calibrated quantitative ultrasound imaging of skeletal muscle using backscatter analysis

ABSTRACT: We evaluated the ability of an ultrasound method, which cancharacterize cardiac muscle pathology and has reliability across differentimaging systems, to obtain calibrated quantitative estimates of backscatterof skeletal muscle. Our procedure utilized a tissue-mimicking phantom toestablish a linear relationship between ultrasound grayscale and backscatterlevels. We studied skeletal muscles of 82 adults: 45 controls and 37 patientswith hereditary myopathies. We found that skeletal muscle ultrasound back-scatter levels varied with probe orientation, age, and muscle contraction andpathology. Reliability was greater with the probe in longitudinal comparedwith transverse planes. Backscatter levels were higher in those �40 years ofage, in muscle extension than flexion, and in myopathic patients thancontrols. Calibrated measurements of muscle backscatter have sensitivityand specificity in identifying and reliably measuring levels of skeletal musclepathology.

Muscle Nerve 38: 893–898, 2008

CALIBRATED QUANTITATIVE ULTRASOUND IMAGING OFSKELETAL MUSCLE USING BACKSCATTER ANALYSIS

CRAIG M. ZAIDMAN, MD,1 MARK R. HOLLAND, PhD,2 CHRISTIAN C. ANDERSON, MA,2 and ALAN PESTRONK, MD1

1 Department of Neurology, Washington University School of Medicine, 660 South Euclid Avenue,Box 8111, St. Louis, Missouri 63110, USA

2 Department of Physics, Washington University, St. Louis, Missouri, USA

Accepted 30 March 2008

Ultrasound imaging has been used as a method forvisualizing skeletal muscle pathology, including neu-ropathic and myopathic changes.5,7,14,20,21,25 Its ad-vantages are that it is rapid, non-invasive, painless,and inexpensive. Analyses of ultrasound images ofmuscle in clinical practice most commonly focus onqualitative features.1,2,5,6 Quantitative measures ofthe degree of neuromuscular pathology, such asgrayscale image (texture) analysis, have also beenemployed.7,14,15,20–22 However, most quantitativemethods have limited reproducibility across differ-ent imaging systems, as they analyze images obtainedwithout calibration.

We report a quantitative ultrasound method thatmeasures levels of skeletal muscle pathology by usingbackscatter analysis of calibrated images from con-trols and patients with neuromuscular disorders.The level of ultrasound backscatter represents theamount of signal reflected by tissue back to theultrasound transducer. Backscatter reflection inmost soft tissues is approximately 1–100 parts from

100,000 of the original signal. The amount of soundbackscattered to the transducer is related to theinherent structural (e.g., fiber size, geometry, andorientation) and material properties (e.g., collagencontent) of tissue.4,9,17,18,23 Measurement of back-scattered signals can provide reliable estimates of theintrinsic tissue properties even among differentultrasound machines and set-ups if reference dataand specific imaging system configurations areknown.10,12,24 Backscatter levels in the myocardiumvary with the state of contraction, muscle fiber ori-entation, and cardiac pathology.3,19 In this study wemeasured the effects of muscle contraction andprobe orientation on backscatter in normal skeletalmuscle and compared backscatter levels in controlsand those with myopathies.

METHODS

The protocol was approved by our institutional re-view board. All subjects provided written consent.We enrolled 45 normal men and women (ages18–80 years). All denied neuromuscular symptomsand a history of disorders involving the studiedlimbs. All controls had images obtained with theultrasound probe oriented longitudinally to the mus-cle. Twenty-four control subjects also had imagesobtained with the probe in a transverse position. We

Abbreviations: cMB-calibrated muscle backscatter; GSL, grayscale levelKey words: backscatter; myopathy; quantitative; skeletal muscle; ultrasoundCorrespondence to: C. Zaidman; e-mail: [email protected]

© 2008 Wiley Periodicals, Inc.Published online 14 June 2008 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/mus.21052

Calibrated Quantitative Ultrasound MUSCLE & NERVE July 2008 893

also enrolled 37 patients (ages 18–81 years) withhereditary myopathies, including limb-girdle or fa-cioscapulohumeral-like dystrophies (n � 14), dystro-phinopathies (n � 10), acid maltase deficiency (n �8), and mitochondrial disorders (n � 5). Thirty-three of 37 subjects had antigravity strength orgreater in the elbow flexors. Images were obtainedin the extended arm in 37 and in the flexed arm in33 patients with the probe in a longitudinal position.Analyses of images were performed using NIH Im-age J (version 1.37) image-analysis software.

Ultrasound examinations were performed usinga Philips HD11XE imaging system with a L12-5 lin-ear-array probe. This system was configured to pro-vide an approximately linear relationship betweenbackscattered ultrasound intensity (expressed indecibels) and the displayed grayscale value over theuseful dynamic range. Postprocessing settings were:fusion 5; 232 dB; compression 1; postprocessing mapA; smoothing 1; and persistence 0. Other imagemodification options were turned off. The calibra-tion procedure used a tissue-mimicking phantom toestablish a known relationship between ultrasoundgrayscale and backscatter values. Serial images of thephantom were obtained as the overall two-dimen-sional (2D) gain was changed in known decibel in-crements over the available range. The knownchanges in overall gain setting provided the relation-ship between the change in mean grayscale level(GSL) and the change in measured signal strength,expressed in decibels. A conversion factor relating achange in grayscale value to the equivalent changein decibels was found by determining the slope ofthe best-fit line in the “linear range” region. Meangrayscale values (representing the level of backscat-ter) could then be converted to ultrasound backscat-ter values, expressed in decibels (Fig. 1).10,11,13,24

For data acquisition from human subjects, im-ages were acquired in the fundamental imagingmode with the ultrasound system configured as justdescribed. The image depth setting was usually set to5 cm. It was occasionally changed to accommodatevery large arms. A single transmit focus position wasset approximately 90% deep into the image relativeto the surface of the transducer and was kept con-stant for all muscle and phantom images. All otherultrasound settings were kept constant. The overall(2D) receive gain and time-gain compensation con-trols were optimized to provide relatively strongbackscatter (unsaturated, mid-level grayscale values)for muscle image acquisition. Time-gain compensa-tion settings were uniform and did not vary withdepth. One investigator (C.M.Z.) obtained all ultra-sound images. Subjects were seated with the entire

arm anteriorly extended, supinated, and supportedby a pillow on a table at approximately mid-thoracicheight. Images were acquired with either the elbowflexed against gravity to 90° or extended. Subjectswere instructed to relax their hand and avoid activemuscle contraction except as needed to maintainelbow flexion against gravity. The middle of theprobe was placed at a region corresponding to thearea of maximal enlargement over the elbow flexors(biceps brachii and brachialis), two thirds of the wayfrom the lateral tip of the acromion to the lateralepicondyle of the humerus. The probe position wasadjusted to yield an image of the elbow flexor mus-cles with the greatest superficial bone reflection andangled to achieve the brightest and narrowest bonereflection. Three images each were obtained in ex-tended and flexed positions. Optimal imaging wasobtained by using minimal compression of tissueand generous amounts of ultrasound coupling gel.To evaluate intrarater reliability, 20 control subjectsunderwent two sets of three measurements in eacharm position and probe orientation. The arm andprobe were repositioned between sets of measure-ments.

Eight-bit bitmap images were digitally exportedfrom the ultrasound system for analysis. The entiredepth of the elbow flexor muscles, visualized be-tween the subcutaneous fat layer and bone, was usedas a region-of-interest. This resulted in a wide anddeep area for analysis of a grayscale histogram andmean grayscale value for each image. In the longi-tudinal images, the region-of-interest excluded ap-

FIGURE 1. Relationship between changes in the backscattersignal levels (in dB) and changes in grayscale level (GSL) for theimaging configuration system used in this study. Images of skel-etal muscle were obtained exhibiting a mean brightness in themost linear range. The conversion factor leading to a change ingrayscale value to the equivalent change in decibels was foundby determining the slope of the best-fit line in this linear region.

894 Calibrated Quantitative Ultrasound MUSCLE & NERVE July 2008

proximately 0.5 cm of the lateral margins to avoidartifact. For each subject and position, the meangrayscale values were averaged across three separateimages. Calibrated muscle backscatter values (cMBs)were calculated by dividing the average grayscalelevel by the slope of the best-fit line relating GSL tobackscatter (GSL/dB) and subtracting the backscat-ter of a reference phantom. Calibration of the ultra-sound system with the configuration employed re-sulted in a slope of 6.6 GSL/dB (Fig. 1). A referencemean backscatter value of 13.7 dB was obtained byimaging a grayscale ultrasound phantom (Model047; Computerized Imaging Reference Systems,Inc.) using the same imaging system configurationparameters used during muscle image acquisition.The region-of-interest in the reference phantom wasof similar depth and size as the region-of-interest inthe muscle. Calculation of cMB was (GSL/6.6) �13.7. The variation in cMB between extended andflexed positions of muscles (“positional variation ofbackscatter”) was calculated by subtracting cMB inthe flexed position from cMB in the extended posi-

tion. Statistics were performed using the SPSS Grad-uate Pack (version 14.0) and expressed as mean(standard error).

RESULTS

All ultrasound measurements gave mean grayscalevalues in the linear range of the calibration curve.Intrarater reliability for cMB values was greater withthe probe oriented longitudinally than transversely(Table 2). cMB was higher (images were brighter)with the arm extended than when flexed (P � 0.001)(Table 1 and Fig. 2). Positional variation was greaterwith the probe oriented longitudinally than trans-

Table 2. Intrarater reliability of muscle backscatter.

Probeorientation

Armposition

Intraclass correlation

Singlemeasures

Mean of threemeasures

Longitudinal Flexion 0.87 0.95Extension 0.81 0.93

Transverse Flexion 0.47 0.82Extension 0.29 0.73

Two sets of three measurements were performed in 20 control subjects.The intraclass correlation coefficient was calculated comparing the single,unaveraged measurements between each set and the averages of the threemeasurements between each set. Reliability is greater in the longitudinalprobe orientation than the transverse and is further increased by usingaveraged measurements.

Table 1. Calibrated muscle backscatter in arm flexors.

Armposition

Controls

MyopathyP vs. allControlAll

Patient age Probe orientation

�40 y �40 y P Longitudinal Transverse P

Flexion 1.1 0.3 2.2 0.02 1.3 1.5 0.68 7.6 �0.001(0.4) (0.5) (0.6) (1.5) (0.6) (0.8)

Extension 4.7 3.8 6.0 0.008 5.0 3.2 �0.005 10.0 �0.001(0.4) (0.5) (0.6) (0.5) (0.6) (0.6)

Positionalvariation 3.6 3.5 3.8 0.71 3.7 1.7 0.01 2.4 0.04

(0.4) (0.4) (0.6) (0.5) (0.6) (0.4)

Probe orientation was longitudinal for all measurements except the transverse measurements. Calibrated muscle backscatter (cMB) is expressed in decibels.Results are expressed as mean (standard error). cMB is higher in the myopathy group than it is in controls in both flexion and extension. Positional variation isgreater in the longitudinal than it is in the transverse images of control muscles.

FIGURE 2. Longitudinal images of elbow flexors. Subcutaneousfat (sc), muscle (m), bone (b), and region-of-interest selected forbackscatter analysis (outlined in white) are shown in (A) (a 25-year-old control). Arm position affects the backscatter signal andis higher in extended (B) (6.6 dB) than flexed (C) (1.8 dB)muscles (a 22-year-old control). cMB is greater with increasingage. Compare (D), an 18-year-old control (�0.7 dB), with (E), a74-year-old control (9.6 dB). In a 45-year-old with limb-girdlemuscular dystrophy (F) (17.3 dB), there is a homogeneouslybright muscle (black arrow) compared with the subcutaneous fat(white arrow) and reduced bone signal.


versely (Table 1). In healthy controls longitudinalcMB was higher in older patients. No associationswere found between cMB and gender, height,weight, or body mass index. There was no differencein age between controls [39.4 (1.9) years] and myo-pathic patients [44.6 (2.8) years] (P � 0.1). cMB washigher in myopathic patients than in controls withthe extremities in either the flexed or extendedposition (P � 0.001) (Table 1 and Fig. 2). A cMB of�6.8 dB in the extended arm and �2.8 dB in theflexed arm resulted in sensitivities of 82% and 85%and specificities of 80% and 76%, respectively, foridentifying myopathy in our patients. The area un-der the receiver operating characteristic curves (Fig.3) was similar for extended (0.88) and flexed (0.91)arm positions.

DISCUSSION

Our goal was to obtain reliable quantitative skeletalmuscle ultrasound data that varied consistentlythrough measurable clinical ranges and could bereproduced across ultrasound systems. To achievethis goal we modified a technique previously em-ployed in analysis of cardiac muscle to quantify skel-etal muscle ultrasound backscatter.10,13,24 Thismethod of backscatter quantitation has two essentialcomponents: (1) configuration of the ultrasoundsystem to establish a linear relationship between themeasured grayscale values and estimated backscatter

(expressed in decibels); and (2) calibration of re-sults with an external reference. We configured ourultrasound system to provide an approximately lin-ear relationship between grayscale and backscattervalues by employing a linear postprocessing map. Wecould then determine the range within whichchanges in displayed grayscale intensity uniformlyreflected changes in the received backscatter signallevel (expressed in decibels). Previous ultrasoundstudies of skeletal muscle have expressed results asuncalibrated grayscale values without ensuring thatvalues fell within a linear range.7,14,15,20,21 Ourmethod also expresses backscatter values relative toan external reference, a tissue-mimicking phantomwith reproducible ultrasound characteristics. Thisallows reproducibility and comparison among differ-ent ultrasound set-ups.12

When quantifying skeletal muscle backscatter,probe orientation (relative to the predominant mus-cle fiber orientation) and arm position (extended orflexed) are important factors. Our reliability mea-surements were improved by imaging with the probein a longitudinal plane relative to muscle fibers. Thereliability of our technique is similar to other quan-titative ultrasound techniques.14,22 Probe orientationdid not alter results in one quantitative ultrasoundstudy of the supraspinatus muscle.16 This discrep-ancy may be due to the different physical propertiesof the two muscles. In our study there were differ-ences in backscatter levels between the longitudinaland transverse images in the extended, but not theflexed, arm. As both probe orientation and muscleposition can have variable effects on backscatter,they should be specified at the time of measurementand reproduced during any repeat studies. Our re-sults suggest that a longitudinal probe position maybe preferable for quantitative ultrasound studies, asit is more reliable. Based on detection of cMB dif-ferences with arm position, it may be more sensitive.

cMB varied with changes in age and the presenceof muscle pathology. The increased cMB levels withage in adults is similar to previously reported da-ta.14,22 Larger studies are required to define theextent of changes in normal cMB over the full rangeof ages in children and adults. Myopathies showedhigher cMB levels than produced by increasing age.The increased cMB with myopathy is similar to pre-vious findings using non-calibrated grayscale analy-sis.14,15,20,21 Additional studies are now being carriedout to determine whether cMB levels correlate withthe degree of muscle weakness and differ amongvarying types of myopathies. In general, cMB of the

FIGURE 3. cMB has similar specificity and sensitivity for identi-fying myopathy in both flexed and extended arms.


biceps brachii and brachialis appears to be useful formeasuring changes in muscle pathology, as signallevels are independent of height, weight, body massindex, and gender.

Positional variation of backscatter, comparinglevels in flexed and extended arms, is another anal-ysis that can be used to quantitate skeletal musclepathology.8 Positional variation minimizes the effectsof system-dependent parameters, as it is derivedfrom referencing one image to another, with onlythe arm position changing between measurements.A similar technique, comparison of the variation ofbackscatter in myocardium during systole and dias-tole, can characterize a wide spectrum of cardiacpathologies.19 We found that cMB is lower in normalskeletal muscle in flexed than in extended arms.Only a slightly smaller amount of positional variationwas found in diseased muscle compared with normalcontrols. Additional studies of cMB relating changesto the strength of contraction of imaged muscles willbe required to assess the utility of this type of self-referencing quantitative backscatter methodology inthe definition of muscle pathology.

Our calibration and ultrasound measurementtechniques were designed to minimize machine-de-pendent and patient effects. Reproduction of ourquantitative results in the clinical setting requiresthat each imaging system and probe combination becalibrated with reference to a common externalphantom. The phantom used in this study is com-mercially available, and the calibration process israpid and straightforward. Our technique minimizesinterpatient variation by using a single, deep, fixedfocus position and a large region-of-interest. Thedeep, fixed focus position (at a distance correspond-ing to 90% of the total image depth) maximizesuniform imaging characteristics across the depth ofthe image. Averaging grayscale values across a largeregion-of-interest minimizes any effects of inhomo-geneity and attenuation within muscle and the needfor depth correction.

In conclusion, calibrated backscatter measure-ments can reliably quantify ultrasound signals ofskeletal muscle in a manner that should be repro-ducible among different ultrasound systems andconfigurations. Calibrated quantitation will allowcollaboration among laboratories in the ultrasoundevaluation of skeletal muscle and should be a usefultechnique for objectively measuring skeletal musclepathology, and possibly for assessing pathologicalchanges over time. Our methods provide a basis foradditional studies to quantify degrees of variationboth within and among ultrasound systems.

This study was supported by the Washington University Neuro-muscular Research Fund and by a National Institutes of HealthNeurological Sciences Academic Development Award. The au-thors thank Stephen A. Mandel, BA, for statistical assistance.

REFERENCES

1. Aydinli N, Baslo B, Caliskan M, Ertas M, Ozmen M. Muscleultrasonography and electromyography correlation for evalu-ation of floppy infants. Brain Dev 2003;25:22–24.

2. Brockmann K, Becker P, Schreiber G, Neubert K, Brunner E,Bonnemann C. Sensitivity and specificity of qualitative muscleultrasound in assessment of suspected neuromuscular diseasein childhood. Neuromuscul Disord 2007;17:517–523.

3. Finch-Johnston AE, Gussak HM, Mobley J, Holland MR,Petrovic O, Perez JE, et al. Cyclic variation of integratedbackscatter: dependence of time delay on the echocardio-graphic view employed and the myocardial segment analyzed.J Am Soc Echocardiogr 2000;13:9–17.

4. Hall CS, Scott MJ, Lanza GM, Miller JG, Wickline SA. Theextracellular matrix is an important source of ultrasoundbackscatter from myocardium. J Acoust Soc Am 2000;107:612–619.

5. Heckmatt JZ, Leeman S, Dubowitz V. Ultrasound imaging inthe diagnosis of muscle disease. J Pediatr 1982;101:656–660.

6. Heckmatt JZ, Pier N, Dubowitz V. Real-time ultrasound imag-ing of muscles. Muscle Nerve 1988;11:56–65.

7. Heckmatt J, Rodillo E, Doherty M, Willson K, Leeman S.Quantitative sonography of muscle. J Child Neurol 1989;4(suppl):S101–106.

8. Helguera M. Quantitative changes in ultrasonic backscatterfrom human skeletal muscle with contraction. Master’s thesis,University of Rochester, Rochester, New York; 1990.

9. Hete B, Shung KK. Scattering of ultrasound from skeletalmuscle tissue. IEEE Trans Ultrason Ferroelectr Freq Control1993;40:354–365.

10. Holland MR, Gibson AA, Peterson LR, Areces M, Schaffer JE,Perez JE, et al. Measurements of the cyclic variation of myo-cardial backscatter from two-dimensional echocardiographicimages as an approach for characterizing diabetic cardiomy-opathy. J Cardiometab Syndr 2006;1:149–152.

11. Holland MR, Kovacs A, Posdamer SH, Wallace KD, Miller JG.Anisotropy of apparent backscatter in the short-axis view ofmouse hearts. Ultrasound Med Biol 2005;31:1623–1629.

12. Knipp BS, Zagzebski JA, Wilson TA, Dong F, Madsen EL.Attenuation and backscatter estimation using video signalanalysis applied to B-mode images. Ultrason Imag 1997;19:221–233.

13. Kovacs A, Courtois MR, Weinheimer CJ, Posdamer SH, Wal-lace KD, Holland MR, et al. Ultrasonic tissue characterizationof the mouse myocardium: successful in-vivo cyclic variationmeasurements. J Am Soc Echocardiogr 2004;17:883–892.

14. Maurits NM, Bollen AE, Windhausen A, De Jager AE, Van DerHoeven JH. Muscle ultrasound analysis: normal values anddifferentiation between myopathies and neuropathies. Ultra-sound Med Biol 2003;29:215–225.

15. Maurits NM, Beenakker EA, van Schaik DE, Fock JM, van derHoeven JH. Muscle ultrasound in children: normal values andapplication to neuromuscular disorders. Ultrasound Med Biol2004;30:1017–1027.

16. Nielsen PK, Jensen BR, Darvann T, Jørgensen K, Bakke M.Quantitative ultrasound image analysis of the supraspinatusmuscle. Clin Biomech 2000;15(suppl 1):S13–S16.

17. O’Brien PD, O’Brien WD Jr, Rhyne TL, Warltier DC, SagarKB. Relation of ultrasonic backscatter and acoustic propaga-tion properties to myofibrillar length and myocardial thick-ness. Circulation 1995;91:171–175.

18. O’Brien WD Jr, Sagar KB, Warltier DC, Rhyne TL. Acousticpropagation properties of normal, stunned, and infarcted


myocardium: morphological and biochemical determinants.Circulation 1995;91:154–160.

19. Perez JE, Holland MR, Barzailai B, Handley SM, VandenbergBF, Miller JG, et al. Ultrasonic characterization of cardiovas-cular tissue. In: Skorton DJ, Seelbert HR, Wolf GL, BrundageBH, editors. Cardiac imaging—a companion to Braunwald’sheart disease. W. B. Saunders; 1996. p 606–622.

20. Pillen S, Scholten RR, Zwarts MJ, Verrips A. Quantitativeskeletal muscle ultrasonography in children with suspectedneuromuscular disease. Muscle Nerve 2003;27:699–705.

21. Pillen S, Verrips A, van Alfen N, Arts I, Sie L, Zwarts M.Quantitative skeletal muscle ultrasound: diagnostic value inchildhood neuromuscular disease. Neuromuscul Disord2007;17:509–516.

22. Reimers CD, Fleckenstein JL, Witt TN, Muller-Felber W, Pon-gratz DE. Muscular ultrasound in idiopathic inflammatorymyopathies of adults. J Neurol Sci 1993;116:82–92.

23. Rose JH, Kaufmann MR, Wickline SA, Hall CS, Miller JG. Aproposed microscopic elastic wave theory for ultrasonic back-scatter from myocardial tissue. J Acoust Soc Am 1995;97:656–668.

24. Sosnovik DE, Baldwin SL, Lewis SH, Holland MR, Miller JG.Transmural variation of myocardial attenuation measuredwith a clinical imager. Ultrasound Med Biol 2001;27:1643–1650.

25. Zuberi SM, Matta N, Nawaz S, Stephenson JB, McWilliam RC,Hollman A. Muscle ultrasound in the assessment of suspectedneuromuscular disease in childhood. Neuromuscul Disord1999;9:203–7.


calibrated quantitative ultrasound imaging of skeletal muscle using backscatter analysis

Documents