Ultrasound in Meal. & Biol. Vol. 18, No. 8. pp. 657-666, 1992 Printed ill the U.S.A,
0301-5629/92 $5.00 + ,00 (c) 1992 Pergamon Press Lid.
OOriginal Contribution DEPENDENCE OF ULTRASONIC ATTENUATION OF LIVER ON PATHOLOGIC FAT AND FIBROSIS: EXAMINATION WITH EXPERIMENTAL FATTY LIVER AND LIVER FIBROSIS MODELS K. S u z u ~ , *
N . HAYASHI, t Y. SASAKI, t M . KONO, t A. KASAHARA, t H . FUSAMOTO, t Y. IMAI * a n d T. KAMADA t
tFirst Department of Medicine, Osaka University Medical School, 1-1-50 Fukushima, Fukushima-ku, Osaka 553, Japan, and *Research and Development Engineering, Process Instrumentation Division, Shimadzu Corporation, Kyoto 604, Japan (Received 23 October 1991; in final form 26 May 1992) Abstract--To clarify the effect of the pathological state of the liver on ultrasonic attenuation, we produced two experimental rabbit models. The influence of fat on ultrasonic attenuation was examined using a fatty liver model without liver fibrosis, and that of fibrosis on attenuation using a liver fibrosis model without fatty infiltration. Ultrasonic data were obtained in vivo directly from the liver, and an acoustic attenuation coefficient slope was obtained by the spectral difference method. Tissue components of the liver, namely the total lipid, hydroxyproline and water contents, were measured precisely by quantitative methods. We revealed that ultrasonic attenuation depends mainly on fatty infiltration of the liver and to a lesser extent on fibrosis, but not on the water content.
Key Words: Attenuation, Ultrasound, Rabbit liver, Experimental fatty liver model, Experimental liver fibrosis model, Fat content, Collagen content, Water content.
INTRODUCTION
ture. A positive correlation between attenuation and fibrosis has been reported (Afschrift et al. 1987; Duerinckx et al. 1988; Lin et al. 1988). However, some authors found little or no elevation of attenuation with fibrosis alone (Bamber et al. 1981; Taylor et al. 1986; Wilson et al. 1987), while others reported that the attenuation value varies with the etiology of the liver cirrhosis (Maklad et al. 1984). There are the following possible reasons for such contradictory findings: i. In a human study, fibrotic livers often exhibit varying degrees of fatty infiltration, and the fatty infiltration itself elevates attenuation, making interpretation of the attenuation results complicated. 2. In most human studies, the contents of fat and fibrosis were not quantitatively measured. Liver specimens were only classified into 3 or 4 groups based upon the microscopic evaluation of the fat and fibrosis by pathologists. In addition, precise grading is not always possible for a small biopsy specimen. 3. The influence of the abdominal wall on the ultrasonic beam and acoustic attenuation has not been thoroughly investigated, and they may affect the attenuation measurement.
Conventional gray scale B-mode ultrasound of the liver is a well-established modality in the diagnosis of focal lesions within the liver. However, diffuse parenchymal diseases of the liver, such as liver cirrhosis and fatty infiltration, are difficult to differentiate on the basis of B-mode imaging alone (Gosink et al. 1979; Sandford et al. 1985). Furthermore, sonographic grading of these diseases is not precise enough to predict the severity of the pathological changes (Needleman et al. 1986). The attenuation of ultrasound in the liver has received considerable attention as a quantitative ultrasonic tissue characterization technique. A positive correlation between fatty infiltration of the liver and attenuation has been reported by many researchers, both in vitro (Bamber et al. 1981; Lin et al. 1988) and in vivo (Afschrift et al. 1987; Duerinckx et al. 1988; Taylor et al. 1986; Wilson et al. 1987). However, as to the influence of fibrosis of the liver on ultrasonic attenuation, there has been some disagreement in the literaAddress correspondence to: Norio Hayashi, M.D., First Department of Medicine, Osaka University Medical School, 1-1-50 Fukushima, Fukushima-ku, Osaka 553, Japan. 657
658
Ultrasound in Medicineand Biology
4. A negative correlation between the water content of the liver and in vitro attenuation has been suggested (Bamber et al. 1981), but the water content was not measured in other studies. 5. The results of in vitro studies may be different from those of in vivo ones and thus difficult to interpret; the vasculature of resected specimens has collapsed and thus contains less blood, of which attenuation is low. On the other hand, postmortem specimens contain microbubbles, which may elevate attenuation. In the present study, to clarify the effect of the pathological state of the liver, namely fibrosis and fat, on ultrasonic attenuation, we used two experimental animal models. One is a liver fibrosis model without fatty infiltration and the other a fatty liver model without liver fibrosis. Ultrasonic data were obtained in vivo directly from the liver through an incision in the abdominal wall, and the determined attenuation coefficient was compared with the contents of fat, fibrosis and water, measured by quantitative methods. MATERIALS AND M E T H O D S
Materials Twenty-four male albino rabbits weighing 2.02.8 kg were used. Experimental fatty liver model (n = 6). Six rabbits were intravenously administered 80 mL/kg of 10% Intrafat (Daigo Nutritive Chemicals, Osaka, Japan) daily for 6 to 9 weeks. Intrafat is a fat emulsion for intravenous infusion, containing soybean oil 10.0 g, phospholipid 1.2 g and glycerol 2.5 g per 100 mL. Experimental liverfibrosis model (n -- 12). Equal volumes of carbon tetrachloride and olive oil were emulsified, and the rabbits received weekly subcutaneous injections of 0.2 mL/kg. Subcutaneous injections of 0.2 mL of 10% phenobarbital were given simultaneously to induce fibrosis in a shorter period (McLean et al. 1969). The rabbits were treated for 3 to 8 months to induce varying degrees of fibrosis. It is well-known that the administration of carbon tetrachloride causes acute liver damage, especially necrosis and steatosis (Cameron and Karunaratne 1936). However, these changes disappear in a short period (Cameron and Karunaratne 1936; Jorgensen et al. 1974; Parker and TuthiU 1986). Therefore, in order to avoid the influence of acute liver damage on ultrasonic attenuation, liver fibrosis model rabbits were subjected to ultrasonic data acquisition 10 days after the last administration of carbon tetrachloride. Normal controls (n = 6). Six rabbits which received no injections were used as controls.
Volume18, Number 8, 1992
Instrumentation The equipment consisted of an A-mode transducer (nominal center frequency 10 MHz, 7 mm in diameter) (Shimadzu Corp., Kyoto, Japan), a pulserreceiver (Shimadzu Corp.), an A/D converter (R/ 390AD; SONY-TEKTRONIX, Tokyo, Japan), a personal computer (PC-9801 VM2; NEC, Tokyo, Japan) and an oscilloscope (T912; TEKTRONIX, Beaverton, OR, USA). A block diagram of the system is presented in Fig. 1. The focal length of the transducer was 16 mm, and the - 2 0 dB bandwidth extended from 6.3 MHz to 13.6 MHz, as was found from the spectrum of the signal reflected from a glass reflector immersed in water. A pulser-receiver was used to excite the transducer and linearly amplify the returning radio frequency (RF) waveforms. The output signals were digitized at a sampling rate of 60 MHz with a resolution of 10 bits with the A/D converter, transferred to the memory of the personal computer, and stored on a floppy disk. Nonlinear processing, such as log-amplification or sensitivity-time control gain was performed on the waveforms during data acquisition. Off-line analysis was applied to the waveforms using the personal computer. Data acquisition A rabbit was anesthetized with pentobarbital, and then its liver was exposed through a midline incision in the abdomen. The transducer was placed on the liver using a holder. Scanning gel was used to keep the transducer in contact with the liver. Ultrasonic waveforms from the liver were monitored with the oscilloscope. Care was taken to avoid specular echoes, such as from large vessels and biliary ducts, during data acquisition. One hundred A-lines were obtained moving the transducer, and each A-line was separated by 1-2 mm. Acoustic attenuation measurement It has already been shown that acoustic attenuation in soft tissues increases nearly linearly with frequency (Pauly and Schwan 1971). Although it is reported that attenuation does not exhibit absolutely linear dependence on frequency (Parker et al. 1988), the attenuation coefficient slope still provides an excellent parameter for practical bandwidths (Ophir et al. 1984). Many studies use the attenuation coefficient slope as a parameter of attenuation, and it is convenient to use the attenuation slope for a comparison with other studies. In this study, we used an acoustic attenuation coefficient slope obtained by the spectral difference method (Kuc 1980). The attenuation computation is shown schematically in Fig. 2. Two non-overlapping data segments were obtained
Ultrasonic attenuationof liver• K. SUZUKIet al.
659
PULSE GENERATORI ---- ATTENUATOR ]
LINEAR AMP I
TRANSDUCER ::::LIVER:::: PERSONALGP'IB A/D COMPUTER: CONVERTER*---
tel
ESClL Fig. 1. Block diagram of the measurement system for the ultrasonic attenuation coefficient.
for the waveforms. Both the near data segment and far data segment were Hamming windowed and fast Fourier transformed to yield power spectra. Then, the spectral difference between two segments was computed. It was assumed that the spectral difference was linear in the frequency range of 6.3-11.7 MHz, and the slope was determined by the least squares regression. The Frequency Dependent Attenuation (FDA) coet~cient slope was obtained by dividing the slope value by twice the distance between two segments.
It is known that the transducer focusing and diffraction affect the attenuation measurement. Beam correction using measured power spectra obtained from a perfect reflector, or tissue mimicking phantom at the same distance with the data segment of the target tissue has been proposed; however, remaining range-dependent bias errors were reported even after the beam correction (Insana et al. 1983; Wilson et al. 1987). In this study, beam correction was not performed. Instead, the distance of data segments from
Reflected Signal (Digitized at a SamplingRate of 60 MHz) -Time
ROI ill (near segment)
RO1121 (far ~ment)
o
FFT 0
I-
~ de ~)/
~ , ~
Averaged Power Spectra
dB
-100
-100 0
FFT
FREQUENCY(MHz) 30
0
, FREQUENCY(MHz} 30
_
ROl II1 - RO1121
a
/]-"l'~V (V;soundvelocity) :
Spectral Difference 0
FREQUENCY(MHz)
30
Frequency Dependent AttenuationCoefficient Slope ( d B / c m - M H z )
Fig. 2. Schematic presentation of the ultrasonic attenuation coefficient calculation. Two data segments are obtained for digitized ultrasonic signals (upper part). The signals inside each region of interest (ROI) are Hamming windowed, fast-Fourier transformed and then averaged to yield the corresponding averaged power spectra (middle part). By subtracting two averaged power spectra, a spectral difference was obtained (lower part), and a slope value was estimated by means of least squares regression. The Frequency Dependent Attenuation (FDA) coefficient slope, denoted by 8, was obtained by dividing the slope value by twice the distance between two segments.
660
Ultrasound in Medicineand Biology
the transducer was fixed to avoid range-dependent attenuation estimation error. In addition, the two data segments were positioned so that diffraction effects were insignificant. The location of the two data segments was determined by a phantom experiment, in which a tissue mimicking graphite-gelatin phantom was used. First, attenuation of the phantom was measured by broadband insertion-loss technique (Ophir et al. 1984), which is a more accurate way of estimating attenuation. The transducer was positioned so that its beam was perpendicular to a glass reflector surface situated at the focus. FDA of the phantom was determined from the difference of the power spectrum of received echo signal with and without the phantom interposed between the transducer and the glass reflector. Second, FDA was also estimated from the backscattered echo signal from the phantom using the spectral difference method. Fifty A-lines, each of them separated by a sufficient distance to insure independent measurements, were acquired from the phantom by laterally moving the transducer. The length of the data segments (4.3 us, corresponding to 3.2 ram) was chosen to be short enough so that attenuation of acoustic energy within the region could be ignored. As to the separation of the two data segments, larger separation is preferable to reduce attenuation estimation error, but separation is limited by the size of the rabbit liver. In this study, separation of the two data segments was determined to be 15 #s, corresponding to 11.3 ram. FDA was computed by the spectral difference between averaged power spectrum of segment pairs with varying distance from the transducer. We found that segment pairs beginning 7.5-11 mm from the transducer were optimal, since computed FDA from these data coincided with that determined by insertion-loss techniques. In the rabbit experiment, FDA was calculated with the near data segment beginning 10 mm from the transducer and the far data segment beginning 21.3 mm from the transducer.
Assessment of the pathological state of the liver After ultrasonic data acquisition, a rabbit was sacrificed by means of an overdose injection ofpentobarbital, and then its liver was excised immediately. The liver was divided into several pieces, which were then subjected to histological examination and measurement of tissue components, namely, fat, fibrosis and water content. The histological examination was done at the location where data acquisition was performed. The histological slices were stained in hematoxylin and eosin, Sudan III and Azan-MaUory, and then examined by an experienced pathologist.
Volume18, Number 8, 1992
Evaluation offatty infiltration of the liver Liver tissue was homogenized, and total lipid content of the liver was extracted using chloroform and methanol (Folch et al. 1957). This was expressed as the total lipid content per g wet weight of the liver.
Evaluation offibrosis of the liver For estimation of the collagen content of the liver specimens, the hydroxyproline content of the liver was measured by the method of Blumenkrantz (Blumenkrantz and Asboe-Hansen 1973). Liver tissue was homogenized and hydrolyzed overnight in 6 N HC1 in a sealed tube at 118°C. The hydrolysate was evaporated and then dissolved in a buffer. A sample containing hydroxyproline was oxidized with periodic acid. Color development was performed with p-dimethylaminobenzaldehyde and the absorbance was read at 565 nm.
Measurement of the water content of the liver After the wet weight of a liver specimen had been determined, the specimen was lyophilized at - 4 0 ° C for 48 h and then weighed. The water content of the liver was determined as follows: (wet weight - dry weight) × 100 (%). wet weight
Statistical methods Values are expressed as mean +_ standard deviation (SD). The unpaired t-test was used to compare the total lipid content, hydroxyproline content, water content and FDA between the experimental groups. Simple linear regression was performed to assess the relationship between FDA and the total lipid content, between FDA and the hydroxyproline content, and between FDA and the water content. RESULTS The livers of rabbits intravenously administered fat emulsion showed mild-to-moderate fatty infiltration, with some predisposition to the pericentral area (Fig. 3). Severe fatty infiltration, i.e., such that more than 60% of the hepatocytes contained fat droplets, was not produced in this experimental model. Apparent fibrosis was not observed. The livers of rabbits given chronic carbon tetrachloride injections showed pericentral and periportal fibrosis. Rabbits subjected to chronic carbon tetrachloride administration for more than six months showed the formation of nodules surrounded with thin fibrotic septa, indicating that liver cirrhosis was induced (Fig. 4). However, apparent fatty infiltration was not observed in Sudan III
Ultrasonic attenuation of liver • K. SuzuKI et
al.
661
Fig. 3. Photomicrograph of a hematoxylin and eosin stained liver sample obtained from a rabbit intravenously administered a fat emulsion. Large fat droplets can be observed mainly in the pericentral area.
stained sections (picture not shown). Apparent necrosis was not observed in the liver fibrosis model, either. The total lipid content of the liver in the fatty liver and liver fibrosis models is given in Fig. 5. The
fatty liver model shows a significantly elevated total lipid content (95.7 + 19.2 mg/g liver) compared with the normal controls (45.0 _+ 7.3 mg/g liver, p < 0.01) and the liver fibrosis model (39.1 _+ 5.5 mg/g liver, p
0301-5629/92 $5.00 + ,00 (c) 1992 Pergamon Press Lid.
OOriginal Contribution DEPENDENCE OF ULTRASONIC ATTENUATION OF LIVER ON PATHOLOGIC FAT AND FIBROSIS: EXAMINATION WITH EXPERIMENTAL FATTY LIVER AND LIVER FIBROSIS MODELS K. S u z u ~ , *
N . HAYASHI, t Y. SASAKI, t M . KONO, t A. KASAHARA, t H . FUSAMOTO, t Y. IMAI * a n d T. KAMADA t
tFirst Department of Medicine, Osaka University Medical School, 1-1-50 Fukushima, Fukushima-ku, Osaka 553, Japan, and *Research and Development Engineering, Process Instrumentation Division, Shimadzu Corporation, Kyoto 604, Japan (Received 23 October 1991; in final form 26 May 1992) Abstract--To clarify the effect of the pathological state of the liver on ultrasonic attenuation, we produced two experimental rabbit models. The influence of fat on ultrasonic attenuation was examined using a fatty liver model without liver fibrosis, and that of fibrosis on attenuation using a liver fibrosis model without fatty infiltration. Ultrasonic data were obtained in vivo directly from the liver, and an acoustic attenuation coefficient slope was obtained by the spectral difference method. Tissue components of the liver, namely the total lipid, hydroxyproline and water contents, were measured precisely by quantitative methods. We revealed that ultrasonic attenuation depends mainly on fatty infiltration of the liver and to a lesser extent on fibrosis, but not on the water content.
Key Words: Attenuation, Ultrasound, Rabbit liver, Experimental fatty liver model, Experimental liver fibrosis model, Fat content, Collagen content, Water content.
INTRODUCTION
ture. A positive correlation between attenuation and fibrosis has been reported (Afschrift et al. 1987; Duerinckx et al. 1988; Lin et al. 1988). However, some authors found little or no elevation of attenuation with fibrosis alone (Bamber et al. 1981; Taylor et al. 1986; Wilson et al. 1987), while others reported that the attenuation value varies with the etiology of the liver cirrhosis (Maklad et al. 1984). There are the following possible reasons for such contradictory findings: i. In a human study, fibrotic livers often exhibit varying degrees of fatty infiltration, and the fatty infiltration itself elevates attenuation, making interpretation of the attenuation results complicated. 2. In most human studies, the contents of fat and fibrosis were not quantitatively measured. Liver specimens were only classified into 3 or 4 groups based upon the microscopic evaluation of the fat and fibrosis by pathologists. In addition, precise grading is not always possible for a small biopsy specimen. 3. The influence of the abdominal wall on the ultrasonic beam and acoustic attenuation has not been thoroughly investigated, and they may affect the attenuation measurement.
Conventional gray scale B-mode ultrasound of the liver is a well-established modality in the diagnosis of focal lesions within the liver. However, diffuse parenchymal diseases of the liver, such as liver cirrhosis and fatty infiltration, are difficult to differentiate on the basis of B-mode imaging alone (Gosink et al. 1979; Sandford et al. 1985). Furthermore, sonographic grading of these diseases is not precise enough to predict the severity of the pathological changes (Needleman et al. 1986). The attenuation of ultrasound in the liver has received considerable attention as a quantitative ultrasonic tissue characterization technique. A positive correlation between fatty infiltration of the liver and attenuation has been reported by many researchers, both in vitro (Bamber et al. 1981; Lin et al. 1988) and in vivo (Afschrift et al. 1987; Duerinckx et al. 1988; Taylor et al. 1986; Wilson et al. 1987). However, as to the influence of fibrosis of the liver on ultrasonic attenuation, there has been some disagreement in the literaAddress correspondence to: Norio Hayashi, M.D., First Department of Medicine, Osaka University Medical School, 1-1-50 Fukushima, Fukushima-ku, Osaka 553, Japan. 657
658
Ultrasound in Medicineand Biology
4. A negative correlation between the water content of the liver and in vitro attenuation has been suggested (Bamber et al. 1981), but the water content was not measured in other studies. 5. The results of in vitro studies may be different from those of in vivo ones and thus difficult to interpret; the vasculature of resected specimens has collapsed and thus contains less blood, of which attenuation is low. On the other hand, postmortem specimens contain microbubbles, which may elevate attenuation. In the present study, to clarify the effect of the pathological state of the liver, namely fibrosis and fat, on ultrasonic attenuation, we used two experimental animal models. One is a liver fibrosis model without fatty infiltration and the other a fatty liver model without liver fibrosis. Ultrasonic data were obtained in vivo directly from the liver through an incision in the abdominal wall, and the determined attenuation coefficient was compared with the contents of fat, fibrosis and water, measured by quantitative methods. MATERIALS AND M E T H O D S
Materials Twenty-four male albino rabbits weighing 2.02.8 kg were used. Experimental fatty liver model (n = 6). Six rabbits were intravenously administered 80 mL/kg of 10% Intrafat (Daigo Nutritive Chemicals, Osaka, Japan) daily for 6 to 9 weeks. Intrafat is a fat emulsion for intravenous infusion, containing soybean oil 10.0 g, phospholipid 1.2 g and glycerol 2.5 g per 100 mL. Experimental liverfibrosis model (n -- 12). Equal volumes of carbon tetrachloride and olive oil were emulsified, and the rabbits received weekly subcutaneous injections of 0.2 mL/kg. Subcutaneous injections of 0.2 mL of 10% phenobarbital were given simultaneously to induce fibrosis in a shorter period (McLean et al. 1969). The rabbits were treated for 3 to 8 months to induce varying degrees of fibrosis. It is well-known that the administration of carbon tetrachloride causes acute liver damage, especially necrosis and steatosis (Cameron and Karunaratne 1936). However, these changes disappear in a short period (Cameron and Karunaratne 1936; Jorgensen et al. 1974; Parker and TuthiU 1986). Therefore, in order to avoid the influence of acute liver damage on ultrasonic attenuation, liver fibrosis model rabbits were subjected to ultrasonic data acquisition 10 days after the last administration of carbon tetrachloride. Normal controls (n = 6). Six rabbits which received no injections were used as controls.
Volume18, Number 8, 1992
Instrumentation The equipment consisted of an A-mode transducer (nominal center frequency 10 MHz, 7 mm in diameter) (Shimadzu Corp., Kyoto, Japan), a pulserreceiver (Shimadzu Corp.), an A/D converter (R/ 390AD; SONY-TEKTRONIX, Tokyo, Japan), a personal computer (PC-9801 VM2; NEC, Tokyo, Japan) and an oscilloscope (T912; TEKTRONIX, Beaverton, OR, USA). A block diagram of the system is presented in Fig. 1. The focal length of the transducer was 16 mm, and the - 2 0 dB bandwidth extended from 6.3 MHz to 13.6 MHz, as was found from the spectrum of the signal reflected from a glass reflector immersed in water. A pulser-receiver was used to excite the transducer and linearly amplify the returning radio frequency (RF) waveforms. The output signals were digitized at a sampling rate of 60 MHz with a resolution of 10 bits with the A/D converter, transferred to the memory of the personal computer, and stored on a floppy disk. Nonlinear processing, such as log-amplification or sensitivity-time control gain was performed on the waveforms during data acquisition. Off-line analysis was applied to the waveforms using the personal computer. Data acquisition A rabbit was anesthetized with pentobarbital, and then its liver was exposed through a midline incision in the abdomen. The transducer was placed on the liver using a holder. Scanning gel was used to keep the transducer in contact with the liver. Ultrasonic waveforms from the liver were monitored with the oscilloscope. Care was taken to avoid specular echoes, such as from large vessels and biliary ducts, during data acquisition. One hundred A-lines were obtained moving the transducer, and each A-line was separated by 1-2 mm. Acoustic attenuation measurement It has already been shown that acoustic attenuation in soft tissues increases nearly linearly with frequency (Pauly and Schwan 1971). Although it is reported that attenuation does not exhibit absolutely linear dependence on frequency (Parker et al. 1988), the attenuation coefficient slope still provides an excellent parameter for practical bandwidths (Ophir et al. 1984). Many studies use the attenuation coefficient slope as a parameter of attenuation, and it is convenient to use the attenuation slope for a comparison with other studies. In this study, we used an acoustic attenuation coefficient slope obtained by the spectral difference method (Kuc 1980). The attenuation computation is shown schematically in Fig. 2. Two non-overlapping data segments were obtained
Ultrasonic attenuationof liver• K. SUZUKIet al.
659
PULSE GENERATORI ---- ATTENUATOR ]
LINEAR AMP I
TRANSDUCER ::::LIVER:::: PERSONALGP'IB A/D COMPUTER: CONVERTER*---
tel
ESClL Fig. 1. Block diagram of the measurement system for the ultrasonic attenuation coefficient.
for the waveforms. Both the near data segment and far data segment were Hamming windowed and fast Fourier transformed to yield power spectra. Then, the spectral difference between two segments was computed. It was assumed that the spectral difference was linear in the frequency range of 6.3-11.7 MHz, and the slope was determined by the least squares regression. The Frequency Dependent Attenuation (FDA) coet~cient slope was obtained by dividing the slope value by twice the distance between two segments.
It is known that the transducer focusing and diffraction affect the attenuation measurement. Beam correction using measured power spectra obtained from a perfect reflector, or tissue mimicking phantom at the same distance with the data segment of the target tissue has been proposed; however, remaining range-dependent bias errors were reported even after the beam correction (Insana et al. 1983; Wilson et al. 1987). In this study, beam correction was not performed. Instead, the distance of data segments from
Reflected Signal (Digitized at a SamplingRate of 60 MHz) -Time
ROI ill (near segment)
RO1121 (far ~ment)
o
FFT 0
I-
~ de ~)/
~ , ~
Averaged Power Spectra
dB
-100
-100 0
FFT
FREQUENCY(MHz) 30
0
, FREQUENCY(MHz} 30
_
ROl II1 - RO1121
a
/]-"l'~V (V;soundvelocity) :
Spectral Difference 0
FREQUENCY(MHz)
30
Frequency Dependent AttenuationCoefficient Slope ( d B / c m - M H z )
Fig. 2. Schematic presentation of the ultrasonic attenuation coefficient calculation. Two data segments are obtained for digitized ultrasonic signals (upper part). The signals inside each region of interest (ROI) are Hamming windowed, fast-Fourier transformed and then averaged to yield the corresponding averaged power spectra (middle part). By subtracting two averaged power spectra, a spectral difference was obtained (lower part), and a slope value was estimated by means of least squares regression. The Frequency Dependent Attenuation (FDA) coefficient slope, denoted by 8, was obtained by dividing the slope value by twice the distance between two segments.
660
Ultrasound in Medicineand Biology
the transducer was fixed to avoid range-dependent attenuation estimation error. In addition, the two data segments were positioned so that diffraction effects were insignificant. The location of the two data segments was determined by a phantom experiment, in which a tissue mimicking graphite-gelatin phantom was used. First, attenuation of the phantom was measured by broadband insertion-loss technique (Ophir et al. 1984), which is a more accurate way of estimating attenuation. The transducer was positioned so that its beam was perpendicular to a glass reflector surface situated at the focus. FDA of the phantom was determined from the difference of the power spectrum of received echo signal with and without the phantom interposed between the transducer and the glass reflector. Second, FDA was also estimated from the backscattered echo signal from the phantom using the spectral difference method. Fifty A-lines, each of them separated by a sufficient distance to insure independent measurements, were acquired from the phantom by laterally moving the transducer. The length of the data segments (4.3 us, corresponding to 3.2 ram) was chosen to be short enough so that attenuation of acoustic energy within the region could be ignored. As to the separation of the two data segments, larger separation is preferable to reduce attenuation estimation error, but separation is limited by the size of the rabbit liver. In this study, separation of the two data segments was determined to be 15 #s, corresponding to 11.3 ram. FDA was computed by the spectral difference between averaged power spectrum of segment pairs with varying distance from the transducer. We found that segment pairs beginning 7.5-11 mm from the transducer were optimal, since computed FDA from these data coincided with that determined by insertion-loss techniques. In the rabbit experiment, FDA was calculated with the near data segment beginning 10 mm from the transducer and the far data segment beginning 21.3 mm from the transducer.
Assessment of the pathological state of the liver After ultrasonic data acquisition, a rabbit was sacrificed by means of an overdose injection ofpentobarbital, and then its liver was excised immediately. The liver was divided into several pieces, which were then subjected to histological examination and measurement of tissue components, namely, fat, fibrosis and water content. The histological examination was done at the location where data acquisition was performed. The histological slices were stained in hematoxylin and eosin, Sudan III and Azan-MaUory, and then examined by an experienced pathologist.
Volume18, Number 8, 1992
Evaluation offatty infiltration of the liver Liver tissue was homogenized, and total lipid content of the liver was extracted using chloroform and methanol (Folch et al. 1957). This was expressed as the total lipid content per g wet weight of the liver.
Evaluation offibrosis of the liver For estimation of the collagen content of the liver specimens, the hydroxyproline content of the liver was measured by the method of Blumenkrantz (Blumenkrantz and Asboe-Hansen 1973). Liver tissue was homogenized and hydrolyzed overnight in 6 N HC1 in a sealed tube at 118°C. The hydrolysate was evaporated and then dissolved in a buffer. A sample containing hydroxyproline was oxidized with periodic acid. Color development was performed with p-dimethylaminobenzaldehyde and the absorbance was read at 565 nm.
Measurement of the water content of the liver After the wet weight of a liver specimen had been determined, the specimen was lyophilized at - 4 0 ° C for 48 h and then weighed. The water content of the liver was determined as follows: (wet weight - dry weight) × 100 (%). wet weight
Statistical methods Values are expressed as mean +_ standard deviation (SD). The unpaired t-test was used to compare the total lipid content, hydroxyproline content, water content and FDA between the experimental groups. Simple linear regression was performed to assess the relationship between FDA and the total lipid content, between FDA and the hydroxyproline content, and between FDA and the water content. RESULTS The livers of rabbits intravenously administered fat emulsion showed mild-to-moderate fatty infiltration, with some predisposition to the pericentral area (Fig. 3). Severe fatty infiltration, i.e., such that more than 60% of the hepatocytes contained fat droplets, was not produced in this experimental model. Apparent fibrosis was not observed. The livers of rabbits given chronic carbon tetrachloride injections showed pericentral and periportal fibrosis. Rabbits subjected to chronic carbon tetrachloride administration for more than six months showed the formation of nodules surrounded with thin fibrotic septa, indicating that liver cirrhosis was induced (Fig. 4). However, apparent fatty infiltration was not observed in Sudan III
Ultrasonic attenuation of liver • K. SuzuKI et
al.
661
Fig. 3. Photomicrograph of a hematoxylin and eosin stained liver sample obtained from a rabbit intravenously administered a fat emulsion. Large fat droplets can be observed mainly in the pericentral area.
stained sections (picture not shown). Apparent necrosis was not observed in the liver fibrosis model, either. The total lipid content of the liver in the fatty liver and liver fibrosis models is given in Fig. 5. The
fatty liver model shows a significantly elevated total lipid content (95.7 + 19.2 mg/g liver) compared with the normal controls (45.0 _+ 7.3 mg/g liver, p < 0.01) and the liver fibrosis model (39.1 _+ 5.5 mg/g liver, p
