- Email: [email protected]
Strain during gastric contractions can be measured using Doppler ultrasonography
Ultrasound in Med. & Biol., Vol. 28, Nos. 11/12, pp. 1457–1465, 2002 Copyright © 2002 World Federation for Ultrasound in Medicine & Biology Printed in the USA. All rights reserved 0301-5629/02/$–see front matter
PII: S0301-5629(02)00614-2
● Original Contribution STRAIN DURING GASTRIC CONTRACTIONS CAN BE MEASURED USING DOPPLER ULTRASONOGRAPHY ODD HELGE GILJA,* ANDREAS HEIMDAL,† TRYGVE HAUSKEN,* HANS GREGERSEN,‡ KNUT MATRE,* ARNOLD BERSTAD* and SVEIN ØDEGAARD* *Institute of Medicine, Haukeland Hospital, University of Bergen, Bergen, Norway; †Dept. of Informatics, University of Oslo, Oslo, Norway; and ‡Department of Surgical Gastroenterology A, Aalborg Hospital and Center for Sensory-motor Interaction, Aalborg University, Aalborg, Denmark (Received 6 February 2002; in final form 26 July 2002)
Abstract—This study was undertaken to explore if strain of the muscle layers within the gastric wall could be measured by transabdominal strain rate imaging (SRI), a novel Doppler ultrasound (US) method. A total of 9 healthy fasting subjects (8 women, 1 man; ages 22 to 55 years) were studied and both grey-scale and Doppler US data were acquired with a 5- to 8-MHz linear transducer in cineloops of 97 to 256 frames. Rapid stepwise inflation (5 to 60 mL) of an intragastric bag was carried out and bag pressure and SRI were measured simultaneously. SRI enabled detailed studies of layers within the gastric wall in all subjects. Great variations in strain distribution of the muscle layers were found. Radial strain was much higher in the circular than in the longitudinal muscle layer. Strains derived from SRI correlated well with strains obtained with B-mode measurements (r ⴝ 0.98, p < 0.05). During balloon distension, we found an inverse correlation between pressure and radial strain (r ⴝ ⴚ0.87, p < 0.05). Intraobserver correlation of strain estimation was r ⴝ 0.98 (p < 0.05) and intraobserver agreement was 0.2% ⴞ 18.6% (mean difference ⴞ 2SD, % strain). Interobserver correlation was r ⴝ 0.84 (p < 0.05) and interobserver agreement was 6.9% ⴞ 56.8%. SRI enables detailed mapping of radial strain distribution of the gastric wall and correlates well with B-mode measurements and pressure increments. (E-mail: [email protected]) © 2002 World Federation for Ultrasound in Medicine & Biology. Key Words: Strain rate imaging, Ultrasonography, Gastric motility, Stomach, Doppler, Strain.
leaves gastric motility undisturbed (Gilja et al. 1996). Hausken and colleagues have developed duplex Doppler methods using both color and pulsed Doppler for the study of transpyloric flow (Hausken et al. 1992, 1998b, 1998a,2001). Tissue Doppler imaging (TDI) enables mapping of local tissue velocities; thus, providing information about moving walls (Uematsu et al. 1997; Bach et al. 1996; Grubb et al. 1995). In color mapping, particularly in M-mode, TDI can accurately delineate the phases of the contractions with high temporal resolution. However, the point velocity of tissue does not differentiate between actively contracting and passively following tissue. Therefore, a novel method based on strain rate imaging (SRI) and estimation of strain has been developed to enable this differentiation. In general terms, strain is a measure of tissue deformation due to an imposed force (stress) (Gregersen and Kassab 1996; Gregersen et al. 2000; Gregersen 2000). It represents the fractional change from the original or unstressed dimension (Lagrangian strain), includes both
INTRODUCTION Distal gastric motor function has been investigated using several methods carried out to study motility and patients with gastroparesis and functional dyspepsia. Antral hypomotility with either decreased frequency or decreased amplitude of postprandial phasic pressure waves has been reported (Malagelada and Stanghellini1985; Camilleri et al. 1986; Kerlin 1989). A wide gastric antrum was observed, both during fasting and postprandially in these patients using ultrasonography (Hausken and Berstad 1992; Hausken et al. 1993), a finding also indicated in an earlier study on gastric emptying (Bolondi et al. 1985). Transabdominal ultrasonography is a noninvasive and radiation-free method that has proven applicable in the study of antral motility, partly due to the fact that it
Address correspondence to: Odd Helge Gilja, M.D., Ph.D., Institute of Medicine, Haukeland University Hospital, N-5021 Bergen, Norway. E-mail: [email protected] 1457
1458
Ultrasound in Medicine and Biology
lengthening, or expansion (positive strains), and shortening, or compression (negative strains). Strain is a dimensionless quantity, such as the Cauchy strain (Gregersen et al. 1999a, 2000; Gregersen 2000). Because zero stress lengths are impossible to measure in vivo, L0 is replaced by the initial muscle length or the precontractile state of the antrum. Strain can alternatively be expressed as natural strain, defined as ln(L/L0), where ln is the natural logarithm. The temporal derivative of natural strain (i.e., the strain rate) is a measure of the rate of deformation. This measure is equivalent to the spatial derivative of the velocity of contraction. Ophir and colleagues have developed ultrasonic methods for quantitative imaging of strain and elastic modulus distributions in soft tissues (Ophir et al. 1991, 2000). This method is not based on the Doppler method, but on external tissue compression with subsequent computation of the strain profile along the transducer axis. The purpose of our study was to explore the applicability of a novel ultrasound (US) Doppler method (SRI) for the measurement of Cauchy’s strain in the wall of the contracting gastric antrum. MATERIALS AND METHODS Subjects A total of 9 healthy volunteers, 8 women and 1 man, 22 to 55 years old (median: 24 years), entered the study. Mean body mass index was 21.9 ⫾ 2.8 kg/m2. The criteria of exclusion from the study were previous surgery of the upper GI tract, present or previous peptic ulcer, alcoholism, and pregnant or lactating women. Definitions Strain in our context means relative strain. Strain is a measure of tissue deformation. Cauchy’s strain is defined as e ⫽ (L ⫺ L0)/L0, where L0 is the reference length and L is the instantaneous length during loading. The reference for relative strain is the precontractile state, which constitutes a resting level; however, not necessarily zero strain. Strain rate (SR) is defined as the temporal derivative of strain, and is a measure of the rate of deformation. See Appendix 1 for more details. Negative strain rate means that the tissue segment is becoming shorter (or thinner), and positive strain rate means that the segment is becoming longer (or thicker). The sample size of strain is analogous to sample volume in ordinary pulsed Doppler. Sample size denotes the distance between the two measurement points along the US beam used when estimating the velocity gradient. Phase 1 of the interdigestive motor pattern cycle is the period without any contractions seen on ultrasonography. Phase 2 denotes sporadic contractions and phase 3
Volume 28, Numbers 11/12, 2002
is the short period of regular contractions with a frequency of 3/min, before phase 1 is restored. Fed state means the postprandial motor pattern consisting of regular contractions of 3/min, lasting much longer than the phase 3 interdigestive pattern. Experimental procedure The protocol consisted of two different acquisition series: one of fasting contractile activity and one of responses to intragastric rapid volume distensions. The volunteers were fasting for a minimum of 8 h. They were scanned while sitting in a chair, leaning slightly backwards, with the transducer positioned in the epigastrium. US images were obtained using a digital scanner (System Five, GE Vingmed Ultrasound, Horten, Norway). We used a B-mode transducer frequency of 8 MHz during most acquisitions and a Doppler frequency of 5.7 MHz. To minimize the noise level and to decrease the Doppler sample size, the pulse repetition frequency (PRF) was set to 250 Hz. The average Doppler sample size, defined as the spatial distance between the two measurement points along the US beam, was 2.0 ⫾ 0.6 mm (SD). The range of frames in the cine-loops was 97 to 256, with a mean of 168 frames recorded at a speed of 45 to 72 ms/frame. A typical sonogram of the antrum including the color Doppler data is shown in Fig. 1. The average maximum depth of scanning was 6 cm. All selected US cineloops were scanned while the subjects held their breath, and the data were temporarily stored on the scanner before transfer through the ethernet to a PC workstation (Dell Optiplex, Austin, TX). The workstation had a 500-MHz Pentium III processor and 250 MB of RAM. A prototype software application (tvi.exe) was used to calculate strain values from the Doppler data. The strain-rate imaging method The velocity component v of every point in the muscle is available from tissue Doppler data, so the spatial gradient can be estimated from two points along the US beam. This estimate can be performed in realtime as a small extension of the tissue Doppler data processing, and this method was termed strain rate imaging (SRI) (Heimdal et al. 1998; Stoylen et al. 1999). Because only the velocity component along the US beam is available, only the strain rate in the beam direction is estimated (Fig. 2). Subsequently, the strain rate was mapped in segments of the gastric antrum wall. A color scheme using a transition from yellow to red for increasingly more negative strain rate (i.e., shortening) and a transition from cyan to blue for increasingly more positive strain rate (i.e., elongation) was chosen for 2-D and M-mode displays, see Fig. 1. In addition, strain rates near zero were colored green to facilitate the definition of areas and periods of low strain rates. This is
Gastric antral strain ● O. H. GILJA et al.
1459
Fig. 2. The schematic drawing shows how the US beams traverse through the gastric wall. The beam direction is depicted in a normal, contracted and relaxed state of the gastric antrum and its relation to circumferential and radial strain directions.
Fig. 1. (top) Photograph showing the human gastric antrum in a sagittal section with a superimposed color Doppler region for SRI. The image was acquired with a linear-array transducer at 8 MHz B-mode frequency. Different wall layers can be observed in the antrum and the main muscle layer is outlined by a white arrow. The skin, the rectus muscle, the liver and the superior mesenteric vein are also visualized. A red line depicts the distance that is viewed in anatomical M-mode. (bottom) A typical anatomical M-mode registration of SRI is shown, where blue denotes expansion and green denotes compression of tissue. The spikes on the graph correspond to the aortic pulsation.
different from tissue Doppler imaging, where there is a sharp color discontinuity around zero velocity. In a prototype computer postprocessing program, tissue velocities and strain rate values from multiple points along an M-mode cursor line were extracted. In this software, the images showing the velocity of tissue motion were superimposed on the 2-D ultrasonic images for real-time display in color. The velocities and strain rates were also available as numerical values for quantitative analysis. The strain measurements In the postprocessing of data, we averaged our measurements over three samples in the lateral direction to the beam and of five samples in the radial beam direction, corresponding to 1.2 mm in the lateral direc-
tion and 1.0 mm in the radial direction. Gain, color rejection and tissue priority were all optimized before strain estimation started. The speed of the cineloop could be varied, adjusted to improve characterization of details. The strain calculations were done relative to the starting point of the cineloop, defined as the beginning of the contraction, as assessed visually by ultrasonography. Two main options for strain calculations existed; one (strain/motion) based on the application of a single spatial point of measurement onto the image. The other was based on curved M-mode analysis, where a line consisting of up to eight measurement points was drawn along a region-of-interest (ROI). Subsequently, time integrals were calculated either based on the labeled values or the mean or median values between the neighboring points. Ordinary B-mode images of the antrum were also analyzed with respect to measurement of the width of the proper muscle layer in a relaxed and contracted state. Based on these measurements, strain values were computed and compared to SRI data for validation purposes. The antral wall changes position in the image frame during a contraction; thus, inducing a potential error in strain estimation. The software inherits a built-in compensation for such movements because the intended point of measurement can be moved to track the position of the anatomical focus of interest. One SRI analysis of a single cineloop took approximately 1 min. One cineloop was traced 3 times and then the average value was recorded. Distension protocol A specially designed distension probe was constructed. The probe was 120-cm long and contained a
1460
Ultrasound in Medicine and Biology
30-m thick polyesterurethane bag 10 cm from the tip of the probe. The cylindrical bag was 5-cm long and could be inflated with fluid through an infusion channel of 3.5 mm to a diameter of 40 mm without stretching the bag wall. The probe contained a metal clip in the middle of the bag as a marker for US. Two side-holes for pressure measurements were placed in the middle of the bag and 2-cm proximal to the bag. The lumens and side-holes both had diameters of 0.5 mm. The perfusion rate for the pressure channels was 0.1 mL min⫺1. The pressure was measured by a low compliance perfusion system connected to external transducers (Medex, Denmark). The pressure data were amplified and analog-to-digital converted at a sampling rate of 10 Hz using a motility data-acquisition system (Gatehouse Ltd., Nørresundby, Denmark). The digitized data were stored on a PC for later analysis in the same software. After calibrating the measurement system, the tube was passed into the stomach via the nostrils. A manometric investigation was carried out to evaluate the position of the catheter and to confirm that the contractile pattern was consistent with phase 1. The middle of the balloon was then placed approximately 3 cm proximal to the pylorus. The position of the catheter was also visualized by ultrasonography. In one subject, fluoroscopic guidance was necessary to ensure correct position of the balloon. The zero pressure level for the distension series was determined before the distension protocol was performed as a volume-controlled series. The fluid volume of the bag was changed intermittently with volumes ranging from 5 to 60 mL for 1-min periods using a hand-held syringe. The volume infusions were done manually as fast as it was possible to empty the syringe volume. The distensions were separated by 2-min resting periods (i.e., the volume was withdrawn from the bag and the bag pressure was kept slightly negative during these periods to ensure complete emptying of the bag). SRI was performed simultaneously with the infusion of water into the balloon. A mark on the skin, as well as the localization of the metal marker on the probe, ensured that the same area was scanned each time. Variability experiments Two independent observers (O. H. Gilja and T. Hausken) estimated intra- and interobserver variabilities of the strain calculations. A total of 14 data sets were analyzed twice by each observer using the dedicated software program, and the mean result was recorded. Some data sets were utilized more than once and traced at different portions of the gastric wall. Because of relatively high strain variations in the different portions of the wall (see results), the two observers had to agree on which part of the antrum to measure before the actual tracing was performed.
Volume 28, Numbers 11/12, 2002
Fig. 3. The box plot shows the magnitude and the differences in strain (%) of the anterior, inferior and superior part of the muscularis propria in a sagittal section of the gastric antrum. Data were acquired in a radial direction by SRI using Doppler ultrasonography.
Statistical analysis Median values and interquartile ranges were calculated and reported, if not otherwise specifically noted. Coefficient of correlation was computed for intra- and interobserver analysis. Limits of agreement were determined as suggested by Bland and Altmann (1986) to evaluate intra- and interobserver agreement. The level of statistical significance was p ⬍ 0.05. All statistical calculations and graphic designs were performed using commercially available software. Ethical aspects The study was approved by the regional ethic committee and was conducted in accordance with the revised Declaration of Helsinki. All volunteers gave written, informed consent to participate in the trial. RESULTS High-frequency real-time ultrasonography and SRI enabled detailed studies of layers within the gastric wall in all subjects. Most of the antral contractions were obtained in phase II according to ultrasonographic evaluation, but one of the contractions was of phase III origin. The anterior part of the gastric antrum was more accessible for SRI than the other parts of the antral circumference. A high positive radial strain was found in the anterior part of the gastric wall (96%); 163% (median values; interquartile range) during contraction. Negative strain was found in the inferior part (⫺33%, 23%) and in the superior part (⫺25%, 6%) (circumferential strain), see Fig. 3. The posterior part of the antral circumference
Gastric antral strain ● O. H. GILJA et al.
1461
Fig. 5. This scatter plot shows the association between strain measured by Doppler ultrasonography and deformation calculated by radial measurements on the ordinary B-mode images. The line of identity is drawn. Fig. 4. These graphics outline the difference in strain of the circular vs. the longitudinal muscle layer of the antral wall. (top) The exact sampling points in the gastric wall are denoted with a square and an asterix in the ultrasonogram. Blue denotes expansion and green denotes compression of tissue. (bottom) The difference between the high positive strain (elongation) of the circular (inner) muscle layer and the low strain of the longitudinal (outer) muscle layer, illustrates the separate biomechanical events in the two portions of the main muscle layer. The green line corresponds to the red square and the yellow line corresponds to the white asterisk (inner circular muscle layer). The data were obtained in a radial direction.
was frequently difficult to image adequately for SRI analysis due to movement of luminal particles and the long distance from the scan head. Local mapping of radial strain distributions of the anterior part of the muscle layers was enabled with this Doppler-based method. In three individuals, it was possible, due to a combination of high resolution and a slim body habitus, to distinguish between strain in the longitudinal and circular muscle layers. During contractions, radial strain was significantly higher in the circular muscle layer compared to the longitudinal layer, where radial strain was closer to zero (e.g., 105% vs. 10%, see Fig. 4). Measurements of the proper muscle diameters on the B-mode ultrasonograms showed good correlation with SRI measurements (r ⫽ 0.98, p ⬍ 0.05), as visualized in Fig. 5. The mean difference was 0.8% and the limits of agreement were ⫺28.6% to 30.2%. The intraobserver correlation between two observers calculating radial strain using the designed software was
r ⫽ 0.98 (p ⬍ 0.01) (Fig. 6) and the intraobserver agreement was 0.2% ⫾ 18.6% (mean difference ⫾ 2 SD, % strain) (Fig. 7). The interobserver correlation was r ⫽ 0.84 (Fig. 8) and the interobserver agreement was 6.9% ⫾ 56.8% (Fig. 9). However, as observed in Fig. 8, there
Fig. 6. Scatter plot showing the intraobserver correlation of strain in the antral muscle layer using SRI. The line of identity is drawn.
1462
Ultrasound in Medicine and Biology
Fig. 7. Plot displaying the intraobserver agreement of strain calculation in antral wall muscle using SRI. The difference between the two observations (Obs 1a – Obs 1b) is depicted on the y-axis. The average of the two observations is plotted along the x-axis. The intraobserver agreement was 0.2% ⫾ 18.6% (mean difference ⫾ 2 SD, % strain), displaying no proportional error effect.
is an obvious outlier in the data, representing the only phase 3 measurement in this study. A phase 3 contraction has much more complex and dynamic movements compared to phase 2 contractions. If we exclude the outlier of the phase 3 contraction, the correlation coefficient is 0.95 and limits of agreement are ⫺0.3% ⫾ 19.6% for the interobserver variation
Volume 28, Numbers 11/12, 2002
Fig. 9. Interobserver agreement of strain calculation in antral wall muscle using SRI. The difference between the 2 observers (Obs 1 – Obs 2) is depicted on the y-axis. Mean of the two observers is plotted along the x-axis. The interobserver agreement was 6.9% ⫾ 56.8% (mean difference ⫾ 2 SD, % strain). However, if the outlier was excluded, interobserver agreement was ⫺0.3% ⫾ 19.6%.
During distensions, we observed that SRI could discriminate between active and passive deformation of the gastric muscle layer. The regular distension pattern was negative radial strain in the anterior wall in response to sudden pressure increments. However, two other distinct patterns were noted during careful postprocessing of image data; distension on contraction (the balloon was inflated when a barely visible contraction had already started) or contraction on distension (the gastric wall was compressing the balloon undergoing inflation). Accordingly, SRI was able to discriminate between active and passive deformation of the wall, not observed during ordinary 2-D scanning. We found an inverse significant relationship between radial strain estimated by SRI and maximum intraballoon pressure (r ⫽ ⫺0.87, p ⬍ 0.05) during distension (Fig. 10). DISCUSSION
Fig. 8. Scatter plot showing the interobserver correlation of strain in the antral muscle layer by using SRI. The line of identity is drawn.
For the first time, a noninvasive method based on Doppler ultrasonography was applied to estimate strain in moving gastrointestinal tissue in humans. In this pilot study, SRI demonstrated high spatial and temporal resolution facilitating detailed analysis of contracting gastric smooth muscle. Using SRI, it was possible to discern biomechanical events in the inner circular vs. the outer longitudinal muscle layer of the antral wall, a phenomenon not observed during ordinary 2-D scanning. Furthermore, SRI enabled mapping of local strain distribution within a muscle layer showing the existence of significant strain variations within small areas. Gastrointestinal strain is a measure of regional deformation and, by definition, negative strain means shortening and positive strain elongation. SRI is an extension
Gastric antral strain ● O. H. GILJA et al.
Fig. 10. Scatter plot demonstrating the association between radial strain estimated using SRI and volume infusion into a balloon distending the antrum. A significant negative correlation is depicted, showing that, as increasing volume is distending the antrum, an increasing negative strain (shortening) is measured radially in the anterior proper muscle layer of the wall.
of existing tissue Doppler methods and makes measurement of deformation possible (Sutherland et al. 1999). Recently, SRI was used to quantify myocardial movement (Voigt et al. 2000; Stoylen et al. 1999) and to assess coronary artery disease (Stoylen et al. 2000; Belohlavek et al. 2001). In a study in anesthetized dogs, SRI was validated against sonomicrometry as a reference method (Urheim et al. 2000). The authors found that, after placing longitudinal ultrasonic segment-length crystals in contracting tissue, strain by Doppler correlated well with strain by sonomicrometry. They concluded that strain estimation by this novel Doppler ultrasonography represented a powerful method for quantifying regional muscle function, and was less influenced by tethering effects than ordinary Doppler tissue imaging. In our study, we found a significant association between SRI measurements and strain calculated on the basis of metric data obtained directly from the B-mode images. Furthermore, intra- and interobserver variabilities of the strain measurements made by SRI were acceptable. SRI correlated well with pressure increments following distension of the antrum by a balloon, showing that strain measured by the SRI method is associated to the pressure force generated by the antral muscle. Interestingly, we found that the outer longitudinal muscle layer of the antral wall had different strain levels than the inner circular muscle layer during a contraction. When radial strain was measured, the longitudinal layer
1463
exhibited a relatively low negative strain (shortening) and the circular muscle layer had higher positive strain (elongation) (Gregersen et al. 1999b). Previously, the separation of the inner and outer muscle layer has been demonstrated with endoluminal ultrasonography with increasing transducer frequencies (Taniguchi et al. 1993, 1995). In these studies, a miniaturized US device attached to the gastrointestinal mucosa by suction has been used to examine motility of esophageal wall layers using M-mode. In that study, an average of 83% thickening of the inner circular muscle layer was observed during contractions. In the present study, a typical reading is shown in Fig. 4, where the positive strain of the circular layer was 105% and strain of the longitudinal layer was approximately 10%. Accordingly, this study shows that the inner circular layer constitutes the driving force during contraction and that the outer layer passively follows the movements during this sequence of phase II contractile activity of the antral wall muscle. To be able to distinguish between the outer and inner muscle layer of muscularis propria of the antral wall, the scanner was preset to a very small strain sample size (average 2 mm), which determines the spatial resolution of the strain estimates. We also took care in not applying too much manual force with the scan head to avoid pressure effects on the gastric wall (Odegaard et al. 1992). However, when a small sample size was used for calculating the strain rate, the random noise became relatively larger. The demand for high spatial resolution was compromised by lower SNR. To reduce the noise, multiple samples from a longer segment along the Mmode beam were averaged during analysis using dedicated software. Furthermore, by the integration procedure from SR to strain the random noise was considerably reduced, and we obtained relatively smooth strain curves. Generally, SRI is limited by the same factors that hamper routine ultrasonography, namely, bowel gas and excessive abdominal fat. However, for ultrasonography of the fasting gastric antrum, gas is seldom a problem for image quality. In our study, the volunteers had normal body habitus and luminal gas was not a significant problem. Another limitation of SRI is marked angle-dependency, more so than for other Doppler modalities. This is due to the fundamental difference between measurement of fluid velocities where particles move freely, and tissue velocities in solid structures where deformation in one direction is always associated with deformation in other directions to keep the volume of the structure constant. To minimize this problem, we took care in aligning the US beams as perpendicular to the anterior part of the muscle as possible; thus, measuring only the axial strain component. While evaluating intra- and interobserver agreement, we observed relatively large variations in our
1464
Ultrasound in Medicine and Biology
strain estimates. This was due partly to the need for selecting a very small strain sample to be able to distinguish the inner and outer muscle layers. Furthermore, we studied a variety of gastric contractions ranging from early phase 2 to phase 3 and, in future studies, variability can probably be reduced by omitting outliers originating from other phases of the interdigestive motor cycle. The data provided by the SRI technique can likely be used for further mechanical analysis, such as to compute longitudinal strains using incompressibility assumptions and to derive mechanical parameters from pressure-strain relationships. In conclusion, we found that Doppler examination based on SRI and dedicated software can be used to noninvasively obtain biomechanical information of the contracting gastric antrum. Fasting antral contractions have different strain distributions within the muscle layers and the circular and longitudinal muscle layers show different strain levels. Moreover, SRI enables discrimination between active and passive deformation of the gastric muscle layer during distension. The high temporal and spatial resolutions inherent in SRI offer very detailed transabdominal imaging, but further studies are needed to establish the precise role of this method, improve the technique to reduce variance in the recorded strain and to investigate possible clinical applications. Acknowledgments—This study was supported by grants from Innovest Strategic Research Programme, Haukeland University Hospital, Bergen, Norway; the Karen Elise Jensens Foundation; the Scandinavian Association for Gastrointestinal Motility (SAGIM); and the Danish Technical Research Council. The authors are grateful for the technical support from GE Vingmed Ultrasound, Horten, Norway.
REFERENCES Bach DS, Armstrong WF, Donovan CL, Muller DW. Quantitative Doppler tissue imaging for assessment of regional myocardial velocities during transient ischemia and reperfusion. Am Heart J 1996;132:721–725. Belohlavek M, Pislaru C, Bae RY, Greenleaf JF, Seward JB. Real-time strain rate echocardiographic imaging: Temporal and spatial analysis of postsystolic compression in acutely ischemic myocardium. J Am Soc Echocardiogr 2001;14:360–369. Bland JM, Altmann DG. Statistical methods for assessing agreement between two methods of clinical measurements. Lancet 1986;1: 307–310. Bolondi L, Bortolotti M, Santi V, et al. Measurement of gastric emptying time by real-time ultrasonography. Gastroenterology 1985; 89:752–759. Camilleri M, Malagelada JR, Kao PC, Zinsmeister AR. Gastric and autonomic responses to stress in functional dyspepsia. Dig Dis Sci 1986;31:1169–1177. Gilja OH, Hausken T, Odegaard S, Berstad A. Three-dimensional ultrasonography of the gastric antrum in patients with functional dyspepsia. Scand J Gastroenterol 1996;31:847–855. Gregersen H. Residual strain in the gastrointestinal tract: A new concept. Neurogastroenterol Motil 2000;12:411–414. Gregersen H, Barlow J, Thompson D. Development of a computercontrolled tensiometer for real-time measurements of tension in tubular organs. Neurogastroenterol Motil 1999a;11:109–118.
Volume 28, Numbers 11/12, 2002 Gregersen H, Kassab GS. Biomechanics of the gastrointestinal tract. Neurogastroenterol Motil 1996;8:77–97. Gregersen H, Kassab GS, Fung YC. The zero-stress state of the gastrointestinal tract: Biomechanical and functional implications. Dig Dis Sci 2000;45:2271–2281. Gregersen H, Lee TC, Chien S, Skalak R, Fung YC. Strain distribution in the layered wall of the esophagus. J Biomech Eng 1999b;121: 442–448. Grubb NR, Fleming A, Sutherland GR, Fox KA. Skeletal muscle contraction in healthy volunteers: Assessment with Doppler tissue imaging. Radiology 1995;194:837–842. Hausken T, Berstad A. Wide gastric antrum in patients with non-ulcer dyspepsia. Effect of cisapride. Scand J Gastroenterol 1992;27:427– 432. Hausken T, Gilja OH, Odegaard S, Berstad A. Flow across the human pylorus soon after ingestion of food, studied with duplex sonography. Effect of glyceryl trinitrate. Scand J Gastroenterol 1998a;33: 484–490. Hausken T, Gilja OH, Undeland KA, Berstad A. Timing of postprandial dyspeptic symptoms and transpyloric passage of gastric contents. Scand J Gastroenterol 1998b;33:822–827. Hausken T, Goldman B, Leotta DF, Odegaard S, Martin RW, Quantification of gastric emptying and duodenogastric reflux stroke volumes using three-dimensional guided digital color Doppler imaging. Eur J Ultrasound 2001(in press). Hausken T, Odegaard S, Matre K, Berstad A. Antroduodenal motility and movements of luminal contents studied by duplex sonography. Gastroenterology 1992;102:1583–1590. Hausken T, Svebak S, Wilhelmsen I, et al. Low vagal tone and antral dysmotility in patients with functional dyspepsia. Psychosom Med 1993;55:12–22. Heimdal A, Stoylen A, Torp H, Skjaerpe T. Real-time strain rate imaging of the left ventricle by ultrasound. J Am Soc Echocardiogr 1998;11:1013–1019. Kerlin P. Postprandial antral hypomotility in patients with idiopathic nausea and vomiting. Gut 1989;30:54–59. Malagelada JR, Stanghellini V. Manometric evaluation of functional upper gut symptoms. Gastroenterology 1985;88:1223–1231. Odegaard S, Kimmey MB, Martin RW, et al. The effects of applied pressure on the thickness, layers, and echogenicity of gastrointestinal wall ultrasound images. Gastrointest Endosc 1992;38:351– 356. Ophir J, Cespedes I, Ponnekanti H, Yazdi Y, Li X. Elastography: A quantitative method for imaging the elasticity of biological tissues. Ultrason Imaging 1991;13:111–134. Ophir J, Garra B, Kallel F, et al. Elastographic imaging. Ultrasound Med Biol 2000;26(Suppl. 1):S23–S29. Stoylen A, Heimdal A, Bjornstad K, Torp HG, Skjaerpe T. Strain rate imaging by ultrasound in the diagnosis of regional dysfunction of the left ventricle. Echocardiography 1999;16:321–329. Stoylen A, Heimdal A, Bjornstad K, et al. Strain rate imaging by ultrasonography in the diagnosis of coronary artery disease. J Am Soc Echocardiogr 2000;13:1053–1064. Sutherland GR, Kukulski T, Voight JU, D’hooge J. Tissue Doppler echocardiography. Future developments. Echocardiography 1999; 16:509–520. Taniguchi DK, Martin RW, Trowers EA, et al. Changes in esophageal wall layers during motility: Measurements with a new miniature ultrasound suction device. Gastrointest Endosc 1993;39:146–152. Taniguchi DK, Martin RW, Trowers EA, Silverstein FE. Simultaneous M-mode echoesophagram and manometry in the sheep esophagus. Gastrointest Endosc 1995;41:582–586. Uematsu M, Nakatani S, Yamagishi M, Matsuda H, Miyatake K. Usefulness of myocardial velocity gradient derived from two-dimensional tissue Doppler imaging as an indicator of regional myocardial contraction independent of translational motion assessed in atrial septal defect. Am J Cardiol 1997;79:237–241. Urheim S, Edvardsen T, Torp H, Angelsen B, Smiseth OA. Myocardial strain by Doppler echocardiography. Validation of a new method to quantify regional myocardial function. Circulation 2000;102:1158– 1164.
Gastric antral strain ● O. H. GILJA et al. Voigt JU, Arnold MF, Karlsson M, et al. Assessment of regional longitudinal myocardial strain rate derived from doppler myocardial imaging indexes in normal and infarcted myocardium. J Am Soc Echocardiogr 2000;13:588–598.
APPENDIX This appendix describes the concept of strain and strain rate, and how these entities can be estimated from the Doppler velocities. In general terms, strain means relative deformation, and strain rate means rate of deformation. If an object has an initial length L0 that, after a certain time, changes to L, strain is defined as:
e⫽
L ⫺ L0 . L0
1465
change in length per unit length equals the velocity gradient times the time increment: dL 2 ⫺ 1 ⫽ dt. L L
Because it is not feasible to track the end points of the object, the velocity gradient is estimated from two points with a fixed distance: dL 共r ⫹ ⌬r兲 ⫺ 共r兲 ⬇ dt ⫽ SR dt. L ⌬r
(4A)
The velocity gradient estimate is termed strain rate (SR). Finally, by integrating this equation, we arrive at:
冕
(1A) log
The instantaneous change in length (dL) in a small time increment (dt) is related to the velocities (1 and 2) of the end points of the object: dL ⫽ 共 2 ⫺ 1兲dt
(3A)
(2A)
By dividing eqn (2A) with L, we see that the instantaneous
L ⫽ L0
t
SR dt,
(5A)
t0
where “log” denotes the natural logarithm. This gives the following relation between strain rate estimated by velocity gradient and strain:
冉冕 冊 t
e ⫽ exp
SR dt
t0
⫺ 1.
(6A)
PII: S0301-5629(02)00614-2
● Original Contribution STRAIN DURING GASTRIC CONTRACTIONS CAN BE MEASURED USING DOPPLER ULTRASONOGRAPHY ODD HELGE GILJA,* ANDREAS HEIMDAL,† TRYGVE HAUSKEN,* HANS GREGERSEN,‡ KNUT MATRE,* ARNOLD BERSTAD* and SVEIN ØDEGAARD* *Institute of Medicine, Haukeland Hospital, University of Bergen, Bergen, Norway; †Dept. of Informatics, University of Oslo, Oslo, Norway; and ‡Department of Surgical Gastroenterology A, Aalborg Hospital and Center for Sensory-motor Interaction, Aalborg University, Aalborg, Denmark (Received 6 February 2002; in final form 26 July 2002)
Abstract—This study was undertaken to explore if strain of the muscle layers within the gastric wall could be measured by transabdominal strain rate imaging (SRI), a novel Doppler ultrasound (US) method. A total of 9 healthy fasting subjects (8 women, 1 man; ages 22 to 55 years) were studied and both grey-scale and Doppler US data were acquired with a 5- to 8-MHz linear transducer in cineloops of 97 to 256 frames. Rapid stepwise inflation (5 to 60 mL) of an intragastric bag was carried out and bag pressure and SRI were measured simultaneously. SRI enabled detailed studies of layers within the gastric wall in all subjects. Great variations in strain distribution of the muscle layers were found. Radial strain was much higher in the circular than in the longitudinal muscle layer. Strains derived from SRI correlated well with strains obtained with B-mode measurements (r ⴝ 0.98, p < 0.05). During balloon distension, we found an inverse correlation between pressure and radial strain (r ⴝ ⴚ0.87, p < 0.05). Intraobserver correlation of strain estimation was r ⴝ 0.98 (p < 0.05) and intraobserver agreement was 0.2% ⴞ 18.6% (mean difference ⴞ 2SD, % strain). Interobserver correlation was r ⴝ 0.84 (p < 0.05) and interobserver agreement was 6.9% ⴞ 56.8%. SRI enables detailed mapping of radial strain distribution of the gastric wall and correlates well with B-mode measurements and pressure increments. (E-mail: [email protected]) © 2002 World Federation for Ultrasound in Medicine & Biology. Key Words: Strain rate imaging, Ultrasonography, Gastric motility, Stomach, Doppler, Strain.
leaves gastric motility undisturbed (Gilja et al. 1996). Hausken and colleagues have developed duplex Doppler methods using both color and pulsed Doppler for the study of transpyloric flow (Hausken et al. 1992, 1998b, 1998a,2001). Tissue Doppler imaging (TDI) enables mapping of local tissue velocities; thus, providing information about moving walls (Uematsu et al. 1997; Bach et al. 1996; Grubb et al. 1995). In color mapping, particularly in M-mode, TDI can accurately delineate the phases of the contractions with high temporal resolution. However, the point velocity of tissue does not differentiate between actively contracting and passively following tissue. Therefore, a novel method based on strain rate imaging (SRI) and estimation of strain has been developed to enable this differentiation. In general terms, strain is a measure of tissue deformation due to an imposed force (stress) (Gregersen and Kassab 1996; Gregersen et al. 2000; Gregersen 2000). It represents the fractional change from the original or unstressed dimension (Lagrangian strain), includes both
INTRODUCTION Distal gastric motor function has been investigated using several methods carried out to study motility and patients with gastroparesis and functional dyspepsia. Antral hypomotility with either decreased frequency or decreased amplitude of postprandial phasic pressure waves has been reported (Malagelada and Stanghellini1985; Camilleri et al. 1986; Kerlin 1989). A wide gastric antrum was observed, both during fasting and postprandially in these patients using ultrasonography (Hausken and Berstad 1992; Hausken et al. 1993), a finding also indicated in an earlier study on gastric emptying (Bolondi et al. 1985). Transabdominal ultrasonography is a noninvasive and radiation-free method that has proven applicable in the study of antral motility, partly due to the fact that it
Address correspondence to: Odd Helge Gilja, M.D., Ph.D., Institute of Medicine, Haukeland University Hospital, N-5021 Bergen, Norway. E-mail: [email protected] 1457
1458
Ultrasound in Medicine and Biology
lengthening, or expansion (positive strains), and shortening, or compression (negative strains). Strain is a dimensionless quantity, such as the Cauchy strain (Gregersen et al. 1999a, 2000; Gregersen 2000). Because zero stress lengths are impossible to measure in vivo, L0 is replaced by the initial muscle length or the precontractile state of the antrum. Strain can alternatively be expressed as natural strain, defined as ln(L/L0), where ln is the natural logarithm. The temporal derivative of natural strain (i.e., the strain rate) is a measure of the rate of deformation. This measure is equivalent to the spatial derivative of the velocity of contraction. Ophir and colleagues have developed ultrasonic methods for quantitative imaging of strain and elastic modulus distributions in soft tissues (Ophir et al. 1991, 2000). This method is not based on the Doppler method, but on external tissue compression with subsequent computation of the strain profile along the transducer axis. The purpose of our study was to explore the applicability of a novel ultrasound (US) Doppler method (SRI) for the measurement of Cauchy’s strain in the wall of the contracting gastric antrum. MATERIALS AND METHODS Subjects A total of 9 healthy volunteers, 8 women and 1 man, 22 to 55 years old (median: 24 years), entered the study. Mean body mass index was 21.9 ⫾ 2.8 kg/m2. The criteria of exclusion from the study were previous surgery of the upper GI tract, present or previous peptic ulcer, alcoholism, and pregnant or lactating women. Definitions Strain in our context means relative strain. Strain is a measure of tissue deformation. Cauchy’s strain is defined as e ⫽ (L ⫺ L0)/L0, where L0 is the reference length and L is the instantaneous length during loading. The reference for relative strain is the precontractile state, which constitutes a resting level; however, not necessarily zero strain. Strain rate (SR) is defined as the temporal derivative of strain, and is a measure of the rate of deformation. See Appendix 1 for more details. Negative strain rate means that the tissue segment is becoming shorter (or thinner), and positive strain rate means that the segment is becoming longer (or thicker). The sample size of strain is analogous to sample volume in ordinary pulsed Doppler. Sample size denotes the distance between the two measurement points along the US beam used when estimating the velocity gradient. Phase 1 of the interdigestive motor pattern cycle is the period without any contractions seen on ultrasonography. Phase 2 denotes sporadic contractions and phase 3
Volume 28, Numbers 11/12, 2002
is the short period of regular contractions with a frequency of 3/min, before phase 1 is restored. Fed state means the postprandial motor pattern consisting of regular contractions of 3/min, lasting much longer than the phase 3 interdigestive pattern. Experimental procedure The protocol consisted of two different acquisition series: one of fasting contractile activity and one of responses to intragastric rapid volume distensions. The volunteers were fasting for a minimum of 8 h. They were scanned while sitting in a chair, leaning slightly backwards, with the transducer positioned in the epigastrium. US images were obtained using a digital scanner (System Five, GE Vingmed Ultrasound, Horten, Norway). We used a B-mode transducer frequency of 8 MHz during most acquisitions and a Doppler frequency of 5.7 MHz. To minimize the noise level and to decrease the Doppler sample size, the pulse repetition frequency (PRF) was set to 250 Hz. The average Doppler sample size, defined as the spatial distance between the two measurement points along the US beam, was 2.0 ⫾ 0.6 mm (SD). The range of frames in the cine-loops was 97 to 256, with a mean of 168 frames recorded at a speed of 45 to 72 ms/frame. A typical sonogram of the antrum including the color Doppler data is shown in Fig. 1. The average maximum depth of scanning was 6 cm. All selected US cineloops were scanned while the subjects held their breath, and the data were temporarily stored on the scanner before transfer through the ethernet to a PC workstation (Dell Optiplex, Austin, TX). The workstation had a 500-MHz Pentium III processor and 250 MB of RAM. A prototype software application (tvi.exe) was used to calculate strain values from the Doppler data. The strain-rate imaging method The velocity component v of every point in the muscle is available from tissue Doppler data, so the spatial gradient can be estimated from two points along the US beam. This estimate can be performed in realtime as a small extension of the tissue Doppler data processing, and this method was termed strain rate imaging (SRI) (Heimdal et al. 1998; Stoylen et al. 1999). Because only the velocity component along the US beam is available, only the strain rate in the beam direction is estimated (Fig. 2). Subsequently, the strain rate was mapped in segments of the gastric antrum wall. A color scheme using a transition from yellow to red for increasingly more negative strain rate (i.e., shortening) and a transition from cyan to blue for increasingly more positive strain rate (i.e., elongation) was chosen for 2-D and M-mode displays, see Fig. 1. In addition, strain rates near zero were colored green to facilitate the definition of areas and periods of low strain rates. This is
Gastric antral strain ● O. H. GILJA et al.
1459
Fig. 2. The schematic drawing shows how the US beams traverse through the gastric wall. The beam direction is depicted in a normal, contracted and relaxed state of the gastric antrum and its relation to circumferential and radial strain directions.
Fig. 1. (top) Photograph showing the human gastric antrum in a sagittal section with a superimposed color Doppler region for SRI. The image was acquired with a linear-array transducer at 8 MHz B-mode frequency. Different wall layers can be observed in the antrum and the main muscle layer is outlined by a white arrow. The skin, the rectus muscle, the liver and the superior mesenteric vein are also visualized. A red line depicts the distance that is viewed in anatomical M-mode. (bottom) A typical anatomical M-mode registration of SRI is shown, where blue denotes expansion and green denotes compression of tissue. The spikes on the graph correspond to the aortic pulsation.
different from tissue Doppler imaging, where there is a sharp color discontinuity around zero velocity. In a prototype computer postprocessing program, tissue velocities and strain rate values from multiple points along an M-mode cursor line were extracted. In this software, the images showing the velocity of tissue motion were superimposed on the 2-D ultrasonic images for real-time display in color. The velocities and strain rates were also available as numerical values for quantitative analysis. The strain measurements In the postprocessing of data, we averaged our measurements over three samples in the lateral direction to the beam and of five samples in the radial beam direction, corresponding to 1.2 mm in the lateral direc-
tion and 1.0 mm in the radial direction. Gain, color rejection and tissue priority were all optimized before strain estimation started. The speed of the cineloop could be varied, adjusted to improve characterization of details. The strain calculations were done relative to the starting point of the cineloop, defined as the beginning of the contraction, as assessed visually by ultrasonography. Two main options for strain calculations existed; one (strain/motion) based on the application of a single spatial point of measurement onto the image. The other was based on curved M-mode analysis, where a line consisting of up to eight measurement points was drawn along a region-of-interest (ROI). Subsequently, time integrals were calculated either based on the labeled values or the mean or median values between the neighboring points. Ordinary B-mode images of the antrum were also analyzed with respect to measurement of the width of the proper muscle layer in a relaxed and contracted state. Based on these measurements, strain values were computed and compared to SRI data for validation purposes. The antral wall changes position in the image frame during a contraction; thus, inducing a potential error in strain estimation. The software inherits a built-in compensation for such movements because the intended point of measurement can be moved to track the position of the anatomical focus of interest. One SRI analysis of a single cineloop took approximately 1 min. One cineloop was traced 3 times and then the average value was recorded. Distension protocol A specially designed distension probe was constructed. The probe was 120-cm long and contained a
1460
Ultrasound in Medicine and Biology
30-m thick polyesterurethane bag 10 cm from the tip of the probe. The cylindrical bag was 5-cm long and could be inflated with fluid through an infusion channel of 3.5 mm to a diameter of 40 mm without stretching the bag wall. The probe contained a metal clip in the middle of the bag as a marker for US. Two side-holes for pressure measurements were placed in the middle of the bag and 2-cm proximal to the bag. The lumens and side-holes both had diameters of 0.5 mm. The perfusion rate for the pressure channels was 0.1 mL min⫺1. The pressure was measured by a low compliance perfusion system connected to external transducers (Medex, Denmark). The pressure data were amplified and analog-to-digital converted at a sampling rate of 10 Hz using a motility data-acquisition system (Gatehouse Ltd., Nørresundby, Denmark). The digitized data were stored on a PC for later analysis in the same software. After calibrating the measurement system, the tube was passed into the stomach via the nostrils. A manometric investigation was carried out to evaluate the position of the catheter and to confirm that the contractile pattern was consistent with phase 1. The middle of the balloon was then placed approximately 3 cm proximal to the pylorus. The position of the catheter was also visualized by ultrasonography. In one subject, fluoroscopic guidance was necessary to ensure correct position of the balloon. The zero pressure level for the distension series was determined before the distension protocol was performed as a volume-controlled series. The fluid volume of the bag was changed intermittently with volumes ranging from 5 to 60 mL for 1-min periods using a hand-held syringe. The volume infusions were done manually as fast as it was possible to empty the syringe volume. The distensions were separated by 2-min resting periods (i.e., the volume was withdrawn from the bag and the bag pressure was kept slightly negative during these periods to ensure complete emptying of the bag). SRI was performed simultaneously with the infusion of water into the balloon. A mark on the skin, as well as the localization of the metal marker on the probe, ensured that the same area was scanned each time. Variability experiments Two independent observers (O. H. Gilja and T. Hausken) estimated intra- and interobserver variabilities of the strain calculations. A total of 14 data sets were analyzed twice by each observer using the dedicated software program, and the mean result was recorded. Some data sets were utilized more than once and traced at different portions of the gastric wall. Because of relatively high strain variations in the different portions of the wall (see results), the two observers had to agree on which part of the antrum to measure before the actual tracing was performed.
Volume 28, Numbers 11/12, 2002
Fig. 3. The box plot shows the magnitude and the differences in strain (%) of the anterior, inferior and superior part of the muscularis propria in a sagittal section of the gastric antrum. Data were acquired in a radial direction by SRI using Doppler ultrasonography.
Statistical analysis Median values and interquartile ranges were calculated and reported, if not otherwise specifically noted. Coefficient of correlation was computed for intra- and interobserver analysis. Limits of agreement were determined as suggested by Bland and Altmann (1986) to evaluate intra- and interobserver agreement. The level of statistical significance was p ⬍ 0.05. All statistical calculations and graphic designs were performed using commercially available software. Ethical aspects The study was approved by the regional ethic committee and was conducted in accordance with the revised Declaration of Helsinki. All volunteers gave written, informed consent to participate in the trial. RESULTS High-frequency real-time ultrasonography and SRI enabled detailed studies of layers within the gastric wall in all subjects. Most of the antral contractions were obtained in phase II according to ultrasonographic evaluation, but one of the contractions was of phase III origin. The anterior part of the gastric antrum was more accessible for SRI than the other parts of the antral circumference. A high positive radial strain was found in the anterior part of the gastric wall (96%); 163% (median values; interquartile range) during contraction. Negative strain was found in the inferior part (⫺33%, 23%) and in the superior part (⫺25%, 6%) (circumferential strain), see Fig. 3. The posterior part of the antral circumference
Gastric antral strain ● O. H. GILJA et al.
1461
Fig. 5. This scatter plot shows the association between strain measured by Doppler ultrasonography and deformation calculated by radial measurements on the ordinary B-mode images. The line of identity is drawn. Fig. 4. These graphics outline the difference in strain of the circular vs. the longitudinal muscle layer of the antral wall. (top) The exact sampling points in the gastric wall are denoted with a square and an asterix in the ultrasonogram. Blue denotes expansion and green denotes compression of tissue. (bottom) The difference between the high positive strain (elongation) of the circular (inner) muscle layer and the low strain of the longitudinal (outer) muscle layer, illustrates the separate biomechanical events in the two portions of the main muscle layer. The green line corresponds to the red square and the yellow line corresponds to the white asterisk (inner circular muscle layer). The data were obtained in a radial direction.
was frequently difficult to image adequately for SRI analysis due to movement of luminal particles and the long distance from the scan head. Local mapping of radial strain distributions of the anterior part of the muscle layers was enabled with this Doppler-based method. In three individuals, it was possible, due to a combination of high resolution and a slim body habitus, to distinguish between strain in the longitudinal and circular muscle layers. During contractions, radial strain was significantly higher in the circular muscle layer compared to the longitudinal layer, where radial strain was closer to zero (e.g., 105% vs. 10%, see Fig. 4). Measurements of the proper muscle diameters on the B-mode ultrasonograms showed good correlation with SRI measurements (r ⫽ 0.98, p ⬍ 0.05), as visualized in Fig. 5. The mean difference was 0.8% and the limits of agreement were ⫺28.6% to 30.2%. The intraobserver correlation between two observers calculating radial strain using the designed software was
r ⫽ 0.98 (p ⬍ 0.01) (Fig. 6) and the intraobserver agreement was 0.2% ⫾ 18.6% (mean difference ⫾ 2 SD, % strain) (Fig. 7). The interobserver correlation was r ⫽ 0.84 (Fig. 8) and the interobserver agreement was 6.9% ⫾ 56.8% (Fig. 9). However, as observed in Fig. 8, there
Fig. 6. Scatter plot showing the intraobserver correlation of strain in the antral muscle layer using SRI. The line of identity is drawn.
1462
Ultrasound in Medicine and Biology
Fig. 7. Plot displaying the intraobserver agreement of strain calculation in antral wall muscle using SRI. The difference between the two observations (Obs 1a – Obs 1b) is depicted on the y-axis. The average of the two observations is plotted along the x-axis. The intraobserver agreement was 0.2% ⫾ 18.6% (mean difference ⫾ 2 SD, % strain), displaying no proportional error effect.
is an obvious outlier in the data, representing the only phase 3 measurement in this study. A phase 3 contraction has much more complex and dynamic movements compared to phase 2 contractions. If we exclude the outlier of the phase 3 contraction, the correlation coefficient is 0.95 and limits of agreement are ⫺0.3% ⫾ 19.6% for the interobserver variation
Volume 28, Numbers 11/12, 2002
Fig. 9. Interobserver agreement of strain calculation in antral wall muscle using SRI. The difference between the 2 observers (Obs 1 – Obs 2) is depicted on the y-axis. Mean of the two observers is plotted along the x-axis. The interobserver agreement was 6.9% ⫾ 56.8% (mean difference ⫾ 2 SD, % strain). However, if the outlier was excluded, interobserver agreement was ⫺0.3% ⫾ 19.6%.
During distensions, we observed that SRI could discriminate between active and passive deformation of the gastric muscle layer. The regular distension pattern was negative radial strain in the anterior wall in response to sudden pressure increments. However, two other distinct patterns were noted during careful postprocessing of image data; distension on contraction (the balloon was inflated when a barely visible contraction had already started) or contraction on distension (the gastric wall was compressing the balloon undergoing inflation). Accordingly, SRI was able to discriminate between active and passive deformation of the wall, not observed during ordinary 2-D scanning. We found an inverse significant relationship between radial strain estimated by SRI and maximum intraballoon pressure (r ⫽ ⫺0.87, p ⬍ 0.05) during distension (Fig. 10). DISCUSSION
Fig. 8. Scatter plot showing the interobserver correlation of strain in the antral muscle layer by using SRI. The line of identity is drawn.
For the first time, a noninvasive method based on Doppler ultrasonography was applied to estimate strain in moving gastrointestinal tissue in humans. In this pilot study, SRI demonstrated high spatial and temporal resolution facilitating detailed analysis of contracting gastric smooth muscle. Using SRI, it was possible to discern biomechanical events in the inner circular vs. the outer longitudinal muscle layer of the antral wall, a phenomenon not observed during ordinary 2-D scanning. Furthermore, SRI enabled mapping of local strain distribution within a muscle layer showing the existence of significant strain variations within small areas. Gastrointestinal strain is a measure of regional deformation and, by definition, negative strain means shortening and positive strain elongation. SRI is an extension
Gastric antral strain ● O. H. GILJA et al.
Fig. 10. Scatter plot demonstrating the association between radial strain estimated using SRI and volume infusion into a balloon distending the antrum. A significant negative correlation is depicted, showing that, as increasing volume is distending the antrum, an increasing negative strain (shortening) is measured radially in the anterior proper muscle layer of the wall.
of existing tissue Doppler methods and makes measurement of deformation possible (Sutherland et al. 1999). Recently, SRI was used to quantify myocardial movement (Voigt et al. 2000; Stoylen et al. 1999) and to assess coronary artery disease (Stoylen et al. 2000; Belohlavek et al. 2001). In a study in anesthetized dogs, SRI was validated against sonomicrometry as a reference method (Urheim et al. 2000). The authors found that, after placing longitudinal ultrasonic segment-length crystals in contracting tissue, strain by Doppler correlated well with strain by sonomicrometry. They concluded that strain estimation by this novel Doppler ultrasonography represented a powerful method for quantifying regional muscle function, and was less influenced by tethering effects than ordinary Doppler tissue imaging. In our study, we found a significant association between SRI measurements and strain calculated on the basis of metric data obtained directly from the B-mode images. Furthermore, intra- and interobserver variabilities of the strain measurements made by SRI were acceptable. SRI correlated well with pressure increments following distension of the antrum by a balloon, showing that strain measured by the SRI method is associated to the pressure force generated by the antral muscle. Interestingly, we found that the outer longitudinal muscle layer of the antral wall had different strain levels than the inner circular muscle layer during a contraction. When radial strain was measured, the longitudinal layer
1463
exhibited a relatively low negative strain (shortening) and the circular muscle layer had higher positive strain (elongation) (Gregersen et al. 1999b). Previously, the separation of the inner and outer muscle layer has been demonstrated with endoluminal ultrasonography with increasing transducer frequencies (Taniguchi et al. 1993, 1995). In these studies, a miniaturized US device attached to the gastrointestinal mucosa by suction has been used to examine motility of esophageal wall layers using M-mode. In that study, an average of 83% thickening of the inner circular muscle layer was observed during contractions. In the present study, a typical reading is shown in Fig. 4, where the positive strain of the circular layer was 105% and strain of the longitudinal layer was approximately 10%. Accordingly, this study shows that the inner circular layer constitutes the driving force during contraction and that the outer layer passively follows the movements during this sequence of phase II contractile activity of the antral wall muscle. To be able to distinguish between the outer and inner muscle layer of muscularis propria of the antral wall, the scanner was preset to a very small strain sample size (average 2 mm), which determines the spatial resolution of the strain estimates. We also took care in not applying too much manual force with the scan head to avoid pressure effects on the gastric wall (Odegaard et al. 1992). However, when a small sample size was used for calculating the strain rate, the random noise became relatively larger. The demand for high spatial resolution was compromised by lower SNR. To reduce the noise, multiple samples from a longer segment along the Mmode beam were averaged during analysis using dedicated software. Furthermore, by the integration procedure from SR to strain the random noise was considerably reduced, and we obtained relatively smooth strain curves. Generally, SRI is limited by the same factors that hamper routine ultrasonography, namely, bowel gas and excessive abdominal fat. However, for ultrasonography of the fasting gastric antrum, gas is seldom a problem for image quality. In our study, the volunteers had normal body habitus and luminal gas was not a significant problem. Another limitation of SRI is marked angle-dependency, more so than for other Doppler modalities. This is due to the fundamental difference between measurement of fluid velocities where particles move freely, and tissue velocities in solid structures where deformation in one direction is always associated with deformation in other directions to keep the volume of the structure constant. To minimize this problem, we took care in aligning the US beams as perpendicular to the anterior part of the muscle as possible; thus, measuring only the axial strain component. While evaluating intra- and interobserver agreement, we observed relatively large variations in our
1464
Ultrasound in Medicine and Biology
strain estimates. This was due partly to the need for selecting a very small strain sample to be able to distinguish the inner and outer muscle layers. Furthermore, we studied a variety of gastric contractions ranging from early phase 2 to phase 3 and, in future studies, variability can probably be reduced by omitting outliers originating from other phases of the interdigestive motor cycle. The data provided by the SRI technique can likely be used for further mechanical analysis, such as to compute longitudinal strains using incompressibility assumptions and to derive mechanical parameters from pressure-strain relationships. In conclusion, we found that Doppler examination based on SRI and dedicated software can be used to noninvasively obtain biomechanical information of the contracting gastric antrum. Fasting antral contractions have different strain distributions within the muscle layers and the circular and longitudinal muscle layers show different strain levels. Moreover, SRI enables discrimination between active and passive deformation of the gastric muscle layer during distension. The high temporal and spatial resolutions inherent in SRI offer very detailed transabdominal imaging, but further studies are needed to establish the precise role of this method, improve the technique to reduce variance in the recorded strain and to investigate possible clinical applications. Acknowledgments—This study was supported by grants from Innovest Strategic Research Programme, Haukeland University Hospital, Bergen, Norway; the Karen Elise Jensens Foundation; the Scandinavian Association for Gastrointestinal Motility (SAGIM); and the Danish Technical Research Council. The authors are grateful for the technical support from GE Vingmed Ultrasound, Horten, Norway.
REFERENCES Bach DS, Armstrong WF, Donovan CL, Muller DW. Quantitative Doppler tissue imaging for assessment of regional myocardial velocities during transient ischemia and reperfusion. Am Heart J 1996;132:721–725. Belohlavek M, Pislaru C, Bae RY, Greenleaf JF, Seward JB. Real-time strain rate echocardiographic imaging: Temporal and spatial analysis of postsystolic compression in acutely ischemic myocardium. J Am Soc Echocardiogr 2001;14:360–369. Bland JM, Altmann DG. Statistical methods for assessing agreement between two methods of clinical measurements. Lancet 1986;1: 307–310. Bolondi L, Bortolotti M, Santi V, et al. Measurement of gastric emptying time by real-time ultrasonography. Gastroenterology 1985; 89:752–759. Camilleri M, Malagelada JR, Kao PC, Zinsmeister AR. Gastric and autonomic responses to stress in functional dyspepsia. Dig Dis Sci 1986;31:1169–1177. Gilja OH, Hausken T, Odegaard S, Berstad A. Three-dimensional ultrasonography of the gastric antrum in patients with functional dyspepsia. Scand J Gastroenterol 1996;31:847–855. Gregersen H. Residual strain in the gastrointestinal tract: A new concept. Neurogastroenterol Motil 2000;12:411–414. Gregersen H, Barlow J, Thompson D. Development of a computercontrolled tensiometer for real-time measurements of tension in tubular organs. Neurogastroenterol Motil 1999a;11:109–118.
Volume 28, Numbers 11/12, 2002 Gregersen H, Kassab GS. Biomechanics of the gastrointestinal tract. Neurogastroenterol Motil 1996;8:77–97. Gregersen H, Kassab GS, Fung YC. The zero-stress state of the gastrointestinal tract: Biomechanical and functional implications. Dig Dis Sci 2000;45:2271–2281. Gregersen H, Lee TC, Chien S, Skalak R, Fung YC. Strain distribution in the layered wall of the esophagus. J Biomech Eng 1999b;121: 442–448. Grubb NR, Fleming A, Sutherland GR, Fox KA. Skeletal muscle contraction in healthy volunteers: Assessment with Doppler tissue imaging. Radiology 1995;194:837–842. Hausken T, Berstad A. Wide gastric antrum in patients with non-ulcer dyspepsia. Effect of cisapride. Scand J Gastroenterol 1992;27:427– 432. Hausken T, Gilja OH, Odegaard S, Berstad A. Flow across the human pylorus soon after ingestion of food, studied with duplex sonography. Effect of glyceryl trinitrate. Scand J Gastroenterol 1998a;33: 484–490. Hausken T, Gilja OH, Undeland KA, Berstad A. Timing of postprandial dyspeptic symptoms and transpyloric passage of gastric contents. Scand J Gastroenterol 1998b;33:822–827. Hausken T, Goldman B, Leotta DF, Odegaard S, Martin RW, Quantification of gastric emptying and duodenogastric reflux stroke volumes using three-dimensional guided digital color Doppler imaging. Eur J Ultrasound 2001(in press). Hausken T, Odegaard S, Matre K, Berstad A. Antroduodenal motility and movements of luminal contents studied by duplex sonography. Gastroenterology 1992;102:1583–1590. Hausken T, Svebak S, Wilhelmsen I, et al. Low vagal tone and antral dysmotility in patients with functional dyspepsia. Psychosom Med 1993;55:12–22. Heimdal A, Stoylen A, Torp H, Skjaerpe T. Real-time strain rate imaging of the left ventricle by ultrasound. J Am Soc Echocardiogr 1998;11:1013–1019. Kerlin P. Postprandial antral hypomotility in patients with idiopathic nausea and vomiting. Gut 1989;30:54–59. Malagelada JR, Stanghellini V. Manometric evaluation of functional upper gut symptoms. Gastroenterology 1985;88:1223–1231. Odegaard S, Kimmey MB, Martin RW, et al. The effects of applied pressure on the thickness, layers, and echogenicity of gastrointestinal wall ultrasound images. Gastrointest Endosc 1992;38:351– 356. Ophir J, Cespedes I, Ponnekanti H, Yazdi Y, Li X. Elastography: A quantitative method for imaging the elasticity of biological tissues. Ultrason Imaging 1991;13:111–134. Ophir J, Garra B, Kallel F, et al. Elastographic imaging. Ultrasound Med Biol 2000;26(Suppl. 1):S23–S29. Stoylen A, Heimdal A, Bjornstad K, Torp HG, Skjaerpe T. Strain rate imaging by ultrasound in the diagnosis of regional dysfunction of the left ventricle. Echocardiography 1999;16:321–329. Stoylen A, Heimdal A, Bjornstad K, et al. Strain rate imaging by ultrasonography in the diagnosis of coronary artery disease. J Am Soc Echocardiogr 2000;13:1053–1064. Sutherland GR, Kukulski T, Voight JU, D’hooge J. Tissue Doppler echocardiography. Future developments. Echocardiography 1999; 16:509–520. Taniguchi DK, Martin RW, Trowers EA, et al. Changes in esophageal wall layers during motility: Measurements with a new miniature ultrasound suction device. Gastrointest Endosc 1993;39:146–152. Taniguchi DK, Martin RW, Trowers EA, Silverstein FE. Simultaneous M-mode echoesophagram and manometry in the sheep esophagus. Gastrointest Endosc 1995;41:582–586. Uematsu M, Nakatani S, Yamagishi M, Matsuda H, Miyatake K. Usefulness of myocardial velocity gradient derived from two-dimensional tissue Doppler imaging as an indicator of regional myocardial contraction independent of translational motion assessed in atrial septal defect. Am J Cardiol 1997;79:237–241. Urheim S, Edvardsen T, Torp H, Angelsen B, Smiseth OA. Myocardial strain by Doppler echocardiography. Validation of a new method to quantify regional myocardial function. Circulation 2000;102:1158– 1164.
Gastric antral strain ● O. H. GILJA et al. Voigt JU, Arnold MF, Karlsson M, et al. Assessment of regional longitudinal myocardial strain rate derived from doppler myocardial imaging indexes in normal and infarcted myocardium. J Am Soc Echocardiogr 2000;13:588–598.
APPENDIX This appendix describes the concept of strain and strain rate, and how these entities can be estimated from the Doppler velocities. In general terms, strain means relative deformation, and strain rate means rate of deformation. If an object has an initial length L0 that, after a certain time, changes to L, strain is defined as:
e⫽
L ⫺ L0 . L0
1465
change in length per unit length equals the velocity gradient times the time increment: dL 2 ⫺ 1 ⫽ dt. L L
Because it is not feasible to track the end points of the object, the velocity gradient is estimated from two points with a fixed distance: dL 共r ⫹ ⌬r兲 ⫺ 共r兲 ⬇ dt ⫽ SR dt. L ⌬r
(4A)
The velocity gradient estimate is termed strain rate (SR). Finally, by integrating this equation, we arrive at:
冕
(1A) log
The instantaneous change in length (dL) in a small time increment (dt) is related to the velocities (1 and 2) of the end points of the object: dL ⫽ 共 2 ⫺ 1兲dt
(3A)
(2A)
By dividing eqn (2A) with L, we see that the instantaneous
L ⫽ L0
t
SR dt,
(5A)
t0
where “log” denotes the natural logarithm. This gives the following relation between strain rate estimated by velocity gradient and strain:
冉冕 冊 t
e ⫽ exp
SR dt
t0
⫺ 1.
(6A)