediatric Radiology, University Hospital Berne, University of Berne, Freiburgstrasse, CH-3010 Berne, Switzerland.
Thomas D. Ruder
Institute of Diagnostic, Interventional, and Pediatric Radiology, University Hospital Berne, University of Berne, Freiburgstrasse, CH-3010 Berne, Switzerland.
Hanno Hoppe
Institute of Diagnostic, Interventional, and Pediatric Radiology, University Hospital Berne, University of Berne, Freiburgstrasse, CH-3010 Berne, Switzerland.
Daniele Marin
Department of Radiology, Duke University Medical Center, Durham, NC.
Rendon C. Nelson
Department of Radiology, Duke University Medical Center, Durham, NC.
Zsolt Szucs-Farkas
Institute of Diagnostic, Interventional, and Pediatric Radiology, University Hospital Berne, University of Berne, Freiburgstrasse, CH-3010 Berne, Switzerland.

Notes

Address correspondence to S. T. Schindera ([email protected]).

Metrics & Citations

Metrics

Citations

Export Citations

To download the citation to this article, select your reference manager software.

Articles citing this article

View Options

View options

PDF

View PDF

PDF Download

Download PDF

Media

Figures

Other

Tables

Share

Share

Copy the content Link

Share on social media