Computerized Medical Imaging and Graphics 28 (2004) 289–293

www.elsevier.com/locate/compmedimag

Liver echogenicity: measurement or visual grading?

Tapio Vehmasa,*, Ari Kaukiainena, Katariina Luomaa,b, Martina Lohmanc,
Markku Nurminena, Helena Taskinena,d
a
The Finnish Institute of Occupational Health (FIOH), Topeliuksenkatu 41 a A, FIN-00250 Helsinki, Finland
b
Helsinki University Central Hospital, Peijas Hospital, Sairaalakatu 1, FIN-01400 Vantaa, Finland
c
Orton Hospital, Tenholantie 10, FIN-00280 Helsinki, Finland
d
School of Public Health, University of Tampere, Tampere, Finland
Received 12 June 2003; revised 24 March 2004; accepted 24 March 2004

Abstract
Objective. Two methods to assess liver echogenicity were compared.
Methods. Liver/kidney echogenicity ratio was measured in 41 persons with the ultrasound software and visually graded by two radiologists
and a radiographer. These echogenicity ratios and grades were related to risk factors for fatty liver and to liver enzyme levels.
Results. These determinants explained 55% of the radiologists’ mean grades, 14% of the radiographer’s and 31% of the measured
echogenicity ratios.
Conclusion. Radiologists’ visual gradings correlated best with the indirect determinants of early liver pathology. Computerized
measurements may be inferior to visual grading due to the lack of holistic tissue diagnostics.
q 2004 Elsevier Ltd. All rights reserved.
Keywords: Echogenicity; Enzymes; Fat; Liver; Measurement; Observer; Ultrasound

1. Introduction detect small differences in optical density [9]. In this study

The increased echogenicity of liver, or ‘bright liver’, was
recognized in the 1970s. A prevalence of 20% was reported
from Italy [1]. Body mass index (BMI), age, serum
cholesterol, triglycerides and apo B levels [1], and
hypertransaminasemia [2] are all independent predictors
of a bright liver. In recent studies non-alcoholic fatty liver
has been found to be associated with insulin resistance [3]
and the metabolic syndrome [4]. The increase of liver
echogenicity is caused mainly by fatty degeneration, and
fibrosis is a minor contributor [5,6]. The sensitivity of
ultrasound (US) to detect liver fatty degeneration was 89%
[7] and the method has been considered superior to
computed tomography [8].
Liver echogenicity is usually estimated by comparing it
to that of the right kidney cortex. Subjective estimation is
observer dependent and may be inaccurate. Densitometry
has been regarded indispensable for the accurate density
assessment due to the suboptimal potential of human eye to
* Corresponding author. Tel.: þ 358-474-72481; fax: þ358-9-241-2414.
E-mail address: tapio.vehmas@ttl.fi (T. Vehmas).
0895-6111/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compmedimag.2004.03.003
we compared the quantitative estimation of liver
echogenicity to visual assessment. The determinants
known to be associated with early liver pathology
(liver enzymes) or liver fatty degeneration (BMI, serum
triglyceride level) were regarded as indirect determinants of
liver changes and used as a gold standard in this study.
2. Subjects and methods
2.1. Subjects
The study group consisted of 22 workers occupationally
exposed to solvents (3 women and 19 men, mean age 52
years, range 29 –66 years) and 19 non-exposed persons
(4 women and 15 men, mean age 45 years, range 30 – 57
years), who were employees of the institute (FIOH).
The exposed persons were initially referred to medical
examinations because of suspected solvent-related chronic
toxic encephalopathy or polyneuropathy. Exclusion criteria
included systemic diseases or medication with known
290 T. Vehmas et al. / Computerized Medical Imaging and Graphics 28 (2004) 289–293
hepatic effects, positive hepatitis serology and current
pregnancy. BMI varied between 19 and 36 (mean 26) kg/m2.
This study was conducted with the approval from the
local ethics committee. All subjects participated voluntarily
and provided informed consent.
2.2. Liver sonography and laboratory tests
The SonoAce 8800 MT ultrasound scanner (Medison,
Soul, South Korea; http://www.trimedic.com/ult/8800.htm)
and the convex (3.5 – 5 MHz) probe were used.
The histogram feature of this scanner is capable of
measuring 256 shades in the grey scale. The time-gain
compensation was set to adjust the tissue echogenicity as
constant as possible regardless of depth. Two foci were
used. A representative longitudinal US slice showing the
liver parenchyma and the neighboring right kidney was
selected for the measurement of echogenicity by a
radiologist (TV) and printed with Mitsubishi P 91E thermic
printer (Mitsubishi, Kyoto, Japan) for blinded visual
grading. The echogenicity of the liver parenchyma was
measured in three different regions of interest (ROI) close to
the adjacent right kidney in the same slice and depth,
avoiding vessels. Similar measurements were performed in
the kidney cortex (Fig. 1). Mean values were accepted.
The echogenicity ratio (mean liver echogenicity/mean
kidney echogenicity) was then calculated.
The degree of liver echogenicity was also visually
assessed by two experienced radiologists (KL, ML) and by a
radiographer with little experience in sonography.
They compared the echogenicity of the liver to that of the
adjacent kidney cortex in the printed images by using
reference images [10] (scale: normal ¼ 1, mild
Fig. 1. Echogenicity measurement. A representative area from liver close to the right kidney (1M:31) and the kidney cortex at the same depth (2M:28) show
echogenicities.
abnormality ¼ 2, severe abnormality ¼ 3). The observers
used decimals if they could not match the particular US
image with any of the reference images. Both radiologists’
mean grade was also calculated for each subject.
Blood samples were obtained in the morning after fasting
for 12 h, before liver sonography was performed.
They included serum alanine aminotransferase (ALT),
aspartate aminotransferase (AST), gamma-glutamyl
transferase (GGT), carbohydrate-deficient transferrin
(CDT) and triglycerides. Standard laboratory methods
were used.
2.3. Statistical methods
The agreement between the three repeated echogenicity
measurements and, on the other hand, the agreement
between the readers’ grades was assessed by using
the intraclass correlation coefficient (ICC). Measured
(liver/kidney) echogenicity ratios and readers’ grades were
compared with Pearson’s correlation.
The relation between the indirect determinants of early
liver pathology and liver echogenicity was studied using
linear regression modeling. To comply better with the
assumption of the normality of the dependent echogenicity
variables (e.g. the symmetry of the distribution) we
subjected them to square root transformation. The
determinants were entered into the model as quadratic
variables because of the non-linearity of some of the
relations. From the regression analysis results we calculated
the coefficient of determination, that is, the square of the
product-moment correlation between two variables, r 2 :
This quantity expresses the proportion of the variance of
one variable given by the other [11]. We related both
 T. Vehmas et al. / Computerized Medical Imaging and Graphics 28 (2004) 289–293 291

ðp , 0:001Þ for radiologist 1, 0.28 ðp ¼ 0:07Þ for radiologist

Fig. 2. Radiologists’ grading vs. echogenicity measurement. The curve
represents a locally weighted scatterplot smooth regression. Estimate of the
linear regression: y ¼ 1:23 þ 0:35x; r2 ¼ 0:2; p ¼ 0:002:
the radiologists’ echogenicity grades separately, their mean
values and the measured echogenicity ratios to the indirect
determinants of early liver pathology by using linear
multiple regression modeling. In the model, the grade and
the ratio had a multiple correlation, R; with a set of their
determinants. The R-squared, R2 ; indicates the fraction of
the total variation in the liver echogenicity, which is
accounted for by the regression model. S-PLUS 2000
system was applied for data analysis (MathSoft, Inc.,
Seattle, WA, USA, 1999).
3. Results
The ICC between the three repeated echogenicity
measurements was 0.93 for the liver and 0.94 for the right
kidney. The ICC of the visual grades was 0.65 between
radiologists 1 and 2, 0.45 between radiologist 1 and
the radiographer, and 0.37 between radiologist 2 and the
radiographer.
Correlations between the measured (liver/kidney)
echogenicity ratio and the readers’ grades were: 0.55
2, and 0.02 ðp ¼ 0:9Þ for the radiographer. There were slight
differences between the readers’ mean grades of
echogenicity (radiologist 1: 1.92, radiologist 2: 1.79,
radiographer: 1.69). The average grade of the radiologists
was plotted as a function of the measured echogenicity
ratio (Fig. 2).
Coefficients of determination between echogenicity
(estimated by the readers or measured with US) and the
indirect determinants of early liver pathology
(study subjects’ BMI, serum triglyceride and liver enzyme
levels) are given in Table 1. The measured echogenicity was
poorly explained by BMI, triglycerides, and CDT, while the
readers’ echogenicity grade was better explained by ALT,
AST, and best by BMI. The radiologists’ grades correlated
better with the indirect determinants of early liver pathology
than the radiographer’s grades. The mean grades of the two
radiologists correlated better with these determinants than
individual radiologists’ grades.
4. Discussion
The radiologists’ echogenicity grades showed better
correlation with the indirect determinants of early liver
pathology than did the echo ratio measurements or the
grades of the radiographer. This fact was especially
pronounced in relation to the BMI.
Liver biopsies were not clinically indicated and thus
could not be obtained for ethical reasons. Due to the lacking
histological gold standard we compared our echogenicity
findings with the combination of variables previously
known to be associated with fatty liver or non-specific
liver parenchymal damage. We did not consider histology
mandatory, because our aim was to compare the visually
based and computerized US methods with each other and
not to study associations between liver histology and US.
The measured echogenicity deep in tissue reflects both
the intrinsic echogenicity of the area itself and the acoustic
properties of all tissues between the probe and this area.
Especially in obese patients, the heavy contribution of

Table 1
Indirect determinants of early liver pathology in relation to its echogenicity

Echogenicity grade/measurement BMI Trigly AST ALT GGT CDT R2 (total)

Radiologist 1 0.30 0.07 0.24 0.28 0.19 0.02 0.48

Radiologist 2 0.25 0.13 0.29 0.24 0.07 0.08 0.48
Mean of radiologists 1 and 2 0.33 0.12 0.32 0.32 0.15 0.05 0.55
Radiographer 0.10 0.00 0.02 0.03 0.03 0.03 0.14
Mean of all readers 0.36 0.09 0.27 0.27 0.15 0.06 0.52
US measurement (liver/kidney ratio) 0.03 0.07 0.15 0.21 0.16 0.00 0.31

Coefficients of determination ðR2 Þ between liver echogenicity (estimated by the readers or by measured echo ratio) and body mass index (BMI) and
laboratory tests (trigly, serum triglycerides; AST, aspartate amino transferase; ALT, alanine amino transferase; GGT, gamma glutamyl transferase; CDT,
carbohydrate-deficient transferrin).
292 T. Vehmas et al. / Computerized Medical Imaging and Graphics 28 (2004) 289–293
superficial tissues may cause unreliability of measurement.
Our good ICCs between repeated echogenicity measure-
ments suggest that this method may theoretically be the best
to quantify the shades of grey according to what
Smith-Levitin et al. [9] have found. However, this technical
ability may not be sufficient for reliable grading of liver
echogenicity. Computerized measurements are restricted to
ROI areas, which are usually set small to exclude disturbing
structures such as vessels, bile ducts or acoustic shadows
from other tissues. On the contrary, the human eye may
register the whole liver parenchyma and kidney cortex
without artificial ‘sampling’ of tissue, in a more holistic
manner. Medically trained observers may also be able to
exclude the effect of artifacts. In addition to echogenicity,
the reference images focused also to the visibility of
vascular markings and the diaphragm. Our observers could
take into account the effect of these features and of liver size
in their evaluation.
Our radiologists had better results than the radiographer
did. This observation contrasts the finding of Smith-Levitin
et al. [9], who found no difference in echogenicity
estimation due to medical qualification. Their study was
designated to compare mere tissue echo density, however.
As a conclusion, computerized measurements of liver
echogenicity suffer from the lack of an integrated
diagnostic approach, which can be reached only by
medically trained observers. US echogenicity measure-
ments should therefore, be dealt with care, even if they
are echo ratios. Experienced observers’ grading is
currently the preferred method for ultrasonic evaluation
of liver echo texture.
5. Summary
We compared the visually based and computerized
measurement method for assessing the liver echogenicity
and its minimal changes. The echogenicity of the liver and
adjacent kidney cortex was measured in 41 persons (aged
29 –66 years) with the US software to determine the
liver/kidney echo ratio. Also, two radiologists and a
radiographer graded liver echogenicity visually by using
reference images. The echo ratios and visual grades were
related to risk factors for fatty liver (BMI and the level of
serum triglycerides) and to liver enzymes, the combination
of which was used as an indirect gold standard of early liver
pathology. These determinants explained 48% of both
radiologists’ visual gradings, 55% of the radiologists’ mean
gradings, 14% of the radiographer’s gradings, and 31% of
the measured echogenicity ratios. The radiologists’ visual
grades were thus best indicators of early liver pathology.
Local echo artifacts especially in obese individuals and the
lack of holistic tissue diagnostics may impede computerized
echogenicity measurements. An experienced radiologist
pays attention to vascular architecture and the effect of
liver size, diaphragmatic visibility and artifacts in addition
to the general echogenicity of liver. Experienced observers’
grading is currently the preferred method for the diagnostic
US evaluation of early liver abnormality.
Acknowledgements
We wish to thank radiographer Elise Koskenseppä for
estimating liver echogenicity and Ms. Terttu Kaustia, MA,
for language revision.
References
[1] Lonardo A, Bellini M, Tartoni P, Tondelli E. The bright liver
syndrome. Prevalence and determinants of a bright liver echopattern.
Italian J Gastroenterol Hepatol 1997;29(4):351– 6.
[2] Frenzese A, Vajro P, Argenziano A, Puzziello A, Iannucci MP,
Saviano MC, Brunetti F, Rubino A. Liver involvement in obese
children. Ultrasonography and liver enzyme levels at diagnosis and
during follow-up in an Italian population. Dig Dis Sci 1997;42(7):
1428–32.
[3] Marchesini G, Brizi M, Morselli-Labate AM, Bianchi G, Bugianesi E,
McCullough AJ, Forlani G, Melchionda N. Association of nonalco-
holic fatty liver disease with insulin resistance. Am J Med 1999;
107(5):450–5.
[4] Lonardo A, Bellini M, Tondelli E, Frazzoni M, Grisendi A,
Pulvirenti M, Della Casa G. Nonalcoholic steatohepatitis and the
bright liver syndrome: should a recently expanded clinical entity
be further expanded? (letter). Am J Gastroenterol 1995;90(11):
2072–3.
[5] Layer G, Zuna I, Lorenz A, Zerban H, Haberkorn U, Bannasch P, van
Kaick G, Rath U. Computerized ultrasound B-scan texture analysis of
experimental diffuse parenchymal liver disease: correlation with
histopathology and tissue composition. J Clin Ultrasound 1991;
19(11):193–201.
[6] Suzuki K, Hayashi N, Sasaki Y, Kono M, Kasahara A, Fusamoto H,
Imai Y, Kamada T. Dependence of ultrasound attenuation of liver
on pathologic fat and fibrosis: examination with experimental fatty
liver and liver fibrosis models. Ultrasound Med Biol 1992;18(8):
657 –66.
[7] Joseph AE, Saverymuttu SH, Al Sam S, Cook MG, Maxwell JD.
Comparison of liver histology with ultrasonography in
assessing diffuse parenchymal liver disease. Clin Radiol 1991;43(1):
26– 31.
[8] Mendler MH, Bouillet P, Le Sidaner A, Lavoine E, Labrousse F,
Sautereau D, Pillegand B. Dual-energy CT in the diagnosis and
quantification of fatty liver: limited clinical value in comparison to
ultrasound scan and single-energy CT, with special reference to iron
overload. J Hepatol 1998;28(5):785 –94.
[9] Smith-Levitin M, Blickstein I, Albrecht-Shach AA, Goldman RD,
Gurewitsch E, Streltzoff J, Chervenak FA. Quantitative assessment of
grey-level perception: observers’ accuracy is dependent on density
differences. Ultrasound Obstet Gynecol 1997;10(5):346–9.
[10] Brodkin CA, Daniell W, Checkoway H, Echeverria D, Johnson J,
Wang K, Sohaey R, Green D, Redlich C, Gretch D, Rosenstock L.
Hepatic ultrasonic changes in workers exposed to perchloroethylene.
Occup Environ Med 1995;52(10):679–85.
[11] Matthews DE, Farewell VT. Using and understanding medical
statistics, 3rd ed. Basel: Karger; 1996.
 T. Vehmas et al. / Computerized Medical Imaging and Graphics 28 (2004) 289–293
Tapio Vehmas completed his MD from Helsinki University in 1984.
He received his specialist degree in radiology in 1991 and published his
doctoral thesis work in the same year. He has worked in the Helsinki
University in 1991–1993 as a lecturer (assistant professor) teaching
radiology and in the Helsinki University Central hospital as a senior
radiologist during 1991–1998. He was appointed to a senior lecturer
(docent) of radiology in Helsinki University in 1997. Since 1999 he has
worked as the chief radiologist at the Finnish Institute of Occupational
Health.
Ari Kaukiainen completed his MD from Helsinki University in 1989
and received his specialist degrees in occupational health in 1998 and
occupational medicine in 2000. Since 1998 he has worked at the
Finnish Institute of Occupational Health, and since 2001 in charge of a
work ability assessment unit at the Department of Occupational
Medicine. He is currently preparing his doctoral thesis about health
effects of organic solvents and paint compounds.
Katariina Luoma completed her MD from Tampere University in
1979 and received her specialist degree in radiology in 1986 and in
neuroradiology in 2002. During 1992–1993 she worked at the Finnish
Institute of Occupational Health. She published her thesis work in 2000.
She has worked as a senior radiologist during 1991–1996 and since
1997 as the deputy head of department at the Department of radiology,
Peijas Hospital, Helsinki University Central Hospital. Her scientific
interest is mainly focused on MRI and epidemiology of musculoske-
letal disease.
293
Martina Lohman completed her MD from the University of Helsinki
in 1986. She made a two-year residency in the field of surgery and
orthopedics before shifting to radiology. In 1995 she completed her
specialization in radiology, and published her doctoral thesis in 2001.
From 1996 to 2002 she worked as a senior radiologist at the Helsinki
University Central Hospital. During 2002–2003 she was appointed as a
Clinical Assistant Professor at the University of Michigan, USA. Her
publications are mainly in the field of musculoskeletal MRI.
Markku Nurminen works as a senior research scientist in the
Department of Epidemiology and Biostatistics, Finnish Institute of
Occupational Health. He holds the position of senior lecturer (docent)
in biometry at the Department of Public Health, University of Helsinki,
being a specialist in biostatistics and epidemiological methods. His
research has focused on the field of occupational and public health for
30 years, and recently on estimating working life expectancies. He is a
(co)author of about 150 publications.
Helena Taskinen is the director of the Department of Occupational
Medicine at the Finnish Institute of Occupational Health, and Professor
of Occupational Health at the School of Public Health at the University
of Tampere, Finland. Her research experience covers occupational
reproduction epidemiology, occupational diseases (solvent induced
diseases, musculo-skeletal diseases of hand and arm from repetitive
work, respiratory diseases of workers at water damaged buildings),
work conditions and well being of pupils and personnel at schools,
training and education of occupational health personnel and work
conditions, as well as health and occupational services of contingent
workers.