Abstract
Ultrasonic backscatter measurements from vertebral bodies (L3 and L4) in 9 women were performed using a clinical ultrasonic imaging system. Measurements were made through the abdomen. The location of a vertebra was identified from the bright specular reflection from the vertebral anterior surface. Backscattered signals were gated to isolate signal emanating from the cancellous interiors of vertebrae. The spectral centroid shift of the backscattered signal, which has previously been shown to correlate highly with bone mineral density (BMD) in human calcaneus in vitro, was measured. BMD was also measured in the 9 subjects’ vertebrae using a clinical bone densitometer. The correlation coefficient between centroid shift and BMD was r = −0.61. The slope of the linear fit was −160 kHz / (g/cm2). The negative slope was expected because the attenuation coefficient (and therefore magnitude of the centroid downshift) is known from previous studies to increase with BMD. The centroid shift may be a useful parameter for characterizing bone in vivo.
Keywords: bone, vertebra, backscatter
INTRODUCTION
Ultrasonic attenuation in bone in vivo is often measured using a through-transmission method at the calcaneus (Langton et al. 1984; Langton and Njeh, 2004; Laugier, 2008). Measurements at other clinically important sites, such as hip and spine, could potentially contain additional useful diagnostic information. Through-transmission measurements are problematic at these sites, however, due to complex bone shapes and the amount of intervening soft tissue.
Pulse-echo measurement has an advantage over through-transmission measurement in that it requires access to only one side of a bone. The spectral properties of the backscattered signal contain useful information regarding the acoustic properties of the scattering medium. For example, since frequency-dependent attenuation in biologic tissues has the effect of a low pass filter, attenuation within a scattering medium (e.g. bone) causes a downshift in the center frequency (i.e. centroid) of the backscattered spectrum. The magnitude of this downshift increases with attenuation coefficient. For a Gaussian pulse (a common form in biomedical ultrasound), the downshift is directly proportional to the attenuation coefficient (Dines and Kak 1978; Narayana and Ophir 1983). Centroid shift from the backscattered signal is an index of attenuation and has been previously used to characterize soft tissues (Fink et al., 1983).
In a previous investigation, centroid shifts from signals backscattered from 30 calcaneal cancellous bone samples in vitro were measured (Wear, 2003). Attenuation slope was also measured using a through-transmission method. The correlation coefficient between backscattered centroid shift and attenuation slope was −0.71 (95% confidence interval: −0.86 to −0.47). This result suggested that the backscattered spectral centroid shift may contain useful diagnostic information. Moreover, unlike through-transmission attenuation measurements, the backscattered centroid shift is potentially applicable to hip and spine, sites commonly associated with osteoporotic fractures.
The objective of this clinical investigation was to investigate the feasibility of characterizing vertebral cancellous bone in vivo from the spectral centroid shift of the backscattered ultrasound signal. We have previously reported methodology for acquisition of backscattered ultrasound data from the spine in vivo (Garra et al. 2002; Wear et al. 2002).
METHODS
A. Subjects
The protocol was approved by the University of Vermont Institutional Review Board. Informed consent was obtained from all 9 women. Volunteer ages ranged from 37 – 69 yr (mean: 49 yr, std. dev.: 9 yr). Volunteer weights ranged from 59 – 116 kg (mean: 82 kg, std. dev.: 22 kg).
B. Data Acquisition
Scanning of the L3 and L4 vertebral bodies was performed through the abdomen, using a GE-Vingmed System Five Ultrasound Scanner with a 2.5 MHz phased-array ultrasound probe. (While ultrasonic measurements on bone are typically performed in the vicinity of 500 kHz – 1 MHz (Langton, 1984; Laugier, 2008), such frequencies are rarely if ever available on clinical imaging systems). The probe was pressed against the skin so that the overlying abdominal tissue was compressed to a standard thickness (either 3 or 6 cm, whichever was feasible for a particular subject—on three occasions, data from both depths were acquired from the same subject). See Figure 1. Since soft tissue behaves as an almost incompressible fluid, the tissue expanded laterally in order to compensate for the longitudinal compression. Therefore, attenuation due to overlying abdominal tissue between transducer and vertebra remained roughly constant at either the 3 or 6 cm level. However, it is conceivable that compression affected the sound speed somewhat so as to create some uncertainty in the true depth of the vertebra. Digitized radio frequency (RF) backscattered ultrasound data were acquired from a region of interest (ROI) just beyond the specular echo from anterior surface of the vertebra. Cortical bone was avoided so that the ROI contained pure cancellous bone. The digitization rate was 20 MHz. Data were acquired in both transverse and longitudinal planes. At least 3 backscatter datasets were acquired in each plane for each subject. Backscatter spectra were averaged over all (at least 6) datasets for each subject. In order to correct for instrument gain settings, focusing effects, and diffraction effects, RF data were also acquired from a calibration phantom (with attenuation coefficient matched to liver), scanned immediately after the human subject using the same instrument gain settings. The position and size of the ROI were identical for human and phantom data acquisition. The average ROI width was 1.3 cm. ROI lengths ranged from 0.5 to 2.0 cm. (Another candidate calibration measurement would have been the specular echo off the anterior surface of the vertebra. However, the phantom measurement allowed closer spatial registration between tissue and calibration measurements.) Areal (i.e. projected) bone mineral density (BMD) in units of gm/cm2 was also measured using a GE-Lunar Prodigy fan-beam whole body DEXA system.
Figure 1.
Data acquisition.
C. Data Analysis
RF lines were gated to isolate the proximal 0.5 cm from the ROI (immediately distal to the specular echo from the anterior surface of the vertebra). Power spectra were computed as the average squared modulus of the FFT’s of all gated RF lines within each ROI. The centroid shift was computed as the difference of the centroids of the vertebral and phantom spectra.
RESULTS
Figure 2 shows a B-mode image. The bright crescent corresponds to the specular echo from the anterior surface of the vertebra. The specular echo was used to localize the ROI for RF data acquisition.
Figure 2.
Ultrasound image. The bright crescent is the proximal surface of a vertebra.
Figure 3 shows average power spectra from one vertebra and the reference phantom. The vertebral backscattered spectrum has a higher magnitude than the phantom spectrum because cancellous bone has a higher backscatter coefficient than the soft-tissue-mimicking phantom. The centroid from the vertebra was lower than that of the phantom because of the high attenuation within cancellous bone.
Figure 3.
Spectra from vertebra and phantom. Centroids are marked with dotted lines.
Figure 4 shows centroid shift vs. BMD for acquisitions from a depth of 3 cm. From measurements from 11 vertebrae in 7 women, the correlation coefficient between spectral centroid shift and BMD was r = −0.61 (95% CI: −0.90 – 0.00). The slope of the linear fit was −160 kHz / (g/cm2). The negative slope was expected because the attenuation coefficient (and therefore magnitude of the centroid downshift) is known from previous studies to increase with BMD.
Figure 4.
Centroid shift vs. BMD in 11 vertebrae for acquisitions from a depth of 3 cm. Two points at (0.87, −240) coincide.
Figure 5 shows centroid shift vs. BMD for acquisitions from a depth of 6 cm. From measurements from 7 vertebrae in 5 women, the correlation coefficient between spectral centroid shift and BMD was r = −0.44 (95% CI: −0.90 – 0.49) (not statistically significantly different from 0). The slope of the linear fit was −80 kHz/(g/cm2).
Figure 5.
Centroid shift vs. BMD in 7 vertebrae for acquisitions from a depth of 6 cm.
CONCLUSION
Acquisition of ultrasonic backscatter data from human spine in vivo is feasible. The spectral centroid shift from vertebrae exhibits a moderate negative correlation with BMD in vivo. The centroid shift may be a useful parameter for characterizing bone in vivo. Clinical utility will depend on reproducibility of this measurement and on correlation with fracture risk.
ACKNOWLEDGEMENTS
The assistance of Jonathan Mai is acknowledged. The mention of commercial products, their source, or their use in connection with material reported herein is not to be construed as either an actual or implied endorsement of such products by the U.S. Food and Drug Administration. This project was supported by the Vermont General Clinical Research Center, NIH Grant RR 00109, and the FDA Office of Women’s Health.
REFERENCES
- Dines KA, and Kak A, Signal processing of broad band pulses ultrasound measurements of attenuation in soft tissues, IEEE Trans. Biomed. Eng, 1978; BME-25, 321–44. [DOI] [PubMed] [Google Scholar]
- Fink M, Hottier F, and Cardoso JF, Ultrasonic signal processing for in vitro attenuation measurement: short time Fourier analysis, Ultrasonic Imag. 1983; 5, 117–35. [DOI] [PubMed] [Google Scholar]
- Garra BS, Wear KA, Pinet MC, Felker S, and Mai J, Measurements of Bone Mineral Density and Backscattered Ultrasound in the Spine and Hip In Vivo– Methodology & Preliminary Results, Proc. 27th International Symp. On Ultrasonic Imaging and Tissue Characterization, Arlington, VA, June 3–5, 2002, Ultrasonic Imag. 23(3), 179–180 (2001). [Google Scholar]
- Langton CM, Palmer SB, and Porter RW, “The measurement of broadband ultrasonic attenuation in cancellous bone.” Eng. in Med 1984; 13, 89–91. [DOI] [PubMed] [Google Scholar]
- Langton CM and Njeh CF, The physical measurement of bone. Bristol, UK: Institute of Physics Publishing, 2004. [Google Scholar]
- Laugier P, “Instrumentation for in vivo ultrasonic characterization of bone strength,” IEEE Trans Ultrason., Ferro., and Freq. Cont, in press., 2008. [DOI] [PubMed] [Google Scholar]
- Narayana PA, and Ophir J, A closed form method for the measurement of attenuation in nonlinearly dispersive media, Ultrason. Imag, 1983; 5, 17–21. [DOI] [PubMed] [Google Scholar]
- Wear KA, Characterization of Trabecular Bone Using the Backscattered Spectral Centroid Shift, IEEE Trans. Ultrason., Ferro. Freq. Cont, 2003; 50, 402–407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wear KA, Garra BS, Pinet MC, Felker S, and Mai J, Measurements of Ultrasonic Backscattered Spectral Centroid Shift from Spine In Vivo – Methodology & Preliminary Results, 24th annual conference Amer. Soc. Bone. & Miner. Res., San Antonio, TX, Sept. 20 – 24, 2002, J. Bone & Mineral Research, 17, suppl. 1, p. S 419, 2002. [Google Scholar]