Relation of Ultrasonic Backscatter and Acoustic Propagation Properties to Myofibrillar Length and Myocardial Thickness
Background Ultrasonic backscatter demonstrates a cardiac cycle–dependent modulation. The exact mechanism of the modulation is under debate. The objective of the present study was to test the hypothesis that a change in size and configuration of myofilaments from systole to diastole alters acoustic propagation properties and backscatter.
Methods and Results In vivo measurements were made of integrated backscatter at 5 MHz (IBR5), followed by in vitro measurements of ultrasonic attenuation, speed, and heterogeneity index using a scanning laser acoustic microscope at 100 MHz. Studies were performed in canine hearts (16) arrested in systole (8) with calcium chloride or arrested in diastole (8) with potassium chloride. Sarcomere length was measured with a calibrated eyepiece on a Ziess microscope. Wall thickness was measured with calipers. The attenuation coefficient of 220±34 dB/cm during systole was significantly higher than the coefficient of 189±24 dB/cm during diastole (P<.01); the IBR5 of −44.7±1.2 dB during systole was significantly greater than the IBR5 of −47.0±1.0 dB during diastole (P<.01); the ultrasonic speed of 1591±11 m/s during systole was higher than the speed of 1575±4.2 m/s during diastole (P<.01); and the heterogeneity index of 7.4±1.8 m/s during systole was significantly lower than the index of 9.0±2.0 m/s during diastole (P<.02). The sarcomere length of 1.804±0.142 μm during diastole was significantly higher than the length of 1.075±0.177 μm during systole (P<.01). Wall thickness was significantly greater during systole than during diastole (20±3 versus 9±3 mm, P<.01).
Conclusions Ultrasonic backscatter and propagation properties are directly related to sarcomere length and myocardial thickness and may be responsible for cardiac cycle–dependent variation in backscatter.
Ultrasonic backscatter demonstrates a modulation across the cardiac cycle,1 and this modulation closely follows changes in myocardial contractility.2 This modulation in backscatter is abolished during myocardial ischemia.1 2 The exact mechanism of modulation in ultrasonic backscatter with cardiac cycle is under debate. Wickline and associates3 argue that the change in acoustic impedance, which in turn is related to changes in passive elastance with contraction, is responsible for the modulation of ultrasonic backscatter. In studies of isolated papillary muscle, Wear and coworkers4 have shown that modulation occurs in isotonically contracting muscles and not in isometrically contracting muscle. In vitro studies by O’Donnell and others5 provide frequency dependence data that support Shung and Reid’s suggestion6 that backscatter arises from small cylindrical scatters. Based on the shape of the backscatter frequency-response and time-varying statistics, the in vivo work from our laboratory argues for a geometric model in which multiple small structures, such as myofilaments, are responsible for myocardial backscatter.7 The present study was designed to test the hypothesis that cyclic variation in backscatter is dependent on active contraction, ie, changes in size and configuration of myofilaments from systole to diastole alter acoustic propagation properties and backscatter. The acoustic propagation properties of myocardium during systole and diastole were measured with a scanning laser acoustic microscope (SLAM), which operates at a frequency of 100 MHz. The SLAM provides acoustic and optical images plus attenuation coefficient, speed of sound, and heterogeneity index data. Conventional imaging frequencies of 3 to 7 MHz were studied using a well-described integrated backscatter measurement (IBR5, or integrated backscatter Rayleigh spectral intensity measurement, referred to 5 MHz).
Adult mongrel dogs of either sex weighing 15 to 25 kg were anesthetized with phenobarbital sodium (30 mg/kg IV) supplemented with 5 mg/kg per hour IV and ventilated with Harvard Apparatus respirator model 607. Atelectasis was prevented by maintaining an end-respiratory pressure of 5 to 7 cm H2O. A thoracotomy was performed in the left fifth intercostal space, and the lungs were gently retracted. The heart was suspended in a pericardial cradle.
IBR5 was measured by coupling the transducer to the epicardial surface of the left ventricle from the left anterior descending coronary artery (LAD) and left circumflex coronary artery (LCx) territories. A research prototype instrument was used to measure IBR5; the details have been published,2 8 and a brief description follows. After in vivo measurements were made, hearts were arrested during systole (8) and during diastole (8). Systolic arrest was produced by rapid intravenous infusion of calcium chloride (20 to 40 μmol/L). Diastolic arrest was produced by rapid intravenous injection of potassium chloride. Myocardial wall thickness was measured with calipers. Approximately square, transmural tissue samples (5×5 mm multiplied by wall thickness) for electron microscopic and SLAM examination were obtained from the regions of LAD and LCx, with the surfaces of the sample oriented along the axis of the ventricle as shown in Fig 1A⇓ and 1B⇓. The SLAM samples were mounted with their endocardial surfaces against a circular cork with Ames Tissue-Tek O.C.T. (optimal cutting temperature; a polyvinyl alcohol, benzalkonium chloride, and polyethyleneglycol gel—an embedding medium for frozen tissue specimens). The base and apical sides of the samples were carefully marked on the corks. The samples were frozen in liquid nitrogen, placed into Ziploc bags, and shipped on dry ice from the Medical College of Wisconsin to the Bioacoustics Research Laboratory at the University of Illinois for storage in a Revco freezer at −70°C until analysis.
Immediately before SLAM analysis, each tissue specimen was removed from the freezer, and the corks were mounted on an object disk of a Lipshaw cryostat (maintained at −35°C). Approximately 50 to 100 μm of the upper surface of the specimen was sectioned and discarded to provide a flat, even surface for subsequent sections. Consequently, all subsequent sections were from planes parallel to the endocardial and epicardial surfaces and of known orientation. This procedure ensured that all specimens remained frozen between the times of initial freezing and the SLAM analysis. Previous studies have shown that rapid freezing in liquid nitrogen has no effect on the ultrasonic impedance and other mechanical properties of several different tissues.9 10 Specimens of known thicknesses (50, 75, 100, 200, and 400 μm) were mounted on a 1-in2 sheet of mylar (25 μm thick) with albumin fixative and marked so that the depth within the wall and the orientation relative to the axis of the ventricle were known.
Tissue specimens (5×5 mm) were also obtained for measurement of sarcomere length. Sarcomere length was determined on samples fixed in glutaraldehyde. Fixed muscle tissue was minced in iso-osmotic glycerol (0.28 mol/L) and then briefly (≈5 seconds) ground with a Polytron homogenizer. The resulting suspension of fixed myofibril fragments was allowed to settle for 5 minutes to remove air bubbles, and then a drop of the suspension was placed on a slide and covered with a coverslip. Sarcomere length was measured with a calibrated eyepiece micrometer on a Ziess microscope at a total magnification of 2000× (10× eyepiece, 2× multiplier, and 100× lens). Sarcomere length was determined on myofibrils with four to eight clearly defined sarcomeres by measuring the total length and dividing by the number of sarcomeres included. Only myofibrils with clearly defined Z-lines, A bands, and I bands were measured or counted. A minimum of 20 myofibrils with a total of at least 100 sarcomeres were measured for each sample.
Ultrasonic Backscatter Instrumentation and Data Analysis
In vivo ultrasonic backscatter was measured using a research instrument used in previous studies.2 8 The myocardium was examined with a 6-mm-diameter, unfocused circular transducer that was fitted with a water-filled fixture that maintained the transducer approximately 18 mm from the epicardium. Echoes were range-gated on the midmyocardium, thus avoiding the specular echoes at the endocardial and epicardial surfaces. The prototype obtained echo signals with a 0.5-μsec resolution and a uniform bandwidth extending from 4.0 to 6.5 MHz. A digitized ECG waveform was used to synchronize the data acquisition with the QRS complex. Backscatter echoes were performed at evenly spaced intervals over 16 cardiac cycles, and data were stored for later analysis.
According to previous methods,2 8 the echoes were adjusted using calibration data7 and further adjusted for the frequency effects of tissue absorption, diffraction, and power-law scattering, resulting in measurements of absolute reflectivity integrated over the frequency spectrum from 4.0 to 6.5 MHz. This is called the IBR5 and represents an absolute measure of the backscatter in square centimeters of reflecting surface per cubic centimeter of volume.11 The IBR5 was constructed at selected ranges and expressed in decibels.
The magnitude and phase of the cardiac amplitude modulation of the IBR5 measurements were determined by independently averaging the IBR5 signals from six evenly spaced phases of the cardiac cycle (relative to the midpoint of the QRS complex). The Fourier coefficient for the amplitude modulation (of the fundamental of the cardiac cycle) was determined by mean square fit to the six data points. The ratio of the IBR5 at the positive and negative peaks of the Fourier sine wave (expressed in decibels) was used as an index of the amplitude modulation and called the Fourier coefficient of amplitude modulation. Also, the phase of the Fourier coefficient was measured from the negative peak of the Fourier sine wave to the QRS complex and expressed as a percent of the cardiac cycle.
Scanning Laser Acoustic Microscope
A SLAM (Sonomicroscope 100, Sonoscan, Inc), operating at an ultrasonic frequency of 100 MHz, was used to determine the specimen’s attenuation coefficient, propagation speed, and heterogeneity index. Operational details of the SLAM have been published.11 12 13 14 15 16 Briefly, the specimen is placed on a sonically activated fused silica stage along with a thin layer of normal saline and covered with a semireflective coverslip. A focused scanning laser beam probe detects a dynamic ripple (surface displacement) by reflection from the coverslip’s lower surface. The reflected laser beam is processed to yield the acoustic image, from which the attenuation coefficient is determined, and the interference image, from which the speed and heterogeneity index are determined. The laser-transmitted beam also is transmitted through the specimen to yield an optical image. All three images are displayed in real time on a standard television monitor representing a specimen area measuring approximately 3 mm horizontally by 2 mm vertically (typically 100×). A frame grabber is used to digitize the video signal of the acoustic and interference images17 and performs frame averaging (up to 256 frames), convolutions (for filtering), and histograms (for dynamic range determination). In addition, attenuation coefficient, speed, and heterogeneity index estimations are performed with software using the digitized video image.
An insertion loss procedure is used to estimate the attenuation coefficient.12 15 Attenuation coefficient is a measure of energy loss per unit distance (decibels per meter). In principle, this procedure compares the received signal amplitude of the specimen of known thickness in the sound path with that of the reference medium, normal saline. A subimage area of approximately 400×250 μm (96 pixels horizontally by 32 pixels vertically) in the acoustic image is used. The signals received from the subimage area are digitized to yield an average amplitude value (V) expressed in decibels. A minimum of five V values are recorded for normal saline. The specimen is then moved into the subimage area, and a minimum of three V values are recorded at each of five specimen locations. An insertion loss (IL) value, given in decibels, is estimated from the following:
where Vr is the average V value recorded from normal saline, and Vs represents individual V values from the specimen. This process yields five IL values for each specimen thickness. Five specimen thicknesses ranging from 50 to 400 μm are used, and the slope of IL versus specimen thickness (using a linear least-squares fit with all 25 IL thickness values) yields the attenuation coefficient.
The spatial frequency domain technique is used to yield speed from the interference image.13 15 Speed represents propagation of ultrasound (in meters per second). The 2×3-mm field of view contains approximately 39 vertical interference lines equally spaced by about 85 μm. The propagation speed is estimated by the horizontal shift of the fringe lines relative to where the fringe lines would be without the specimen, ie, relative to the reference (with known speed value) medium. The specimen is positioned so that each interference line includes both the reference fluid above and below the specimen. From the digitized interference image, basically each interference line is digitized and processed, yielding a vertical (relative to the image orientation) speed profile. For each sample, five specimen thicknesses are analyzed, with each yielding a mean speed value from the speed profile region within the specimen. The individual mean speed values are then averaged to yield the ultrasonic speed of that sample.
The heterogeneity index represents an estimate of the spatial distribution of the ultrasonic speed.11 An advantage of determining the ultrasonic speed using the speed profile is that many speed values, each representative of a 4×8-μm specimen region, are obtained. Not only is the mean value estimated, yielding speed, but also the standard deviation is estimated. This standard deviation of the spatial speed distribution is an estimate of the heterogeneity index of the specimen at the specified thickness.
The values of the heterogeneity index are an inverse function of thickness. Therefore, the separate heterogeneity index values are reported, at a thickness of 100 μm as representative of all thicknesses.
Estimates of attenuation coefficient and speed from the SLAM have been assessed for solutions of known acoustic properties and duplicate samples of skin and healing wound tissue.15 With a homogeneous medium, the accuracy (proximity to the true value) was ±2.9% and the precision (reproducibility of successive independent measurements) was ±0.4% for speed; for attenuation coefficient, accuracy and precision were ±12% and ±15%, respectively. With heterogenous samples of normal canine skin and wound tissue, the speed and attenuation coefficient precision were ±1.7% and ±16%, respectively.
All SLAM values and sarcomere length values are reported as mean±SD. Comparison between systolic and diastolic measurements were performed with a paired t test, and changes were considered significant when values were P<.05.
Ultrasonic attenuation coefficient, speed, heterogeneity index, IBR5, sarcomere length, and wall thickness were measured in 16 dogs. A total of 32 data sets were obtained for each parameter—16 during systole and 16 during diastole.
Fig 2A⇓ shows the attenuation coefficient and IBR5 during systole and during diastole. Attenuation coefficient of 220±34 dB/cm during systole was significantly higher than the coefficient of 189±24 dB/cm during diastole (P<.01). IBR5 of −47±1.0 dB during systole was significantly weaker than the IBR5 of −44.7±1.2 dB (P<.01) during diastole.
Ultrasonic speed of 1591±11 m/s during systole was about 1% higher than the speed of 1575±4.2 m/s during diastole; the difference is small but statistically significant (P<.01) (Fig 2B⇑). The heterogeneity index of 7.4±1.8 m/s during systole was significantly lower than the index of 9.0±2.0 m/s during diastole (P<.02) at a specimen thickness of 100 μm (Fig 2B⇑). This gradient in heterogeneity index between systole and diastole was observed for all specimen thicknesses.
Sarcomere length and wall thickness are shown in Fig 3⇓. As expected, the sarcomere length of 1.804±0.142 μm during diastole was significantly higher than the length of 1.075±0.177 μm during systole, thus confirming the change in sarcomere size with contraction. Wall thickness was significantly greater during systole than during diastole (20±3 versus 9±3 mm, P<.01; Fig 3⇓).
Regional variability in acoustic propagation properties was determined by comparing specimens from territories supplied by LAD and LCx. No significant regional differences were observed for any of the parameters.
The above data show that cardiac cycle–dependent changes in myofilament length and thickness are associated with modulation in acoustic propagation properties and ultrasonic backscatter. Attenuation coefficient and ultrasonic speed are higher in systole, whereas heterogeneity index and IBR5 are lower in systole.
Myocardial contraction causes several alterations in the cardiac muscle: increased thickness and tissue density, changes in collagen matrix, change in size, and shape of ultrasonic scatterer within the myofilament. Any of these changes alone or in combination can alter attenuation, speed, heterogeneity index, and backscatter. This report suggests a geometrical model for the cardiac cycle dependence of backscatter seen at frequencies of 3 to 7 MHz.
A compilation of the attenuation coefficient, speed, and protein and water concentrations in various tissues indicates that the attenuation coefficient and speed in tissue increase as the protein concentration increases and water decreases.18 19 Collagen also plays an important role in the acoustic properties of tissues. The higher the collagen content, the higher are the attenuation, speed, and backscatter. During systole, the muscle is expected to be most dense. In systole, myocardium has higher density, the collagen and protein per unit mass of the muscle are higher, and thus attenuation coefficient is greater. The attenuation measured by SLAM is roughly 15% higher (in terms of decibels) in systole. Mass density also plays a role in the speed of propagation. At the SLAM frequency of 100 MHz, the alteration of the ultrasonic speed is on the order of 1%, with the higher speed occurring in systole. Because speed is reciprocal to the square root of the product of mass density and compressibility, the product of density and compressibility must be about 2% higher in systole than in diastole. It appears that in systole there is an increase in mass density and a decrease in compressibility. Systolic decrease in compressibility may arise from alteration in collagen cross-linkage or non-Hookien behavior of the elastic elements as contraction distorts the elastic components of the myocardium.3
It is possible that cyclic variation in backscatter could be related to myocardial tissue attenuation and possible cyclic changes in attenuation.3 20 In the present study, attenuation was significantly greater in systole, although it was not as high as predicted by Wickline and associates.3 However, when one calculates the cyclic backscatter that arises from the measured attenuation, it is insufficient to account for the cyclic backscatter observed at frequencies of 3 to 7 MHz, as suggested by Wickline et al.3 Consequently, variation in attenuation alone is not sufficient basis for cyclic backscatter. Thus, it does not appear that either the attenuation or propagation speed properties directly address the mechanisms of cyclic backscatter.
Wickline and associates3 suggested that cyclic variation in backscatter was related to myofibrillar elastic characteristics. They speculated that cyclic backscatter changes arise from an impedance mismatch between intracellular and extracellular domains, with the intracellular domain associated with the series elastic components and the extracellular domain associated with parallel elastic components, of a Maxwell-type model for the myocardium. The change in acoustic impedance of elastic components due to differential stretching and compression should give rise to the observed cyclic variation. However, Wear and associates4 reported little cyclic variation in backscatter for isometrically contracting canine muscle, where the series elastic components should have been cyclically stretched while parallel components were held fixed. By the model of Wickline et al, this should have produced a cyclic variation in the intracellular and extracellular acoustic impedance. We suspect that the basis of cyclic backscatter may not be due to this mechanism.
Mottley and Miller21 delineated an angular dependence relative to the muscle fiber orientation. The present study demonstrates a change in size and thickness of myofilaments with contraction but does not directly address the question of anisotropy of the scatterers. The samples were sectioned in planes parallel to the endocardial surface (see Fig 1⇑), which ensured that muscle fiber orientation was always in the plane of the sections. The in vivo data were similarly obtained with the ultrasound traversing the myocardial wall at a normal incidence. Thus, the sound waves in both IBR5 and SLAM analyses were always perpendicular to the fibers, removing this variable from the investigation.
The SLAM operates at 100 MHz and produces images with an acoustic resolution of approximately 29 μm. The paired acoustic and optical images were evaluated to determine if the backscatter, which is seen in clinical cardiac images at frequencies from 2.5 to 7.5 MHz, could be ascribed to structures resolved by the SLAM at 100 MHz. It has been reported by our laboratory that the through transmission acoustic images of the myocardium were of very low contrast, which delineated fiber margins and general structure, and correlated well with these features in the paired optical images.11 Consequently, it is concluded that the basis of myocardial backscatter is not resolvable at the 29-μm level and that the backscatter must arise from structures that are much smaller. In addition, previous work at our laboratory shows that the backscatter exhibits almost Rayleigh scattering, indicating that the scattering elements are much smaller than a wavelength (one-fourth wavelength at 5 MHz is 77 μm).7 This study suggests that the basis of ultrasonic backscatter may be at the myofibrillar level.
Heterogeneity index and ultrasonic backscatter are weaker during systole than during diastole. The heterogeneity index is a through transmission marker for backscatter, with the variation in transmission indicating that energy is being removed from the through transmitted ultrasound by the backscatter. The relatively higher index during diastole indicates more backscatter than during systole, which is consistent with the cyclic modulation of myocardial backscatter in IBR5 data. Consequently, the SLAM data support the conclusion that the basis of the backscatter is at the ultrastructural level. Moreover, SLAM acoustic images previously reported by these researchers11 clearly indicate lineal structures following the muscle fibers of the paired optical image. The small size of scatterers is consistent with previously reported data5 7 concerning the spectral shape of the backscattering spectra, as discussed above. Moreover, the scattering spectra obtained independently in systole and diastole maintain the same shape, with the diastolic spectrum being stronger everywhere than the systolic spectrum.7 This indicates that in cyclic backscatter, the elemental scatterers alter the product of their scattering size and density.
The basis of backscatter is the impedance mismatch at the surface of a scattering element. The surface area of each small scattering element determines its reflectivity to the ultrasound. The total backscatter is the summation of the reflection intensities of all the elements in a given volume (thus, IBR5 has the dimensions of area per volume). Data from our laboratory7 indicate that it is the product of scatterer area and scatterer density that is altered by the contraction. It is proposed that a “geometric” model be used to account for cyclic backscatter, based primarily on changes in the density of scatterers with contraction.
Available knowledge indicates that clinical backscatter arises from small scatterers that are linearly arranged, which most likely are the collagenous connective tissues plus other heavy molecules, such as the mycin and actin fibers of the active contracting elements. Each of these small scattering objects presents a given scattering surface area to the ultrasound, and the summation of the intensity of individual reflections makes up the overall backscatter. The scattering elements are present in the microstructure with other nonreflecting material, such as mitochondria (≈40% of myocardium), intracellular and extracellular fluids, and so on. Systolic contraction results in the partial compression and partial reshaping of the noncontracting constituents of the ultrastructure, and this contributes to the gross thickening of the myocardial wall. Consequently, in the contracted state, there is greater mean free distance separating connective tissues and myofilaments. Increased separation constitutes a reduction in the number of reflecting elements in a given volume, which is the scatterer density. One index of this density is wall thickening with contraction, which, in the present study, was a thickening of a factor of ≈2. It would be expected that the density of scatterers would be reduced approximately by this factor of 2. The observed change in the intensity of IBR5 is a factor of 1.7 (ie, 2.3 dB), which is consistent with the change of scatterer density. The lack of cyclic variation in backscatter with isometric contraction is also consistent with this hypothesis.
In future studies, it may be possible to accurately correlate cyclic changes in tissue histology with the proposed geometric model of cyclic backscatter. In particular, variation in the mean free spacing of suspected scattering components with contraction is of interest, as well as myofibril dimensional and density changes.
This work was supported in part by National Institutes of Health grant HL-39014. We thank Kathi Annous and Sandra Wilton for technical assistance and Cindy Ketola for preparing the manuscript.
- Received May 12, 1994.
- Accepted July 27, 1994.
- Copyright © 1995 by American Heart Association
Sagar KB, Rhyne TL, Warltier DC, Pelc LR, Wann LS. Intramyocardial variability in integrated backscatter: effects of coronary occlusion and reperfusion. Circulation. 1987;75:436-442.
Shung KK, Reid JM. Ultrasonic scattering from tissue. 1977 Ultrasonics Symposium Proceedings, IEEE catalog no. 77CH1264-ISU, 230, 1977.
Sagar KB, Pelc LR, Rhyne TL, Wann LS, Warltier DC. Influence of heart rate, preload, afterload, and inotropic state on myocardial ultrasonic backscatter. Circulation. 1988;77:478-486.
Marangore RD, Glasser AA, Mosi JS, Brody AS, Beckwith TH, Walter GR, White WL. Effects of storage and handling techniques on skin tissue properties. Ann N Y Acad Sci. 1966;136:439-454.
Gellski JV, Shung KK. Studies on acoustic impedance of major bovine blood vessel walls. J Acoust Soc Am. 1983;71:467-470.
Sagar KB, Agemura DH, O’Brien WD Jr, Pelc LR, Rhyne TL, Wann LS, Komorowski RA, Warltier DC. Quantitative ultrasonic assessment of normal and ischaemic myocardium with an acoustic microscope: relationship to integrated backscatter. Cardiovasc Res. 1990;24:447-455.
Tervola KMU, Foster SC, O’Brien WD Jr. Attenuation coefficient measurement technique at 100 MHz with the scanning laser acoustic microscope. IEEE Trans Sonics Ultrasonics. 1985;SU-32:259-265.
Tervola KMU, O’Brien WD Jr. Spatial frequency domain technique: an approach for analyzing the scanning laser acoustic microscope interferogram images. IEEE Trans Sonics Ultrasonics. 1985;SU-32:544-554.
Embree PM, Tervola KMU, Foster SG, O’Brien WD Jr. Spatial distribution of the speed of sound in biological materials with the scanning laser acoustic microscope. IEEE Trans Sonics Ultrasonics. 1985;SU-32:341-350.
Nicozisin DD. An automated image data acquisition and analysis system for the scanning laser acoustic microscope. Master of Science Thesis, University of Illinois, 1989.
Dunn F. Ultrasonic properties of biological media. In: Melner R, Rosenfield E, Cobet U, eds. Ultrasound Interactions in Biology and Medicine. New York, NY: Plenum Publishing Corp; 1983:1-6.
O’Brien WD Jr. The role of collagen in determining ultrasonic propagation properties in tissues. In: Kessler LW, ed. Acoustical Holography. New York, NY: I. Plenum Press; 1977:37-50.