- Email: [email protected]
Clinical evaluation of sonoelasticity measurement in liver using ultrasonic imaging of internal forced low-frequency vibration
Ultrasound in Med. & Biol., Vol. 26, No. 9, pp. 1455–1460, 2000 Copyright © 2001 World Federation for Ultrasound in Medicine & Biology Printed in the USA. All rights reserved 0301-5629/00/$–see front matter
PII: S0301-5629(00)00307-0
● Original Contribution CLINICAL EVALUATION OF SONOELASTICITY MEASUREMENT IN LIVER USING ULTRASONIC IMAGING OF INTERNAL FORCED LOWFREQUENCY VIBRATION MASAHIKO SANADA,* MASAAKI EBARA,* HIROYUKI FUKUDA,* MASAHARU YOSHIKAWA,* NOBUYUKI SUGIURA,* HIROMITSU SAISHO,* YOSHIKI YAMAKOSHI,‡ KEISUKE OHMURA,‡ AKIRA KOBAYASHI§ and FUKUO KONDOH† *First Department of Medicine and †Second Department of Pathology, Chiba University School of Medicine, Chiba, Japan; ‡Faculty of Engineering, Gunma University, Kiryu, Gunma, Japan; and §Faculty of Informatics, Teikyo Heisei University, Chiba, Japan (Received 17 January 2000; in final form 15 August 2000)
Abstract—The aim of this study is to evaluate liver elasticity noninvasively. We have already proposed an ultrasonic imaging system that can reconstruct vibration maps inside tissue under forced mechanical vibration. With this system, shear elastic properties of soft tissue can be evaluated as vibration velocities. Theoretically, these velocities increase with the increase of tissue elasticity. Sonoelasticity imaging was performed on 236 patients with chronic hepatitis and liver cirrhosis, and 50 healthy volunteers. The average of the velocities was 598.8 ⴞ 151.7 cm/s in healthy volunteers, 984.4 ⴞ 362.5 cm/s in chronic hepatitis and 1189.0 ⴞ 411.7 cm/s in liver cirrhosis. The average velocity of Child C group was statistically faster than those of Child A and B groups. Fibrotic rate from biopsy specimens statistically had the strongest positive correlation with velocities. With our system, the degree of liver fibrosis and function can be estimated objectively and noninvasively. © 2001 World Federation for Ultrasound in Medicine & Biology. Key Words: Chronic hepatitis, Liver cirrhosis, Liver elasticity, Palpation, Sonoelasticity imaging system, Liver fibrosis, Liver function.
change of the liver largely based on experience. But, the palpation technique is subjective, at best. Pathologic examinations by liver biopsy must be performed for a conclusive diagnosis. Furthermore, liver biopsy has become increasingly more essential for deciding the indication for interferon treatment for HCV. However, biopsy has problems, including the possibility of hemorrhage, infection and injury to the organ and usage of time and other resources that could be directed to the care of others. In spite of the limitation of manual palpation, many physicians, especially surgeons, have attached particular importance to liver elasticity (Nishizaki et al. 1995; Yamanaka et al. 1985), because they know from experience that elasticity is related to liver function and, therefore, also the functional reserve of the liver before hepatectomy. So, if new imaging systems are developed that can noninvasively evaluate the degree of elasticity of the liver, and close relationship with histologic findings and function is proven, they may play an important supportive role to conventional imagings in the diagnosis of CLD.
INTRODUCTION AND LITERATURE REVIEW Recently, with the advent of medical imagings such as ultrasonography (US), computed tomography (CT) and magnetic resonance imaging (MRI), it has become possible to diagnose liver disease correctly and noninvasively. Abnormal lesions in the liver have come to be detected more easily and accurately. However, even with the use of these morphologic imaging techniques, diagnosis of chronic liver disease (CLD) is still sometimes difficult. For example, the differentiation between chronic hepatitis (CH) and liver cirrhosis (LC), especially in alcoholic liver disease, and estimation of the clinical stage of CH are difficult even if we use these systems in combination. For such cases, as supportive diagnostic methods, we used to investigate liver elasticity by palpation and estimate the degree of fibrotic Address correspondence to: Masahiko Sanada, M.D., First Department of Medicine, Chiba University School of Medicine, Inohana 1-8-1, Chuo-ku, Chiba 260-0870, Japan. E-mail: [email protected]. ne.jp 1455
1456
Ultrasound in Medicine and Biology
Objective estimation of tissue elasticity has been reported by many investigators. For example, low-frequency wave velocities were measured for long, narrow strip samples (Nolle 1947). By this method, a vibrator is attached to one side of the strip sample to elicit mechanical vibration, and the standing wave pattern of the sample is measured with a receiving element. However, this method may provide information on the mechanical properties of formative tissues only, or give information only regarding the surface of the samples. Sekiya et al. (1979) devised a gauge that can indicate hardness of the liver in numerical values with a peritoneoscopic examination. They measured the hardness from the degree of liver deformity by depressing the liver surface with the tip of this gauge. Yamanaka et al. (1985) measured liver consistency using their devised needle during laparotomy and on the resected specimen. They introduced a needle that was connected to a pressure transducer and infusion pump into the liver tissue and measured consistency from the slope of pressure rise on a pressurevolume curve obtained at constant saline infusion. Nishizaki et al. (1995) measured hepatic consistency using the Harpenden skinfolds caliper intraoperatively. All these methods, however, were invasively performed intraoperatively or under peritoneoscopy. More recently, noninvasive techniques of measuring elasticities with conventional imaging system (US and MRI) were introduced. These are classified into correlation-based techniques and Doppler-based techniques, both depending on the basic method of the measurement of motion. The correlation-based technique makes direct estimation of tissue elasticity by calculating the strain through measurements of local tissue displacement excited internally (Dickinson and Hill 1982; Tristam et al. 1986) or externally (Ophir et al. 1991). Dickinson and Hill (1982) measured the mechanical response of tissue due to cardiac pulsation using ultrasonic waves, estimating mechanical characteristics from patterns of small movements in response to cardiac pulsation. Ophir et al. (1991) introduced elastography, which is a new ultrasonic elasticity imaging technique that produces images of the strain distribution in compliant tissue. This strain distribution, which is a result of tissue elasticity variation, is derived from ultrasonically estimated longitudinal internal motion induced by an external quasistatic compression of tissue. The elasticity information is displayed in the form of a grey-scale 2-D image called an elastogram. Muthupillai et al. (1995) described a method of spatial mapping and quantitate displacement patterns corresponding to harmonic mechanical waves with amplitudes in tissue with an MRI technique. Strain and other mechanical characteristics related to wave propa-
Volume 26, Number 9, 2000
gation can be computed from these displacement patterns. Such images could be used noninvasively to deduce important mechanical properties of the internal structure of a wide range of solid materials. The Doppler-based technique, which consists of measurement of the velocity of propagation of a lowfrequency mechanical wave, has been proposed by many investigators (Krouskop et al. 1987; Lerner et al. 1988; Tristam et al. 1986; Yamakoshi et al. 1990). Lerner and coworkers named their method sonoelasticity imaging, which incorporates tissue elasticity features into US imaging, with their goal being mainly to differentiate between healthy tissue and localized lesions (Parker et al. 1990). Our proposed method is theoretically the same as theirs, with the aim of our study being the evaluation of whole liver elasticity using our designed system. We have already presented an ultrasonic imaging system that can reconstruct vibration maps inside soft tissue under forced mechanical vibration (Yamakoshi et al. 1990). When an elastically incompressible tissue is deformed using a low-frequency source of vibration (1 kHz), the velocity of the propagating shear wave is generally proportional to the square root of its shear modulus. Sonoelasticity imaging produces vibrational maps (phase and amplitude of the internal tissue vibration) using Doppler instrumentation, from which shear-wave velocity can be calculated. Theoretically, the increase of elasticity is directly related to the square of low-wave vibration velocities. Finally, the relationship between elasticity and the histologic and biochemical findings was investigated. MATERIALS AND METHODS A schematic diagram of the sonoelasticity imaging system is shown in Fig. 1. A small mechanical vibrator with a round-surfaced, stick-like vibration head (20-mm diameter with a hemispheric shape) is used to apply low-frequency vibration (40 Hz) to the liver. Ultrasonic waves with a center frequency of 3.5 MHz, used as probing waves, are transmitted and then received by an ultrasonic transducer array, which is used for scanning by narrow ultrasonic beam in a lateral direction with a 45-mm depth and 0.75-mm width. Doppler frequency shift signals of the received ultrasonic probing waves are detected using a commercial ultrasonic pulsed Doppler imaging instrument (TOSHIBA SSA-250, Tokyo, Japan), and the detected signals are then fed into a microcomputer through a pair of high-speed A/D converters (sampling rate in depth direction: 1 MHz, sampling rate on time axis: 2 MHz, the number of bits of the A/D: 10). In the microcomputer, both internal vibration amplitude and phase were estimated for all measurement points, and these procedures were repeated by scanning
Liver elasticity, liver function and fibrosis ● M. SANADA et al.
1457
Fig. 1. Schematic diagram of signal detection and imaging system. (a) The amplitude and phase of vibration in the liver tissue are estimated from the Doppler signal (b). Low-frequency vibration velocities are measured from the observed amplitude and phase maps.
the measuring point two-dimensionally, and the amplitude and phase maps are obtained (B-mode). In addition, the propagation of wavefronts of low-frequency vibration is displayed as a motion picture (M-mode), which was produced as follows: first, positions of wavefronts at the desired instance were calculated from the phase map, and then the amplitudes of low-frequency vibration only at the positions of wavefronts were displayed to make a subimage. After repeating these procedures, the motion picture was displayed by showing these subimages sequentially. In this system, the time required to get one image is about 90 s. The M-mode type vibration map is reconstructed in real time and is easily judged as to whether it is presenting useful data or not. On this map, the vibration wave penetrating liver tissue can be seen and its velocity can be measured. To limit the influence of cardiovascular and respiratory movement, and suppress the refraction and/or reflection wave at the internal tissue surface giving rise to a standing wave, which would degrade the quality of the internal vibration map, we applied a two-step data selection method to the measurement of vibration velocity, as shown in Fig. 2. First, an M-mode type vibration phase map was obtained for the left lobe of the liver. This map shows the vibration phase advance due to the vibration wave propagation in the depth direction. To this map, we first applied the step of data selection to select an optimum section where the vibration wave phase advances linearly in the depth direction, because the vibration wave that propagates only in the forward direction without producing any standing wave shows such a linear phase advance. Then, a one-dimensional vibration phase map is obtained and, to this map, we introduced the second step of data selection to select an
Fig. 2. Two-step data selection in vibration velocity estimation. (a) Measured phase in the left lobe of the liver; (b) M-mode type vibration phase image. As the first step, an optimum section where the vibration wave phase advanced linearly in the depth direction was selected (A1, A2, A3). (c) Example of a 1-D vibration phase map in the depth direction calculated from section A2. As the second step, a section where the vibration wave phase advanced linearly in the depth direction (B2) was selected, and the velocity was calculated.
1458
Ultrasound in Medicine and Biology
optimum section where the vibration wave phase advances linearly in the depth direction. From this section, vibration velocity was calculated. From June 1993 to November 1997, at the First Department of Medicine, Chiba University School of Medicine, and its affiliated hospitals, sonoelasticity was measured in 50 healthy volunteers, 83 patients with CH, and 153 with LC with or without hepatocellular carcinoma (HCC). The average age was 65.7 years, ranging from 21 to 74 years. A total of 151 patients were seropositive for antibody to hepatitis C virus (HCV), 12 for hepatitis B surface antigen (HbsAg), and 3 for autoimmune antibodies (ANA and/or AMA). All patients were checked for levels of glutamic oxaloacetic transaminase (GOT), glutamic pyruvic transaminase (GPT), total bilirubin (T-Bil), total protein (TP), albumin (Alb), total cholesterol (T-cho), cholinesterase (Che), zinc-sulfate turbidity test (ZTT), tymol turbidity test (TTT) and platelet count (Plt), and were classified into groups A, B and C by Child’s classification (Child and Turcottee 1964). Indocyanine Green retention rate at 15 min after injection of 0.5 mg/kg (ICGR15, %) was examined in 155 patients and the serum level of type 4 collagen was measured in 216 patients. Biopsy specimens were obtained with a 14 or 17 gauge biopsy needle (Hakko, Tokyo, Japan; Sonoguide CHIBA biopsy needle TypeC2) in 63 patients (18 CH and 45 LC). In all patients, in a supine position, velocity data were measured in the left lobe of the liver. Patients with HCC in the left lobe of the liver were excluded. The velocity was measured five times for each patient during breath-holding for 10 s. The average of these five velocities in each patient was adopted as one value. To select data objectively, we used four criteria for each optimum section: minimum values for length of time, mean absolute error, section depth and average vibration amplitude. In 63 patients who underwent liver biopsy by Quick-cut needle, the fibrotic rate (FR; ratio of rate of fibrosis to whole liver) of their specimens was measured by Masson trichrome staining and was compared with the velocities statistically. FR was calculated by a commercially available imaging processing package (Adobe Photoshop and NIH Image Ver. 1.56: free software developed by Dr. Wayne Rasband et al. at the National Institutes of Health). The relationship between FR of each specimen and the low-wave vibration velocities was then determined.
Volume 26, Number 9, 2000
Table 1. Repeatability of velocities and interobserver variation Case Healthy control
CH
LC
1 2 3 4 5 av(cm/s) SD variant coefficient 1 2 3 4 5 av(cm/s) SD variant coefficient 1 2 3 4 5 av(cm/s) SD variant coefficient
Dr. 1
Dr. 2
Dr. 3
486 520 516 532 512 513.2 17.0 3.31 798 830 780 836 810 810.8 23.0 2.84 1254 1302 1286 1320 1060 1244.4 105.9 4.89
540 534 484 504 512 514.8 22.8 4.43 824 840 780 768 838 810.0 33.7 4.16 980 1230 1206 1020 1340 1155.2 151.0 7.22
542 496 506 542 552 527.6 24.9 4.72 756 768 760 810 780 774.8 21.7 2.8 1208 980 1342 982 1034 1109.2 160.1 7.32
The velocity was measured 5 times for each case by three different doctors.
three adults (healthy control, CH patient and LC patient), five times for each. Results are shown in Table 1. No statistical differences were observed among the three doctors (Kruskal–Wallis test). The variant coefficient of velocities among the three cases was different statistically, and that of LC was different from healthy control and CH statistically significant (Fisher’s protected least significant difference; Fisher’s PLSD). Velocities of the three groups; healthy volunteers, patients with CH and with LC The average velocities in the liver were 631.7 ⫾ 147.9 cm/s in healthy volunteers, 983.1 ⫾ 373.4 cm/s in
RESULTS Repeatability of velocities and interobserver variation To check the repeatability of velocities and interobserver variation, experimental measurements were performed. Three doctors measured velocities in the same
Fig. 3. The results of average velocities of the three groups. * ⫽ p ⬍ 0.01 (Fisher’s PLSD); LC ⫽ liver cirrhosis; CH ⫽ chronic hepatitis.
Liver elasticity, liver function and fibrosis ● M. SANADA et al.
1459
Table 2. Relationship between velocities and histologic and biochemical findings
FR Serum level of type 4 collagen Plt ICG R15 Che
Correlation coefficient*
p value
0.488 0.369 ⫺0.263 0.297 ⫺0.195
⬍ 0.001 ⬍ 0.001 ⬍ 0.001 ⬍ 0.001 ⬍ 0.001
* Pearson’s correlation coefficient.
Fig. 4. The relation between liver function (Child’s classification) and FR. * ⫽ p ⬍ 0.01 (Fisher’s PLSD).
patients with CH and 1200.5 ⫾ 415.8 cm/s in those with LC (Fig. 3). Statistically significant differences were observed among all groups (Kruskal–Wallis test). Velocities increased with progression of CLD. Velocities and FR of the three groups according to Child’s classification The average velocities in the liver were 986.6 ⫾ 397.8 cm/s in Child’s A group (n ⫽ 156), 1188.6 ⫾ 386.7 cm/s in Child’s B group (n ⫽ 55), and 1287.1 ⫾ 400.5 cm/s in Child’s C group (n ⫽ 25) (Figs. 4 and 5). Statistically significant differences were observed among all groups (Kruskal–Wallis test). In addition, the differences between groups A and C, and between B and C were statistically significant (Fisher’s PLSD). The average FR (n ⫽ 63) was 14.4 ⫾ 8.7% in Child’s A (n ⫽ 41), 19.5 ⫾ 11.9% in Child’s B (n ⫽ 17) and 39.5 ⫾ 15.6% in Child’s C (n ⫽ 5). Statistically significant differences were observed among all three groups (Kruskal–Wallis test). FR increased statistically with worse liver function. Correlation between velocities and histologic and biochemical findings The relationships between velocities and histologic and biochemical findings are presented in Table 2 (Pear-
Fig. 5. The results of average velocities of the three groups classified by Child’s classification. * ⫽ p ⬍ 0.01 (Fisher’s PLSD).
son’s correlation coefficient). FR was most significantly correlated with velocities (r ⫽ 0.488, p ⬍ 0.001, n ⫽ 63). The serum level of type 4 collagen was also statistically correlated with velocities (r ⫽ 0.369, p ⬍ 0.001, n ⫽ 216). Using multivariate analysis (stepwise regression), velocities were defined by the following equation: square of velocities (m/s) ⫽ 59.01 ⫹ 469.34 ⫻ FR (adjusted r2 ⫽ 0.333, p value ⬍ 0.01). DISCUSSION Sonoelasticity imaging has been the subject of many studies. Parker et al. (1990) reported their objective of sonoelasticity imaging to be the differentiation between mass and surrounding tissue. Our method was theoretically the same as theirs, but our goal was to evaluate the elasticities of liver in various diffuse liver diseases. Our method used a conventional Doppler imaging system to visualize the internal vibration of the liver in real-time under forced low-frequency vibration (40 Hz). From the amplitude and phase maps, the propagation of vibration was observed on a monitor and the velocities apparently related to elasticity were measured. All measurements could be noninvasively performed within 10 s. However, the right lobe of the liver did not lend itself well to this procedure because the costal bones prevented the penetration of effective vibration power. Fortunately, CLD is a diffuse process, so the left lobe can be used to measure elasticities of patients. Interobserver variation of velocities was negligible, but the variant coefficient of LC was statistically different from healthy controls and CH patients, indicating that the repeatability of LC velocities is uneven. This result shows either the possibility that unbalanced liver elasticities reflect histologic construction of LC, or that this was a technical problem of our system. A recent study of sonoelasticity technique (Catheline et al. 1999) reported biases that decrease the measurement accuracy of tissue velocities: reflected waves created at boundaries, nonnegligible size of the piston source which causes diffraction effects, and the influence of a low-frequency compressional wave. Further investigations are also being performed to improve our system.
1460
Ultrasound in Medicine and Biology
Low-frequency vibration velocities of patients with LC were statistically faster than those of healthy controls and CH patients. This result coincides with our clinical experience by palpation, namely that, with progression of CLD, there is elastic change in the liver. It was proved statistically that both FR and velocities reflected liver function defined by Child’s classification, and multivariate analysis between velocities and histologic and biochemical findings showed that velocities were most significantly correlated with FR. In conclusion, by measuring liver elasticity as the velocity of low-frequency vibration by our system, the degree of static liver fibrosis and liver function can be estimated. In time, liver biopsy is sure to be replaced by noninvasive methods, and then our sonoelasticity system will have a useful role. More recently, Desmet et al. (1994) reported new guidelines for a revised system of classifying CH. Further recent studies have suggested that carcinogenesis by HCV infection takes place as CLD progresses to LC (Takano et al. 1995). New classifications of CLD are, therefore, placing more and more emphasis on the histology of fibrosis. A characteristic of Asians with primary HCC is their having liver cirrhosis (Foster and Berman 1977). The high rate of postoperative intrahepatic recurrence of HCC might be due to undetected and unremoved tumorous tissue, so, to improve the postoperative prognosis, a wider field of hepatectomy would be called for if the functional capacity of the remaining liver permits (Utsunomiya et al. 1992). However, modern techniques for estimating the functional reserve of the liver are not sufficient, and when the experienced surgeon faces the difficult choices when operating on HCC, they often rely on the findings of palpation before hepatectomy (Nishizaki et al. 1995). Preoperative measurement of velocities of the liver would have a good possibility of showing them new supportive clinical information about liver function. Of course, although fiber is the main factor for defining liver elasticity, we think it is impossible to explain liver elasticity by only one component factor. Factors affecting elasticity of the liver are not only the degree of fibrosis, but also many components such as hepatocytes, fat cells, capsules, blood flow, vessels, bile duct and so on. Although we did not examine the possible correlation between elasticity and hepatic blood circulation, Yamanaka et al. (1985) reported that there is no significant difference of consistency between in vivo and ex vivo liver, concluding that blood flow played no important role in determining liver elasticity. Furthermore, we measured the velocities of a patient with severe fulminant hepatitis. Postmortem examination showed the extent of liver fibrosis and fibrotic rate to be high, al-
Volume 26, Number 9, 2000
though hepatocytes had drastically decreased in number. However, the velocities of this patient were much lower than expected from the degree of fibrosis (unpublished data). From this, it can be speculated that expansive regeneration of liver cells might have a particular effect on elasticity. In the future, we will examine in greater detail what elasticity might indicate at the histologic and biochemical levels, and continue to attempt to increase the accuracy of the measurement of tissue velocities. REFERENCES Catheline S, Wu F, Fink M. A solution to diffraction biases in sonoelasticity: The acoustic impulse technique. J Acoust Soc Am 1999; 105(5):2941–2950. Child CG, Turcotte JG. Surgery and portal hypertension. In: Child CG, ed. The liver and portal hypertension. Philadelphia: WB Saunders, 1964:50. Desmet V, Gerber M, Hoofnagle JH, Manns M, Scheuer P. Classification of chronic hepatitis; Diagnosis, grading and staging. Hepatology 1994;19:1513–1520. Dickinson RJ, Hill CR. Measurement of soft tissue motion using correlation between A-scans. Ultrasound Med Biol 1982;8(3):263– 271. Foster JH, Berman MM. Solid liver tumors. Philadelphia. WB Saunders, 1977:9 –27, 62–104. Krouskop TA, Dougherty DR, Vinson FS. A pulsed Doppler ultrasonic system for making noninvasive measurements of the mechanical properties of soft tissue. J Rehab Res Dev 1987;24(2):1– 8. Lerner RM, Parker KJ, Holen J, Gramiak R, Waar RC. Sono-elasticity: Medical elasticity images derived from ultrasound signals in mechanically vibrated targets. Acoust Imaging 1988;16:317–327. Muthupillai R, Lomas DJ, Rossman PJ, Greenleaf JF, Manduca A, Ehman RL. Magnetic resonance elastography by direct visualization of propagating acoustic strain waves. Science 1995;269(5232): 1854 –1857. Nishizaki T, Matsumata T, Kamakura T, Adachi E, Sugimachi K. Significance of intraoperative measurement of liver consistency prior to hepatic resection. Hepato-Gastroenterol 1995;42:5– 8. Nolle AW. Acoustic determination of the physical constants of rubberlike materials. J Acoust Soc Am 1947;19;194 –201. Ophir J, Cespedes I, Ponnekanti H, Yazdi Y, Li X. Elastography: A quantitative method for imaging the elasticitiy of biological tissues. Ultrasonic Imaging 1991;13(2):111–134. Parker KJ, Huang SR, Musulin RA, Lerner RM. Tissue response to mechanical vibrations for sonoelasticity imaging. Ultrasound Med Biol 1990;16(3):241–246. Sekiya C, Yasuyuki Y, Takahashi A, Numazaki A, Namiki M. New device of gauge for measuring hardness of liver. Gastroenterol Endosc 1979;21(8):974 –981. Tristam M, Barbosa DC, Cosgrove DO, Nassiri DK, Bamber JC, Hill CR. Ultrasonic study of in vivo kinetic characteristics of human tissues. Ultrasound Med Biol 1986;12(12):927–937. Takano S, Yokosuka O, Imazeki F, Tagawa M, Omata M. Incidence of hepatocellular carcinoma in chronic hepatitis B and C: A prospective study of 251 patients. Hepatology. 1995;21(3):650 – 655. Utsunomiya T, Matsumata T, Adachi E, Honda H, Sugimachi K. Limitations of current preoperative liver imaging techniques for intrahepatic metastatic nodules of hepatocellular carcinoma. Hepatology 1992;16:694 –701. Yamakoshi Y, Sato J, Sato T. Ultrasonic imaging of internal vibration of soft tissue under forced vibration. IEEE Trans Ultrason Ferroelec Frequency Control 1990;37(2):45–53. Yamanaka N, Okamoto E, Toyosaka A, Ohashi S, Tanaka N. Consistency of human liver. J Surg Res 1985;39:192–198.
PII: S0301-5629(00)00307-0
● Original Contribution CLINICAL EVALUATION OF SONOELASTICITY MEASUREMENT IN LIVER USING ULTRASONIC IMAGING OF INTERNAL FORCED LOWFREQUENCY VIBRATION MASAHIKO SANADA,* MASAAKI EBARA,* HIROYUKI FUKUDA,* MASAHARU YOSHIKAWA,* NOBUYUKI SUGIURA,* HIROMITSU SAISHO,* YOSHIKI YAMAKOSHI,‡ KEISUKE OHMURA,‡ AKIRA KOBAYASHI§ and FUKUO KONDOH† *First Department of Medicine and †Second Department of Pathology, Chiba University School of Medicine, Chiba, Japan; ‡Faculty of Engineering, Gunma University, Kiryu, Gunma, Japan; and §Faculty of Informatics, Teikyo Heisei University, Chiba, Japan (Received 17 January 2000; in final form 15 August 2000)
Abstract—The aim of this study is to evaluate liver elasticity noninvasively. We have already proposed an ultrasonic imaging system that can reconstruct vibration maps inside tissue under forced mechanical vibration. With this system, shear elastic properties of soft tissue can be evaluated as vibration velocities. Theoretically, these velocities increase with the increase of tissue elasticity. Sonoelasticity imaging was performed on 236 patients with chronic hepatitis and liver cirrhosis, and 50 healthy volunteers. The average of the velocities was 598.8 ⴞ 151.7 cm/s in healthy volunteers, 984.4 ⴞ 362.5 cm/s in chronic hepatitis and 1189.0 ⴞ 411.7 cm/s in liver cirrhosis. The average velocity of Child C group was statistically faster than those of Child A and B groups. Fibrotic rate from biopsy specimens statistically had the strongest positive correlation with velocities. With our system, the degree of liver fibrosis and function can be estimated objectively and noninvasively. © 2001 World Federation for Ultrasound in Medicine & Biology. Key Words: Chronic hepatitis, Liver cirrhosis, Liver elasticity, Palpation, Sonoelasticity imaging system, Liver fibrosis, Liver function.
change of the liver largely based on experience. But, the palpation technique is subjective, at best. Pathologic examinations by liver biopsy must be performed for a conclusive diagnosis. Furthermore, liver biopsy has become increasingly more essential for deciding the indication for interferon treatment for HCV. However, biopsy has problems, including the possibility of hemorrhage, infection and injury to the organ and usage of time and other resources that could be directed to the care of others. In spite of the limitation of manual palpation, many physicians, especially surgeons, have attached particular importance to liver elasticity (Nishizaki et al. 1995; Yamanaka et al. 1985), because they know from experience that elasticity is related to liver function and, therefore, also the functional reserve of the liver before hepatectomy. So, if new imaging systems are developed that can noninvasively evaluate the degree of elasticity of the liver, and close relationship with histologic findings and function is proven, they may play an important supportive role to conventional imagings in the diagnosis of CLD.
INTRODUCTION AND LITERATURE REVIEW Recently, with the advent of medical imagings such as ultrasonography (US), computed tomography (CT) and magnetic resonance imaging (MRI), it has become possible to diagnose liver disease correctly and noninvasively. Abnormal lesions in the liver have come to be detected more easily and accurately. However, even with the use of these morphologic imaging techniques, diagnosis of chronic liver disease (CLD) is still sometimes difficult. For example, the differentiation between chronic hepatitis (CH) and liver cirrhosis (LC), especially in alcoholic liver disease, and estimation of the clinical stage of CH are difficult even if we use these systems in combination. For such cases, as supportive diagnostic methods, we used to investigate liver elasticity by palpation and estimate the degree of fibrotic Address correspondence to: Masahiko Sanada, M.D., First Department of Medicine, Chiba University School of Medicine, Inohana 1-8-1, Chuo-ku, Chiba 260-0870, Japan. E-mail: [email protected]. ne.jp 1455
1456
Ultrasound in Medicine and Biology
Objective estimation of tissue elasticity has been reported by many investigators. For example, low-frequency wave velocities were measured for long, narrow strip samples (Nolle 1947). By this method, a vibrator is attached to one side of the strip sample to elicit mechanical vibration, and the standing wave pattern of the sample is measured with a receiving element. However, this method may provide information on the mechanical properties of formative tissues only, or give information only regarding the surface of the samples. Sekiya et al. (1979) devised a gauge that can indicate hardness of the liver in numerical values with a peritoneoscopic examination. They measured the hardness from the degree of liver deformity by depressing the liver surface with the tip of this gauge. Yamanaka et al. (1985) measured liver consistency using their devised needle during laparotomy and on the resected specimen. They introduced a needle that was connected to a pressure transducer and infusion pump into the liver tissue and measured consistency from the slope of pressure rise on a pressurevolume curve obtained at constant saline infusion. Nishizaki et al. (1995) measured hepatic consistency using the Harpenden skinfolds caliper intraoperatively. All these methods, however, were invasively performed intraoperatively or under peritoneoscopy. More recently, noninvasive techniques of measuring elasticities with conventional imaging system (US and MRI) were introduced. These are classified into correlation-based techniques and Doppler-based techniques, both depending on the basic method of the measurement of motion. The correlation-based technique makes direct estimation of tissue elasticity by calculating the strain through measurements of local tissue displacement excited internally (Dickinson and Hill 1982; Tristam et al. 1986) or externally (Ophir et al. 1991). Dickinson and Hill (1982) measured the mechanical response of tissue due to cardiac pulsation using ultrasonic waves, estimating mechanical characteristics from patterns of small movements in response to cardiac pulsation. Ophir et al. (1991) introduced elastography, which is a new ultrasonic elasticity imaging technique that produces images of the strain distribution in compliant tissue. This strain distribution, which is a result of tissue elasticity variation, is derived from ultrasonically estimated longitudinal internal motion induced by an external quasistatic compression of tissue. The elasticity information is displayed in the form of a grey-scale 2-D image called an elastogram. Muthupillai et al. (1995) described a method of spatial mapping and quantitate displacement patterns corresponding to harmonic mechanical waves with amplitudes in tissue with an MRI technique. Strain and other mechanical characteristics related to wave propa-
Volume 26, Number 9, 2000
gation can be computed from these displacement patterns. Such images could be used noninvasively to deduce important mechanical properties of the internal structure of a wide range of solid materials. The Doppler-based technique, which consists of measurement of the velocity of propagation of a lowfrequency mechanical wave, has been proposed by many investigators (Krouskop et al. 1987; Lerner et al. 1988; Tristam et al. 1986; Yamakoshi et al. 1990). Lerner and coworkers named their method sonoelasticity imaging, which incorporates tissue elasticity features into US imaging, with their goal being mainly to differentiate between healthy tissue and localized lesions (Parker et al. 1990). Our proposed method is theoretically the same as theirs, with the aim of our study being the evaluation of whole liver elasticity using our designed system. We have already presented an ultrasonic imaging system that can reconstruct vibration maps inside soft tissue under forced mechanical vibration (Yamakoshi et al. 1990). When an elastically incompressible tissue is deformed using a low-frequency source of vibration (1 kHz), the velocity of the propagating shear wave is generally proportional to the square root of its shear modulus. Sonoelasticity imaging produces vibrational maps (phase and amplitude of the internal tissue vibration) using Doppler instrumentation, from which shear-wave velocity can be calculated. Theoretically, the increase of elasticity is directly related to the square of low-wave vibration velocities. Finally, the relationship between elasticity and the histologic and biochemical findings was investigated. MATERIALS AND METHODS A schematic diagram of the sonoelasticity imaging system is shown in Fig. 1. A small mechanical vibrator with a round-surfaced, stick-like vibration head (20-mm diameter with a hemispheric shape) is used to apply low-frequency vibration (40 Hz) to the liver. Ultrasonic waves with a center frequency of 3.5 MHz, used as probing waves, are transmitted and then received by an ultrasonic transducer array, which is used for scanning by narrow ultrasonic beam in a lateral direction with a 45-mm depth and 0.75-mm width. Doppler frequency shift signals of the received ultrasonic probing waves are detected using a commercial ultrasonic pulsed Doppler imaging instrument (TOSHIBA SSA-250, Tokyo, Japan), and the detected signals are then fed into a microcomputer through a pair of high-speed A/D converters (sampling rate in depth direction: 1 MHz, sampling rate on time axis: 2 MHz, the number of bits of the A/D: 10). In the microcomputer, both internal vibration amplitude and phase were estimated for all measurement points, and these procedures were repeated by scanning
Liver elasticity, liver function and fibrosis ● M. SANADA et al.
1457
Fig. 1. Schematic diagram of signal detection and imaging system. (a) The amplitude and phase of vibration in the liver tissue are estimated from the Doppler signal (b). Low-frequency vibration velocities are measured from the observed amplitude and phase maps.
the measuring point two-dimensionally, and the amplitude and phase maps are obtained (B-mode). In addition, the propagation of wavefronts of low-frequency vibration is displayed as a motion picture (M-mode), which was produced as follows: first, positions of wavefronts at the desired instance were calculated from the phase map, and then the amplitudes of low-frequency vibration only at the positions of wavefronts were displayed to make a subimage. After repeating these procedures, the motion picture was displayed by showing these subimages sequentially. In this system, the time required to get one image is about 90 s. The M-mode type vibration map is reconstructed in real time and is easily judged as to whether it is presenting useful data or not. On this map, the vibration wave penetrating liver tissue can be seen and its velocity can be measured. To limit the influence of cardiovascular and respiratory movement, and suppress the refraction and/or reflection wave at the internal tissue surface giving rise to a standing wave, which would degrade the quality of the internal vibration map, we applied a two-step data selection method to the measurement of vibration velocity, as shown in Fig. 2. First, an M-mode type vibration phase map was obtained for the left lobe of the liver. This map shows the vibration phase advance due to the vibration wave propagation in the depth direction. To this map, we first applied the step of data selection to select an optimum section where the vibration wave phase advances linearly in the depth direction, because the vibration wave that propagates only in the forward direction without producing any standing wave shows such a linear phase advance. Then, a one-dimensional vibration phase map is obtained and, to this map, we introduced the second step of data selection to select an
Fig. 2. Two-step data selection in vibration velocity estimation. (a) Measured phase in the left lobe of the liver; (b) M-mode type vibration phase image. As the first step, an optimum section where the vibration wave phase advanced linearly in the depth direction was selected (A1, A2, A3). (c) Example of a 1-D vibration phase map in the depth direction calculated from section A2. As the second step, a section where the vibration wave phase advanced linearly in the depth direction (B2) was selected, and the velocity was calculated.
1458
Ultrasound in Medicine and Biology
optimum section where the vibration wave phase advances linearly in the depth direction. From this section, vibration velocity was calculated. From June 1993 to November 1997, at the First Department of Medicine, Chiba University School of Medicine, and its affiliated hospitals, sonoelasticity was measured in 50 healthy volunteers, 83 patients with CH, and 153 with LC with or without hepatocellular carcinoma (HCC). The average age was 65.7 years, ranging from 21 to 74 years. A total of 151 patients were seropositive for antibody to hepatitis C virus (HCV), 12 for hepatitis B surface antigen (HbsAg), and 3 for autoimmune antibodies (ANA and/or AMA). All patients were checked for levels of glutamic oxaloacetic transaminase (GOT), glutamic pyruvic transaminase (GPT), total bilirubin (T-Bil), total protein (TP), albumin (Alb), total cholesterol (T-cho), cholinesterase (Che), zinc-sulfate turbidity test (ZTT), tymol turbidity test (TTT) and platelet count (Plt), and were classified into groups A, B and C by Child’s classification (Child and Turcottee 1964). Indocyanine Green retention rate at 15 min after injection of 0.5 mg/kg (ICGR15, %) was examined in 155 patients and the serum level of type 4 collagen was measured in 216 patients. Biopsy specimens were obtained with a 14 or 17 gauge biopsy needle (Hakko, Tokyo, Japan; Sonoguide CHIBA biopsy needle TypeC2) in 63 patients (18 CH and 45 LC). In all patients, in a supine position, velocity data were measured in the left lobe of the liver. Patients with HCC in the left lobe of the liver were excluded. The velocity was measured five times for each patient during breath-holding for 10 s. The average of these five velocities in each patient was adopted as one value. To select data objectively, we used four criteria for each optimum section: minimum values for length of time, mean absolute error, section depth and average vibration amplitude. In 63 patients who underwent liver biopsy by Quick-cut needle, the fibrotic rate (FR; ratio of rate of fibrosis to whole liver) of their specimens was measured by Masson trichrome staining and was compared with the velocities statistically. FR was calculated by a commercially available imaging processing package (Adobe Photoshop and NIH Image Ver. 1.56: free software developed by Dr. Wayne Rasband et al. at the National Institutes of Health). The relationship between FR of each specimen and the low-wave vibration velocities was then determined.
Volume 26, Number 9, 2000
Table 1. Repeatability of velocities and interobserver variation Case Healthy control
CH
LC
1 2 3 4 5 av(cm/s) SD variant coefficient 1 2 3 4 5 av(cm/s) SD variant coefficient 1 2 3 4 5 av(cm/s) SD variant coefficient
Dr. 1
Dr. 2
Dr. 3
486 520 516 532 512 513.2 17.0 3.31 798 830 780 836 810 810.8 23.0 2.84 1254 1302 1286 1320 1060 1244.4 105.9 4.89
540 534 484 504 512 514.8 22.8 4.43 824 840 780 768 838 810.0 33.7 4.16 980 1230 1206 1020 1340 1155.2 151.0 7.22
542 496 506 542 552 527.6 24.9 4.72 756 768 760 810 780 774.8 21.7 2.8 1208 980 1342 982 1034 1109.2 160.1 7.32
The velocity was measured 5 times for each case by three different doctors.
three adults (healthy control, CH patient and LC patient), five times for each. Results are shown in Table 1. No statistical differences were observed among the three doctors (Kruskal–Wallis test). The variant coefficient of velocities among the three cases was different statistically, and that of LC was different from healthy control and CH statistically significant (Fisher’s protected least significant difference; Fisher’s PLSD). Velocities of the three groups; healthy volunteers, patients with CH and with LC The average velocities in the liver were 631.7 ⫾ 147.9 cm/s in healthy volunteers, 983.1 ⫾ 373.4 cm/s in
RESULTS Repeatability of velocities and interobserver variation To check the repeatability of velocities and interobserver variation, experimental measurements were performed. Three doctors measured velocities in the same
Fig. 3. The results of average velocities of the three groups. * ⫽ p ⬍ 0.01 (Fisher’s PLSD); LC ⫽ liver cirrhosis; CH ⫽ chronic hepatitis.
Liver elasticity, liver function and fibrosis ● M. SANADA et al.
1459
Table 2. Relationship between velocities and histologic and biochemical findings
FR Serum level of type 4 collagen Plt ICG R15 Che
Correlation coefficient*
p value
0.488 0.369 ⫺0.263 0.297 ⫺0.195
⬍ 0.001 ⬍ 0.001 ⬍ 0.001 ⬍ 0.001 ⬍ 0.001
* Pearson’s correlation coefficient.
Fig. 4. The relation between liver function (Child’s classification) and FR. * ⫽ p ⬍ 0.01 (Fisher’s PLSD).
patients with CH and 1200.5 ⫾ 415.8 cm/s in those with LC (Fig. 3). Statistically significant differences were observed among all groups (Kruskal–Wallis test). Velocities increased with progression of CLD. Velocities and FR of the three groups according to Child’s classification The average velocities in the liver were 986.6 ⫾ 397.8 cm/s in Child’s A group (n ⫽ 156), 1188.6 ⫾ 386.7 cm/s in Child’s B group (n ⫽ 55), and 1287.1 ⫾ 400.5 cm/s in Child’s C group (n ⫽ 25) (Figs. 4 and 5). Statistically significant differences were observed among all groups (Kruskal–Wallis test). In addition, the differences between groups A and C, and between B and C were statistically significant (Fisher’s PLSD). The average FR (n ⫽ 63) was 14.4 ⫾ 8.7% in Child’s A (n ⫽ 41), 19.5 ⫾ 11.9% in Child’s B (n ⫽ 17) and 39.5 ⫾ 15.6% in Child’s C (n ⫽ 5). Statistically significant differences were observed among all three groups (Kruskal–Wallis test). FR increased statistically with worse liver function. Correlation between velocities and histologic and biochemical findings The relationships between velocities and histologic and biochemical findings are presented in Table 2 (Pear-
Fig. 5. The results of average velocities of the three groups classified by Child’s classification. * ⫽ p ⬍ 0.01 (Fisher’s PLSD).
son’s correlation coefficient). FR was most significantly correlated with velocities (r ⫽ 0.488, p ⬍ 0.001, n ⫽ 63). The serum level of type 4 collagen was also statistically correlated with velocities (r ⫽ 0.369, p ⬍ 0.001, n ⫽ 216). Using multivariate analysis (stepwise regression), velocities were defined by the following equation: square of velocities (m/s) ⫽ 59.01 ⫹ 469.34 ⫻ FR (adjusted r2 ⫽ 0.333, p value ⬍ 0.01). DISCUSSION Sonoelasticity imaging has been the subject of many studies. Parker et al. (1990) reported their objective of sonoelasticity imaging to be the differentiation between mass and surrounding tissue. Our method was theoretically the same as theirs, but our goal was to evaluate the elasticities of liver in various diffuse liver diseases. Our method used a conventional Doppler imaging system to visualize the internal vibration of the liver in real-time under forced low-frequency vibration (40 Hz). From the amplitude and phase maps, the propagation of vibration was observed on a monitor and the velocities apparently related to elasticity were measured. All measurements could be noninvasively performed within 10 s. However, the right lobe of the liver did not lend itself well to this procedure because the costal bones prevented the penetration of effective vibration power. Fortunately, CLD is a diffuse process, so the left lobe can be used to measure elasticities of patients. Interobserver variation of velocities was negligible, but the variant coefficient of LC was statistically different from healthy controls and CH patients, indicating that the repeatability of LC velocities is uneven. This result shows either the possibility that unbalanced liver elasticities reflect histologic construction of LC, or that this was a technical problem of our system. A recent study of sonoelasticity technique (Catheline et al. 1999) reported biases that decrease the measurement accuracy of tissue velocities: reflected waves created at boundaries, nonnegligible size of the piston source which causes diffraction effects, and the influence of a low-frequency compressional wave. Further investigations are also being performed to improve our system.
1460
Ultrasound in Medicine and Biology
Low-frequency vibration velocities of patients with LC were statistically faster than those of healthy controls and CH patients. This result coincides with our clinical experience by palpation, namely that, with progression of CLD, there is elastic change in the liver. It was proved statistically that both FR and velocities reflected liver function defined by Child’s classification, and multivariate analysis between velocities and histologic and biochemical findings showed that velocities were most significantly correlated with FR. In conclusion, by measuring liver elasticity as the velocity of low-frequency vibration by our system, the degree of static liver fibrosis and liver function can be estimated. In time, liver biopsy is sure to be replaced by noninvasive methods, and then our sonoelasticity system will have a useful role. More recently, Desmet et al. (1994) reported new guidelines for a revised system of classifying CH. Further recent studies have suggested that carcinogenesis by HCV infection takes place as CLD progresses to LC (Takano et al. 1995). New classifications of CLD are, therefore, placing more and more emphasis on the histology of fibrosis. A characteristic of Asians with primary HCC is their having liver cirrhosis (Foster and Berman 1977). The high rate of postoperative intrahepatic recurrence of HCC might be due to undetected and unremoved tumorous tissue, so, to improve the postoperative prognosis, a wider field of hepatectomy would be called for if the functional capacity of the remaining liver permits (Utsunomiya et al. 1992). However, modern techniques for estimating the functional reserve of the liver are not sufficient, and when the experienced surgeon faces the difficult choices when operating on HCC, they often rely on the findings of palpation before hepatectomy (Nishizaki et al. 1995). Preoperative measurement of velocities of the liver would have a good possibility of showing them new supportive clinical information about liver function. Of course, although fiber is the main factor for defining liver elasticity, we think it is impossible to explain liver elasticity by only one component factor. Factors affecting elasticity of the liver are not only the degree of fibrosis, but also many components such as hepatocytes, fat cells, capsules, blood flow, vessels, bile duct and so on. Although we did not examine the possible correlation between elasticity and hepatic blood circulation, Yamanaka et al. (1985) reported that there is no significant difference of consistency between in vivo and ex vivo liver, concluding that blood flow played no important role in determining liver elasticity. Furthermore, we measured the velocities of a patient with severe fulminant hepatitis. Postmortem examination showed the extent of liver fibrosis and fibrotic rate to be high, al-
Volume 26, Number 9, 2000
though hepatocytes had drastically decreased in number. However, the velocities of this patient were much lower than expected from the degree of fibrosis (unpublished data). From this, it can be speculated that expansive regeneration of liver cells might have a particular effect on elasticity. In the future, we will examine in greater detail what elasticity might indicate at the histologic and biochemical levels, and continue to attempt to increase the accuracy of the measurement of tissue velocities. REFERENCES Catheline S, Wu F, Fink M. A solution to diffraction biases in sonoelasticity: The acoustic impulse technique. J Acoust Soc Am 1999; 105(5):2941–2950. Child CG, Turcotte JG. Surgery and portal hypertension. In: Child CG, ed. The liver and portal hypertension. Philadelphia: WB Saunders, 1964:50. Desmet V, Gerber M, Hoofnagle JH, Manns M, Scheuer P. Classification of chronic hepatitis; Diagnosis, grading and staging. Hepatology 1994;19:1513–1520. Dickinson RJ, Hill CR. Measurement of soft tissue motion using correlation between A-scans. Ultrasound Med Biol 1982;8(3):263– 271. Foster JH, Berman MM. Solid liver tumors. Philadelphia. WB Saunders, 1977:9 –27, 62–104. Krouskop TA, Dougherty DR, Vinson FS. A pulsed Doppler ultrasonic system for making noninvasive measurements of the mechanical properties of soft tissue. J Rehab Res Dev 1987;24(2):1– 8. Lerner RM, Parker KJ, Holen J, Gramiak R, Waar RC. Sono-elasticity: Medical elasticity images derived from ultrasound signals in mechanically vibrated targets. Acoust Imaging 1988;16:317–327. Muthupillai R, Lomas DJ, Rossman PJ, Greenleaf JF, Manduca A, Ehman RL. Magnetic resonance elastography by direct visualization of propagating acoustic strain waves. Science 1995;269(5232): 1854 –1857. Nishizaki T, Matsumata T, Kamakura T, Adachi E, Sugimachi K. Significance of intraoperative measurement of liver consistency prior to hepatic resection. Hepato-Gastroenterol 1995;42:5– 8. Nolle AW. Acoustic determination of the physical constants of rubberlike materials. J Acoust Soc Am 1947;19;194 –201. Ophir J, Cespedes I, Ponnekanti H, Yazdi Y, Li X. Elastography: A quantitative method for imaging the elasticitiy of biological tissues. Ultrasonic Imaging 1991;13(2):111–134. Parker KJ, Huang SR, Musulin RA, Lerner RM. Tissue response to mechanical vibrations for sonoelasticity imaging. Ultrasound Med Biol 1990;16(3):241–246. Sekiya C, Yasuyuki Y, Takahashi A, Numazaki A, Namiki M. New device of gauge for measuring hardness of liver. Gastroenterol Endosc 1979;21(8):974 –981. Tristam M, Barbosa DC, Cosgrove DO, Nassiri DK, Bamber JC, Hill CR. Ultrasonic study of in vivo kinetic characteristics of human tissues. Ultrasound Med Biol 1986;12(12):927–937. Takano S, Yokosuka O, Imazeki F, Tagawa M, Omata M. Incidence of hepatocellular carcinoma in chronic hepatitis B and C: A prospective study of 251 patients. Hepatology. 1995;21(3):650 – 655. Utsunomiya T, Matsumata T, Adachi E, Honda H, Sugimachi K. Limitations of current preoperative liver imaging techniques for intrahepatic metastatic nodules of hepatocellular carcinoma. Hepatology 1992;16:694 –701. Yamakoshi Y, Sato J, Sato T. Ultrasonic imaging of internal vibration of soft tissue under forced vibration. IEEE Trans Ultrason Ferroelec Frequency Control 1990;37(2):45–53. Yamanaka N, Okamoto E, Toyosaka A, Ohashi S, Tanaka N. Consistency of human liver. J Surg Res 1985;39:192–198.