вход по аккаунту



код для вставкиСкачать
J Clin Ultrasound 24:25-29, January 1996
CCC 0091-2751/96/010025-05
0 1996 by John Wiley & Sons, Inc.
Sonographic Diagnosis of Fatty Liver Using a
Histogram Technique That Compares Liver and
Renal Cortical Echo Amplitudes
Hiroyuki Osawa, MD,* and Yasuaki Mori, MD?
Abstract: Accurate diagnosis of fatty liver using ultrasonography was attempted
based on the difference between the echo intensities of the liver and kidney determined
from ultrasonic histograms. Livers were then classified as having fatty infiltration,
normal histology, or intermediate histology based on CT ratios established previously
in earlier work comparing non-contrast-enhanced liver and spleen. The hepatorenal
difference was significantly greater in the fatty liver group than in the normal liver
group (8.9 2.0 dB vs 2.5 4.5dB, p < 0.001). When a hepatorenal difference of 37.0
dB was taken as the criterion, this method had a sensitivity of 91.3%, a specificity
of 83.8%, and an accuracy of 86.7% for the diagnosis of fatty liver. Thus, quantitative ultrasonic diagnosis of fatty liver can be performed using echo intensity histograms. 0 1996 John Wiley & Sons, Inc.
Indexing Words: Liver . Histogram . Computed tomography . Kidney
In the diagnosis of fatty liver, ultrasonic imaging
is simple and provides useful information, but it
is dependent on the subjective interpretation of
the observer. Quantitative evaluation of echo intensity by means of histograms has been reported
as a way of overcoming the variability due to subjective observer
Computed tomography (CT) is excellent for detecting fat deposition and is a useful imaging method for the
assessment of fatty liver. There is a significant
correlation between the tissue fat content of the
liver and the CT number3 because x-ray absorption decreases as more fat is deposited, and there
is a resultant decrease in the CT number. This
allows even mild fatty infiltration to be detected
by CT, generally on the basis of the ratio of the
CT numbers for the liver and the spleen, ie, the
hepatolienal ratio. In the present study, we compared the echo intensities of the liver and kidney
using ultrasonic histograms based on the hepatolienal ratio values determined by CT, and as-
From the Departments of *Internal Medicine and tSurgery,
Nishiarita Kyoritsu Hospital, Japan. For reprints contact Hiroyuki Osawa, MD, Omiya Medical Center, Jichi Medical
School, 847-1, Amanuma, Omiya City, Saitama 330, Japan.
VOL. 24, NO. 1, JANUARY 1996
sessed the usefulness of this method for making
an objective diagnosis of fatty liver.
Between December 1991 and March 1993, 70 patients (33 men and 37 women; mean age: 39.2
years) among 351 who underwent abdominal CT
scanning were enrolled in this prospective study.
Thirty-one patients had a suspected diagnosis of
fatty liver on the basis of ultrasonic imaging and
subsequently underwent abdominal CT scanning.
Thirty-nine patients were investigated for various abdominal diseases, and a diagnosis of normal liver or fatty liver was made by CT, after
which abdominal ultrasonography was performed. Only patients who gave informed consent
after a full explanation of the study and the procedures involved underwent the second imaging
procedure. Thirty-three patients rejected the second imaging procedure. Two hundred forty-eight
patients with positive HBs antigen, positive HCV
antibody, regular alcohol drinking, liver tumors,
splenomegaly, or renal dysfunction including proteinuria and hematuria were excluded.
CT scanning was performed with a Toshiba
TCT9OOS scanner (Toshiba Co., Tokyo) without
the use of contrast medium. The window level and
window width were kept constant at 50 and 250,
respectively. The hepatic CT number was measured at 2 to 4 sites in the right lobe, and the
splenic CT number was determined at 1to 2 sites
(Figure 1).Then the hepatolienal ratio was calculated from the respective means of these values. On the basis of previously reported crite23 patients with a hepatolienal ratio SO.90
were considered to have fatty liver, 37 patients
with a ratio 31.0 were defined as having normal
livers, and 10 patients with a ratio of 0.91 to 0.99
were classified as intermediate (Table 1).A histological diagnosis was not made in any of the
patients because liver biopsy was not indicated
Abdominal ultrasonic examination was performed using a Yokogawa RT4600 instrument
(Yokogawa Medical Systems, Tokyo) and a convex 3.5 MHz probe. The time gain control was set
at its central position, and the power was set at a
constant level, with a gain of 60 dB. The probe
was positioned in a right intercostal scan in each
patient so that stable parenchymal echo images of
the liver and the right kidney were obtained simultaneously with no vessels in the images. Subsequently the echo intensity of the liver and the
Assignment of Subjects on the Basis of the CT
Hepatolienal Ratio and the Ultrasonic
Hepatorenal Difference
Hepatolienal ratio (CT)
difference (US)
3 7 dB
0 dB
Fatty liver
right renal cortex was measured on these images
using a pixel-based histogram measurement unit
installed in the apparatus.
Each region of interest chosen in the liver and
the right renal cortex was almost in contact with
the region of the other organ in the same ultrasonic image. The region of interest in both organs
was defined as a circle 1cm in diameter (576 pixels) (Figure 2). The histogram showed the frequency distribution of the echo intensity in the
region of interest, and the most frequent intensity
(MODE) was defined as the echo intensity of that
region. Hepatorenal contrast was assessed from
the difference in echo intensity between these organs, and the mean value of four measurements
was used for analysis.
FIGURE 1. Regions of interest used in determining the CT numbers of the liver and spleen.
FIGURE 2. Echo histograms of regions of interest in the liver (upper panels) and kidney (lower panels). Each
histogram shows the frequency distribution of the echo intensity i n the region of interest. The echo intensity
of the region is defined by the most frequent gradient (MODE). The percentage in the figure represents the
percentage area showing intensity of the MODE within the total region of interest. The average MODE value
for four measurements was used for analysis. For example, i n this figure, the average echo intensity of the
liver is 14.5 dB (two of the four pictures used for calculation are shown) and the average echo intensity of the
kidney (also t w o of the four pictures are shown) is 7.5 dB. The hepatorenal difference is calculated t o be 7 dB
by subtracting 7.5 from 14.5. The hepatolienal CT ratio in this case was 0.88, indicating fatty liver.
Statistical Methods
Of differences among the
three groups was performed with student's t test
using the Bonferroni correction, andp< 0'05 was
taken to indicate significance. The sensitivity,
specificity, accuracy, positive predictive value,
and negative predictive value of the diagnosis of
fatty liver by the histogram method were obtained by comparison of the results in patients
with fatty liver and others (the intermediate
group plus the normal liver group) as defined by
VOL. 24, NO. 1, JANUARY 1996
CT scanning. In addition, these parameters were
calculated between patients with a normal liver
and others (the fatty liver group plus the intermediate group), and between those with a fatty
liver and those with a normalliver, excluding intermediate values.
The ultrasonic hepatorenal differences and the
CT hepatolienal ratios showed an inverse corre27
lation ( r = - 0.58, p < 0.001). In the normal liver
group, the mean ultrasonic hepatorenal difference was 2.5 4.5 dB (mean ? SD). In contrast,
the mean hepatorenal difference was 8.9 t 2.0 dB
in the fatty liver group, and it was significantly
greater than in the normal liver group ( p < 0.001,
Bonferroni correction with t test) (Figure 3). In
addition, the mean ultrasonic hepatorenal difference was 4.6 -+. 5.6 dB in the intermediate group,
was midway between the values for the fatty liver
and normal liver groups, and was not significantly different from either of these values.
A hepatorenal difference of 27.0 dB was defined as the criterion indicating fatty liver since
the hepatorenal difference in the fatty liver group
lay mostly within the range 6.9 dB to 10.9 dB and
that in the normal liver group mostly within the
range -2.0 dB to 7.0 dB from the above results.
This criterion provided a sensitivity of 81.8%, a
specificity of 83.8%, an accuracy of 82.9%, a positive predictive value of 81.8%, and a negative
predictive value of 83.8% with respect to CT diagnosis when the fatty liver group diagnosed by
CT included the intermediate group, that is, a
hepatolienal CT ratio <1.0 was defined as fatty
liver and the rest were defined as normal.
When the intermediate group was included in
the normal group, that is, when a hepatolienal CT
ratio ~ 0 . 9 was
0 defined as fatty liver and the rest
were defined as normal, the sensitivity (91.3%)
and the negative predictive value (94.6%) were
increased, but the specificity (74.5%), accuracy
(80.0%), and positive predictive value (63.6%)
were decreased as expected theoretically. When
the patients in the intermediate group (CT ratios
between 0.91 and 0.99) were excluded, this ultra-
P < 0.001
... ....
.. ,,
.......................................................................................... *.......................
Fatty Liver
FIGURE 3. Hepatorenal difference values in the fatty liver group and
the normal liver group. There is a significant difference between the
two groups.
sonic diagnosis achieved a quite high sensitivity
(91.3%),specificity (83.8%),accuracy (86.7%),and
negative predictive value (93.9701, at the expense
of the positive predictive value (77.8%).
Taylor et a15 reported that ultrasound reveals a
high-intensity echo in the hepatic parenchyma of
patients with fatty liver, and since then this has
been called “bright liver.” However, the diagnosis
of “bright liver” based on ultrasound studies may
be influenced by various factors, including an increase in the decrement coefficient, echoes from
intrahepatic vessels, indistinct band signs, and
the hepatorenal ~ o n t r a s tThus,
. ~ ~ ~the same findings may be interpreted differently by different
examiners. Accordingly, quantitative assessment
of echo intensity using histograms has been attempted to provide an objective method for the
diagnosis of fatty 1 i ~ e r . These
l ~ ~ studies were
performed in subjects who were previously demonstrated by biopsy to have fatty or normal livers
and who were not selected randomly.
Because fat deposition in the liver parenchyma
is likely to be a continuing process and thus produce a broad spectrum of values for the hepatorenal difference, we postulated that a better histogram could be devised from randomly selected
subjects, including those in whom it was difficult
to distinguish between normal and fatty liver by
CT scanning. We chose CT as our gold standard.
It has been reported that the CT number of
normal liver ranges widely from 40 t o 70 Hounsfield units (HU), and that it is 7 t o 8 HU higher
than the CT number of the spleen.’ As fat accumulates in the liver, the x-ray absorption of the
tissues decreases, leading to a decrease in the CT
number. Histologically, a liver in which 30% or
more of the hepatocytes contain fat droplets is
generally regarded as showing fatty change.
In the present study, we evaluated the hepatorenal difference using the MODE shown in the
histogram images. In our apparatus the histogram was able to show both the most frequent
gradient (MODE) and the mean calculated from
the frequency distribution of the echo intensity in
the region of interest. Both parameters can be
used as the representative value in the region of
interest because these showed almost the same
values, as shown in Figure 2. However, the
MODE was chosen as the representative echo intensity in the region of interest because it might
be influenced by focally heterogeneous echo imJOURNAL OF CLINICAL ULTRASOUND
ages to a lesser extent than the mean of the echo
Our results revealed that the hepatorenal difference in the fatty liver group was significantly
greater than that in the normal liver group. Our
method achieved high sensitivity (91.3%), specificity (83.8%), accuracy (86.7%), positive predictive value (77.8%),and negative predictive value
(93.9%)with respect to CT diagnosis when a hepatorenal difference 3 7 . 0 dB was taken to indicate
fatty liver. Therefore, a hepatorenal difference of
7.0 dB or more strongly suggests the presence of
fatty liver, and a difference of 8 dB or more indicates an almost definite diagnosis despite the possible influence of observer bias and interpatient
It is generally considered that ultrasound is
inferior to CT for making a diagnosis of fatty
liver. CT and liver biopsy are very useful for making an objective and quantitative diagnosis, but
our study suggests that reliable quantitative
evaluation of fatty liver is also feasible by ultrasound on the basis of the hepatorenal difference.
Accurate diagnosis of fatty liver by ultrasound
alone would be desirable from the point of view of
decreasing costs and the physical burden that CT
and biopsy place on patients.
Although the measurements obtained may
vary from apparatus t o apparatus, the method
presented here shows promise as a simple, noninvasive technique for the detection of fatty liver. It
is important to select adequate regions of interest
in both the liver and kidney in order to obtain
higher accuracy. To standardize this selection so
as to provide a representative sample and avoid
bias, an effort must be made to obtain stable parenchymal echo images of each organ and, in addition, to select regions of interest which are more
VOL. 24, NO. 1, JANUARY 1996
homogeneous and have an echogenicity similar to
that of surrounding images.
We thank H. Nagaishi and M. Nagaishi for their
assistance with radiological examinations. We
are grateful to Dr. H. Origasa, Toyama Medical
and Pharmaceutical University, for his generous
advice about statistical interpretation of our data.
We also thank Dr. M. Kawakami, Omiya Medical
Center, for his help in preparing the manuscript.
1. Itoh K, Yasuda Y, Aihara T: Acoustic intensity histogram pattern diagnosis of the liver diseases. J
Clin Ultrasound 13:449, 1985.
2. Taniguchi N, Itoh K, Suzuki 0: Estimation of hepatorenal echo contrast by echo-intensity histogram
and its problem. Ultrasonics 18:856, 1991.
3. Kawata R, Sakura K, Kunieda T: Quantitative evaluation of fatty liver by computed tomography in rabbits. AJR, 142741, 1984.
4. Tsunetomi S, Itoh S, Ohfuji M: Diagnosis of fatty
liver by ultrasound and computed tomography (in
Japanese). Kan Tan Sui 10:365, 1985.
5 . Taylor KJM, Carpenter DA, Hill CR. Gray scale ultrasound imaging: the anatomy and pathology of the
liver. Radiology 119:415, 1976.
6. Foster KJ, Dewbury KC, Griffith AH: The accuracy
of ultrasound in the detection of fatty infiltration of
the liver. Brit J Radio1 53~440,1980.
7. Hagga J, Reich NE: Computed Tomography of A b dominal Abnormalities. Mosby, St. Louis, 1978, p.
8. Piekarusky J, Goldberg HI, Roal SA: Difference between liver and spleen CT numbers in presence of
diffuse liver diseases, Radiology 137:727, 1980.
Без категории
Размер файла
553 Кб
Пожаловаться на содержимое документа