Characterization of Center Frequency and Bandwidth of Broadband Ultrasound Reflected by the Articular Cartilage to Subchondral Bone Interface

Ultrasound in Med. & Biol., Vol. 37, No. 1, pp. 112–121, 2011 Copyright Ó 2011 World Federation for Ultrasound in Medicine & Biology Printed in the USA. All rights reserved 0301-5629/$ - see front matter

doi:10.1016/j.ultrasmedbio.2010.10.015

d

Original Contribution CHARACTERIZATION OF CENTER FREQUENCY AND BANDWIDTH OF BROADBAND ULTRASOUND REFLECTED BY THE ARTICULAR CARTILAGE TO SUBCHONDRAL BONE INTERFACE SIMO SAARAKKALA,*yz SHU-ZHE WANG,* YAN-PING HUANG,* JUKKA S. JURVELIN,y and YONG-PING ZHENG*x * Department of Health Technology and Informatics, Hong Kong Polytechnic University, Hong Kong, China; y Department of Physics and Mathematics, University of Eastern Finland, Kuopio, Finland; z Department of Diagnostic Radiology, University of Oulu, Oulu, Finland; and x Research Institute of Innovative Products and Technologies, Hong Kong Polytechnic University, Hong Kong, China (Received 21 September 2009; revised 6 October 2010; in final form 12 October 2010)

Abstract—Osteoarthritis (OA) produces degenerative changes both in articular cartilage and subchondral bone. During OA, reflection of high frequency ultrasound from the cartilage-bone interface is affected by both changes in attenuation of the cartilage layer and acoustic properties of the interface. The objective of this study was to experimentally investigate the spectral content of ultrasound reflection from the cartilage-bone interface. Specifically, we analyzed the center frequency and –6 dB bandwidth of the broadband high-frequency (40 MHz) ultrasound signal. Intact bovine articular cartilage samples with and without the underlying subchondral bone (n 5 6) were measured in vitro using a commercial high-frequency ultrasound scanner. Furthermore, the diagnostic potential of the measurement of center frequency and bandwidth for OA was studied with another series of bovine articular cartilage samples (n 5 40) after enzymatic degradations of tissue proteoglycans and collagen. Compared with the reference spectrum at the same depth from a perfect reflector, a major downshift (.51%) of the center frequency and a reduction (.42%) of the bandwidth were observed in both sample groups when analyzing the ultrasound reflection from the cartilage-bone interface. The results suggest that attenuation in the cartilage layer primarily controls the observed downshift of the center frequency and acoustic properties of the subchondral bone play only a minor role in affecting the spectrum of the cartilage-bone interface. Changes in the ultrasound bandwidth of the cartilage-bone interface signals, compared with reference signals, were found to vary more than those in the center frequency in both cartilage sample groups. Compared with pretreatment values, a significant downshift in center frequency (p , 0.01) and a minor reduction in bandwidth of spectra from the cartilage-bone interface were recorded after chemical degradation of proteoglycans with trypsin. In contrast, center frequency and bandwidth of the echoes from the cartilage-bone interface did not change after the chemical degradation of cartilage collagen fibrils. The results suggest that proteoglycan loss, typical to OA, may be detected via the changes in the center frequency of the ultrasound reflected from the cartilage-bone interface. (E-mail: [email protected], [email protected]) Ó 2011 World Federation for Ultrasound in Medicine & Biology. Key Words: Quantitative ultrasound, Articular cartilage, Subchondral bone, Frequency domain analysis, Center frequency, Bandwidth, Osteoarthritis, Collagenase, Trypsin.

initial stage of OA includes an increase of the cartilage surface roughness (Minns et al. 1977; Saarakkala et al. 2006), probably associated with the reduction of collagen content and changes in the organization of the collagen network. Furthermore, depletion of cartilage tissue proteoglycans, especially in the superficial layer, is known to be present in early OA (Buckwalter and Martin 1995). The initial OA changes in subchondral bone include remodeling and sclerosis, as well as osteophyte formation (Buckwalter and Martin 1995; Day et al. 2004). The sensitivity of current diagnostic techniques of OA, i.e.,

INTRODUCTION Osteoarthritis (OA) is a prevalent joint disorder mainly characterized by the degenerative changes in articular cartilage and subchondral bone. In articular cartilage, the

Address correspondence to: Simo Saarakkala, Ph.D., Yong-Ping Zheng, Ph.D., Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong SAR, P.R.China. E-mails: [email protected] (S. Saarakkala), [email protected] (Y.-P. Zheng) 112

Ultrasound spectral analysis of cartilage-bone interface d S. SAARAKKALA et al.

X-ray imaging, arthroscopy and magnetic resonance imaging, is inadequate for detecting these early changes. Consequently, OA is typically diagnosed at an advanced stage, where articular cartilage and subchondral bone have already undergone irreversible damage. The development of more sensitive diagnostic techniques would allow earlier diagnosis of OA and, also, enable the accurate follow-up of articular cartilage after treatment, e.g., after cartilage repair surgery. Quantitative ultrasound imaging (QUI) is a promising method for the characterization of early OA changes in articular cartilage. It has been reported that ultrasound reflection from the proximal cartilage surface decreases as a function of the superficial cartilage degradation in early OA, i.e., when increased surface roughness and degeneration of superficial collagen fibrils take place (Cherin et al. 1998; Hattori et al. 2004; Jaffre et al. 2003; Saarakkala et al. 2006; Saied et al. 1997; Spriet et al. 2005). Furthermore, the intensity of ultrasound backscattering from the internal cartilage layers has been demonstrated to decrease during the development of OA (Cherin et al. 1998). In contrast to collagen degeneration, no change in the intensity of the ultrasound reflection from the proximal surface has been reported after the experimental degradation of cartilage proteoglycans (Pellaumail et al. 2002; Saarakkala et al. 2004). Backscattering intensity from the internal layers is also maintained after proteoglycan degradation (Pellaumail et al. 2002). Thus, it can be anticipated that the ultrasound reflection from the proximal cartilage surface and backscattering from the internal tissue are mainly characterized by the surface roughness, collagen network orientation and collagen content. Ultrasound reflection from the interface between the articular cartilage and the subchondral bone has been also investigated in previous studies. In contrast to ultrasound reflection from the proximal cartilage surface and internal layers, it has been demonstrated that echogenicity of the cartilage-bone interface visually increases in ultrasound images during OA (Disler et al. 2000; Laasanen et al. 2003, 2006; Saied et al. 1997; Spriet et al. 2005). In quantitative analysis, amplitude of the ultrasound reflection from the cartilage-bone interface has also been reported to increase in OA (Jaffre et al. 2003; Saarakkala et al. 2006). It may be proposed that this increase arises from the changes in subchondral bone related to OA, e.g., stiffening of the bone due to sclerosis. On the other hand, the attenuation properties of the overlying cartilage layer are also known to change in OA (Agemura et al. 1990; Joiner et al. 2001; Nieminen et al. 2004; Senzig et al. 1992). This change in attenuation, if not compensated, also affects the reflection amplitude from the cartilage-bone interface. The relative influence of OA-related changes in cartilage

113

and subchondral bone on the frequency content of the ultrasound echo reflected from the cartilage-bone interface has not been characterized. It is well known that ultrasound propagation in human tissues is frequency dependent. Generally, the higher frequencies are attenuated more by the tissues than the lower frequencies. Due to the frequency dependent attenuation, the reflected or backscattered broadband ultrasound pulse experiences a downshift in its center frequency and a reduction in the bandwidth. The extents of these spectral modifications are related to the attenuation characteristics in the tissue (Narayana and Ophir 1983; Narayana et al. 1984; Treece et al. 2005). In the case of articular cartilage, a significant downshift of the center frequency is usually observed with high-frequency (.20 MHz) broadband ultrasound signals reflected from the cartilage-bone interface, compared with a perfect reflection at the same distance after a free propagation. In the current literature, there is only one study investigating the frequency content of ultrasound reflection from the cartilage-bone interface (Brown et al. 2007). However, in that study no significant change of the center frequency of the reflected signals from the proximal and cartilage-bone interfaces was reported after enzymatic degradations. This may be related to the lower ultrasound frequency (10 MHz) and the fact that the spectral changes were only qualitatively depicted (Brown et al. 2007). Up to now, there are no studies focusing on the main factors behind the relatively large downshift of center frequency of the ultrasound reflection from the cartilage-bone interface at higher frequencies. The main goal of this study was to experimentally investigate the center frequency downshift and reduction of bandwidth of broadband high-frequency (40 MHz) ultrasound reflected from the cartilage-bone interface, compared with the reference signal collected at the same distance from the perfect reflector. In principle, the center frequency can be downshifted both by the variable acoustic properties of the cartilage-bone interface and by the attenuation of the overlying cartilage layer. Therefore, the first aim of this study was to find out which factor dominates the change in the center frequency and bandwidth of the signal reflected from the cartilage-bone interface. Furthermore, the diagnostic potential of these frequency domain parameters (center frequency and bandwidth) was studied by analyzing the downshift and bandwidth before and after enzymatic degradations of cartilage tissue proteoglycans and collagen network, respectively. MATERIALS AND METHODS Two different series of bovine articular cartilage samples were used in the study (Fig. 1).

114

Ultrasound in Medicine and Biology

Volume 37, Number 1, 2011

a

b

Fig. 1. (a) The processing protocol for the first cartilage sample series. Ultrasound echoes were first acquired from the interface between cartilage and the subchondral bone. Subsequently, the overlying cartilage layer was carefully removed with a sharp razor blade and the ultrasound echoes were acquired from the distal interface of the cartilage. (b) The processing protocol for the second cartilage sample series. Ultrasound echoes were first acquired from the interface between cartilage and the subchondral bone. Subsequently, samples were exposed to enzymatic degradations with collagenase or trypsin enzymes and ultrasound echoes from the cartilage-bone interface were again recorded.

Series I: Cartilage samples with and without subchondral bone Cylindrical osteochondral samples (n 5 6, diameter 5 6.35 mm, thickness z 4.5 mm) without visible lesions were prepared from the medial upper quadrant of mature bovine patellae (Fig. 1a). Patellae were obtained from a local market and stored in a freezer (–20 C) before preparation. The distal surface of the subchondral bone was ground to be parallel to the proximal cartilage surface and the cartilage-bone plugs were fixed onto the metallic plate using a brand of versatile, reusable putty-like pressure-sensitive adhesive Blu-Tack (Bostik, Thomastown, Australia). The metallic plate with the attached sample was then positioned on the bottom of the container filled with physiologic saline (Fig. 2a). In the first phase, the samples were scanned with the ultrasound system by positioning the proximal cartilage surface at the focal distance. Altogether six two-dimensional (2-D) scans were conducted for each sample. The first three scans were performed in parallel to one direction perpendicular to the cylinder axis, with an interval distance of 0.5 mm and the central scan passing through the center of the disk. Then, another three scans with the same pattern were repeated by rotating an angle of 90 with respect to the previous direction (Fig. 2b). The ultrasound reflection signals from the cartilage-bone interface were analyzed. In the second phase, the overlying cartilage layer was carefully removed with a sharp razor blade. The cartilage

plugs without the bone were again attached onto the polished steel plate. Cartilage samples were then scanned similarly as in the first phase and the ultrasound reflections from the distal cartilage-saline interface were analyzed. Series II: Enzymatically degraded cartilage samples The second cartilage sample series consisted of cylindrical bovine cartilage-bone plugs (n 5 40, diameter 5 6.35 mm, thickness z 4.5 mm) from the medial upper quadrant of patella (Fig. 1b). This sample set was used also in our recent study (Wang et al. 2010). Half of the samples (n 5 20) were exposed to collagenase digestion. Collagenase is known to effectively degrade the collagen network in the cartilage tissue (Shingleton et al. 1996). Another half of the samples (n 5 20) were exposed to trypsin digestion. Trypsin is known to effectively digest proteoglycans in the cartilage tissue and it also has a slight simultaneous effect on the collagen network (Harris et al. 1972). The sample processing for enzymatic degradations has been described with more details in our earlier study (Wang et al. 2008). The samples were scanned with the ultrasound system before and after enzymatic degradations. Osteochondral blocks were fixed at the bottom of the container using the Blu-Tack and the container was filled with physiologic saline (Fig. 2a). Altogether 12 scans were performed for each sample: four different directions in the horizontal plane with an interval of 45 and in each

Ultrasound spectral analysis of cartilage-bone interface d S. SAARAKKALA et al.

a

b

Fig. 2. (a) Schematic presentation of the ultrasound measurement geometry. Osteochondral sample (series I and II) or cartilage layer (series I) was attached on the polished steel plate using the Blu-Tack. Before the measurements, the transducer was carefully tilted to different angles to obtain the maximum amplitude for the reflection signal. (b) Scanning scheme for the cartilage samples (viewed from upside). In the series I, six scans were conducted at two perpendicular directions in the horizontal plane (0 and 90 ), including three different parallel scan lines with an interval of 0.5 mm in each direction. In the series II, 12 scans were performed for each sample: four different directions in the horizontal plane with an interval of 45 and in each direction three different parallel scans with an interval of 0.5 mm.

direction three different parallel scans with an interval of 0.5 mm were recorded (Fig. 2b). It was verified that each stored scan contained the information of reflected echoes from the saline-cartilage and cartilage-bone interfaces. During the ultrasound measurements, the proximal cartilage surface was positioned in the focal zone. The ultrasound reflection signals from the cartilage-bone interface were analyzed. Reference measurements After the cartilage measurements, extensive measurements from the polished steel plate immersed in physiologic saline were conducted with the same ultrasound system to eliminate the variation in center frequency and bandwidth due to defocusing. These

115

measurements from the steel plate provided references to obtain uniform and frequency-independent specular reflections at a certain distance. The reference measurements were conducted by placing the steel plate at 30 different distances from the transducer. These distances ranged from 3.2 to 6.1 mm with a step of 0.1 mm and centered approximately at 4.5 mm, which is the focal length of the current transducer. For the signal collected at each distance, a fast fourier transform (FFT) was performed to obtain its amplitude spectrum in the frequency domain. From this spectrum, the center frequency and bandwidth were determined and used as references for this distance. Ultrasound instrumentation and data acquisition The commercial high-frequency ultrasound scanner (Vevo 770, Visualsonics, Inc., Toronto, CA) equipped with RMV-708 scan head (–6 dB bandwidth 5 18–55 MHz, focal length 5 4.5 mm) was used in the present study. During the experiment, the position of the transducer could be adjusted vertically to move the focal point to a desired location. Radio-frequency (RF) signals were collected from the RF output and digitized at a sampling frequency of 420 MHz by the ultrasound system. The region-of-interest (ROI), i.e., the area in which the digitized RF signals were collected and analyzed, was chosen by placing a window on the software interface. The width of ROI was approximately 1.0 mm containing 100 lines of RF signals and the length of ROI was approximately 3.0 mm, which was generally large enough to cover those parts of the signals reflected from the proximal surface and cartilage-bone (or distal in series I) interfaces. Stored RF signals were later analyzed using a custom-designed Matlab script (Matlab R2007A, Mathworks Inc., Natick, MA, USA). Analysis of spectral data The first step to obtain the spectra from reflections of the cartilage interfaces and reference steel plate was to locate the central position of those reflections. The position was located by searching the local maximum of the envelope at a region nearby those interfaces. The envelope was computed by Hilbert-transforming each RF signal and calculating the absolute values of the transformed signal. The region to search the local maximum was defined by manually placing a 0.71 ms long window that included the interface. After the central position was obtained, it was used as the center of a rectangular window (series I and reference measurement) or Hamming window (series II) of 0.71 ms to obtain the amplitude spectra in the frequency domain for the cartilage interfaces or reference steel plate. In biomedical ultrasound, the broadband pulses are typically Gaussian shaped in the frequency domain. As

116

Ultrasound in Medicine and Biology

Volume 37, Number 1, 2011

deviation of the peak. Thus, by fitting a Gaussian function to the measured amplitude spectrum, one obtains the center frequency directly from the fitting parameter f0 and the bandwidth can be calculated from the standard deviation (s) as follows: pffiffiffiffiffiffiffiffiffi (2) B26dB 5 2 2ln2s;

Fig. 3. (a) Representative frequency spectrum of the ultrasound signal reflected from the cartilage-bone interface (solid line) and the corresponding Gaussian fit (dotted line). Data are from the cartilage sample in the series II before trypsin degradation. (b) Reference spectrum from the steel plate recorded at the same distance as the cartilage-bone interface in A (solid line) and the corresponding Gaussian fit (dotted line). The downshift of the center frequency and decrease in bandwidth can be observed in comparison with A. a.u. 5 arbitrary unit.

this was also the case with the current ultrasound system, we modelled the ultrasound pulse to be of Gaussian shape. Narayana and Ophir (1983) have demonstrated that an ultrasound pulse with Gaussian shape propagating in a lossy medium suffers a center frequency downshift and a decrease in its bandwidth. They also showed that the center frequency downshift and bandwidth decrease are related to the attenuation parameters of the medium (Ophir and Jaeger 1982). By fitting a Gaussian function to the measured spectra, the center frequency and bandwidth were determined from the amplitude spectra (Fig. 3). In the frequency domain, the Gaussian function (A) is presented as: ! ðf 2f0 Þ2 Aðf Þ 5 aexp 2 ; (1) 2s2 where f is frequency, a is a scaling parameter, f0 is the position of the center of the peak and s is the standard

where B26dB is the –6 dB bandwidth of the signal. The fitting was conducted using the Matlab’s fminsearch function. In the series II, four samples (three in collagenase group and one in trypsin group) had to be removed from the study since those samples were so thick that the cartilage-bone interface could not be completely fit in the stored ROI. For each sample, the center frequency and bandwidth were obtained from the fit. When determining the distance between the ultrasound transducer and cartilage-bone interface (series I and II) or distal cartilage-saline interface (series I), the different speeds of sound in saline and cartilage were taken into account. The speed of sound values used in the calculations were taken from the literature: 1495 m/s, 1610 m/s, 1595 m/s and 1580 m/s for physiologic saline (Saarakkala et al. 2004), for healthy cartilage (Laasanen et al. 2002), for trypsin digested cartilage (Laasanen et al. 2002) and for collagenase digested cartilage (Laasanen et al. 2002), respectively. Fitting of the Gaussian function was also conducted for each reference spectrum (Fig. 3b). Consequently, for each sample the relative changes in the center frequency and bandwidth, compared with the reference spectrum at the corresponding distance, were calculated as follows:

Rf 5

fcartilage 2fsteel Bcartilage 2Bsteel ,100% and RB 5 ,100%; fsteel Bsteel (3)

where R represents the relative change in the center frequency or bandwidth, fcartilage or Bcartilage is the center frequency or bandwidth measured from the cartilagebone interface (series I and II) or distal cartilage-saline interface (series I) and fsteel or Bsteel is the center frequency or bandwidth measured from the steel plate at a distance corresponding to that of cartilage measurements. In our earlier study, integrated backscattering (IBS) for the internal cartilage tissue and apparent integrated backscattering from the cartilage-bone interface (AIBbone) were determined before and after enzymatic degradations for series II cartilage samples (Wang et al. 2010). IBS indicates the strength of acoustic backscattering from the internal cartilage and AIBbone reflects the strength of reflection from the cartilage-bone interface after attenuation of the overlying cartilage layer. As they both

Ultrasound spectral analysis of cartilage-bone interface d S. SAARAKKALA et al.

potentially reflect the tissue degeneration state, in the present study we correlated them with the relative changes in the center frequency or bandwidth, respectively, and reported the results. Statistical tests In the series I, a nonparametric Wilcoxon signed rank (1-tailed) test was used to compare the relative change of the center frequency and bandwidth between the reflection from the cartilage-bone interface and the reflection from the distal cartilage-saline interface. In the series II, a paired sample t-test was used to compare the relative change of the center frequency and bandwidth from the cartilage-bone interface before and after collagenase or trypsin digestion. Pearson’s linear correlation analysis was used when obtaining the relations between IBS or AIBbone and Rf or RB. RESULTS Series I: Cartilage samples with and without subchondral bone The center frequency and bandwidth of ultrasound spectra reflected from the cartilage-bone interface (with subchondral bone attached) or the distal cartilage-saline interface (subchondral bone detached) are presented in Table 1. The spectra from the cartilage-bone interface exhibited a major downshift of the center frequency and a reduction in the bandwidth, compared with those obtained from a steel plate at the same depth (upper part of the table). After removing the underlying subchondral bone (lower part of the table), the relative change of the center frequency was significantly smaller than that with the bone attached (–53.7% 6 6.4% vs. –58.6% 6

117

Table 1. The mean values (6 SD) of center frequency (f0) and bandwidth (B-6dB) of ultrasound spectrum (n 5 6, series I) reflected from the cartilage-bone interface (with subchondral bone attached) or the distal cartilage-saline interface (subchondral bone detached) f0 (MHz)

B26dB (MHz)

Cartilage-bone interface: Reference spectra: Relative change (%):

13.8 6 1.3 33.3 6 1.3 258.6 6 3.6

18.5 6 3.4 32.3 6 0.7 242.8 6 10.9

Distal cartilage-saline interface: Reference spectra: Relative change (%):

15.9 6 2.6 34.3 6 1.7 253.7 6 6.4*

17.9 6 3.9 32.4 6 1.2 244.9 6 11.9

Relative change of the center frequency and bandwidth, compared with the reference spectrum from the polished steel plate at the corresponding distance, was calculated individually for each sample and the mean values are compared. * p , 0.05 (Wilcoxon signed rank test, 1-tailed), compared with the change before removal of the subchondral bone.

3.6%, p , 0.05). With regard to bandwidth of the reflected ultrasound signal from the cartilage-bone interface or distal cartilage-saline interface, no significant changes were observed between the samples with the subchondral bone attached and the samples with the bone removed (–44.9% 6 11.9% vs. –42.8% 6 10.9%, p 5 0.422). Series II: Enzymatically degraded cartilage samples B-mode images of a representative cartilage sample before and after trypsin degradation are presented in Figure 4. Only slight and nonconclusive changes in reflection amplitude from the cartilage-bone interface after degradation can be visualized from B-mode images. With regard to quantitative spectral analysis, the center frequency and bandwidth of the ultrasound spectra from the cartilage-bone interface in the samples before and

Fig. 4. B-mode images of a representative cartilage sample (series II) before and after trypsin degradation. Only slight and nonconclusive changes in reflection amplitude from the cartilage-bone interface after degradation, but no corresponding spectral changes, can be visualized from B-mode images.

118

Ultrasound in Medicine and Biology

Table 2. The mean values (6SD) of center frequency (f0) and bandwidth (B26dB) of ultrasound spectrum (n 5 19, series II) reflected from the cartilage-bone interface before and after trypsin degradation

Volume 37, Number 1, 2011

f0 (MHz)

B26dB (MHz)

Before trypsin degradation: Reference spectrum: Relative change (%):

15.9 6 2.3 33.7 6 1.5 252.9 6 6.2

16.2 6 2.8 32.9 6 1.1 250.7 6 8.8

When comparing the present spectral results to the values of other ultrasound parameters reported earlier for the same samples (Wang et al. 2010), i.e., integrated backscattering in the middle part of the cartilage (IBS) and apparent integrated backscattering from the cartilage-bone interface (AIBbone), significant linear correlations (r $ 0.359, p # 0.002) were observed between the relative change in the center frequency or bandwidth and IBS or AIBbone (Fig. 5). Generally, correlations were higher for AIBbone than IBS (Fig. 5).

After trypsin degradation: Reference spectrum: Relative change (%):

14.2 6 2.0 32.9 6 1.4 256.9 6 5.7*

15.3 6 3.0 34.4 6 2.6 255.3 6 9.6

DISCUSSION

Spectrum from the cartilagebone interface

Relative change of the center frequency and bandwidth, compared with the reference spectrum from the polished steel plate at the corresponding distance, was calculated individually for each sample and the mean values are compared. * p , 0.01 (Paired sample t-test), compared with the change before the trypsin digestion.

after proteoglycan and collagen degradation are presented in Tables 2 and 3, respectively. The relative change in the center frequency after trypsin digestion was significantly larger than before the digestion (–56.9% 6 5.7 % vs. –52.9% 6 6.2 %, p , 0.01). Furthermore, the relative change in the bandwidth after trypsin digestion was larger than before the digestion (–55.3% 6 9.6 % vs. –50.7% 6 8.8 %), but this change was not statistically significant (p 5 0.153). After collagenase digestion, the relative change in the bandwidth was slightly smaller than that before the digestion (–54.1% 6 13.6% vs. –56.5% 6 10.9 %), but standard deviations were large and the change was not statistically significant (p 5 0.163). There was no significant difference in the relative change of the center frequency after collagenase digestion (–51.3% 6 8.5 % vs. –52.0% 6 7.5%, p 5 0.577). Table 3. The mean values (6SD) of center frequency (f0) and bandwidth (B26dB) of ultrasound spectrum (n 5 17, series II) reflected from the cartilage-bone interface before and after collagenase degradation Spectrum from the cartilagebone interface f0 (MHz) Before collagenase degradation: Reference spectrum: Relative change (%): After collagenase degradation: Reference spectrum: Relative change (%):

15.6 6 2.5 32.5 6 0.9 252.0 6 7.5 15.7 6 2.7 32.3 6 0.3 251.3 6 8.5

B26dB (MHz) 15.9 6 3.2 36.9 6 3.0 256.5 6 10.9 16.3 6 3.9 36.1 6 2.9 254.1 6 13.6

Relative change of the center frequency and bandwidth, compared with the reference spectrum from the polished steel plate at the corresponding distance, was calculated individually for each sample and the mean values are compared.

Series I: Cartilage samples with and without subchondral bone In the present study, it was observed that the overlying cartilage, mainly through its attenuation, is the dominant factor causing the large downshift of the center frequency of the reflected ultrasound signal from the cartilage-bone interface. The smaller downshift of the center frequency without the underlying subchondral bone may suggest that the acoustic properties of the subchondral bone mildly increase the downshift. However, it should be noted that a manual complete removal of the cartilage layer from the subchondral bone is challenging. To estimate the success of the manual removal, the thickness of the cartilage layer before and after removal was also calculated from the ultrasound data (1.72 6 0.23 mm vs. 1.57 6 0.30 mm). The mean thickness of remaining cartilage after removal was approximately 150 mm. Consequently, a smaller cartilage thickness might also decrease the attenuation and explain the observed smaller downshift of the center frequency after removing the bone. With regard to bandwidth of the reflected ultrasound signal from the cartilage-bone interface, the removal of the subchondral bone induced only minor and insignificant differences to the relative change of the bandwidth. Furthermore, a relatively large standard deviation in the bandwidth showed that this parameter includes more variation and experimental uncertainty than that in the center frequency and it cannot be regarded to be as sensitive an indicator of attenuation properties as the shift in center frequency. In our earlier study, we reported an insignificant correlation between the subchondral bone mineral density and ultrasound reflection amplitude from the cartilage-bone interface in a small sample population (n 5 10) including both normal and osteoarthritic bovine cartilages (Laasanen et al. 2005). Based on this, we concluded that the reflection amplitude from the cartilage-bone interface may be more controlled by the roughness and density of the acoustic interface between the deep layer and calcified cartilage, or the immediate cartilage-bone interface, than the structure and

Ultrasound spectral analysis of cartilage-bone interface d S. SAARAKKALA et al.

119

Fig. 5. (a) Relation between the relative change in the center frequency (Rf) and integrated backscattering in the middle part of the cartilage (IBS). (b) Relation between the relative change in the bandwidth (RB) and IBS. (c) Relation between the relative change in the center frequency (Rf) and apparent integrated backscattering from the cartilage-bone interface (AIBbone). (d) Relation between the relative change in the bandwidth (RB) and AIBbone. Graphs include samples in the series II before and after enzymatic degradations and the data for IBS and AIBbone were extracted from our earlier publication (Wang et al. 2010).

composition of the subchondral bone (Laasanen et al. 2005). The results of the present study also suggest that the subchondral bone plays only a minor role in the observed major frequency downshift of the ultrasound reflection. Thus, it can be hypothesized that ultrasound reflection from the cartilage-bone interface carries more information of the cartilage layer attenuation properties, i.e., absorption and scattering, than that of the subchondral bone acoustic properties. This raises a new question about which cartilage layer is mainly responsible for the frequency downshift. This issue could be investigated in the future by measuring the frequency-dependent amplitude attenuation and center frequency downshift for separated cartilage layers at different depths. Series II: Enzymatically degraded cartilage samples A statistically significant downshift of the center frequency and a decrease of the bandwidth (although not statistically significant) after trypsin digestion were simultaneously observed. As the enzymatic digestions for the full osteochondral blocks should not cause any changes in the cartilage-bone interface, an observed increase in attenuation after trypsin digestion should originate in the overlying cartilage layer. One possible reason might be the increase of cartilage attenuation after trypsin digestion. The attenuation model used in the present study takes into account all attenuating phenomena occurring before the cartilage-bone interface. These include

reflection from the proximal cartilage surface, reflection and scattering from the internal structures of the cartilage and reflection and scattering from the cartilage-bone interface. These individual effects cannot be directly separated from each other and they altogether contribute to the observed center frequency downshift and decrease in bandwidth. In our earlier study, thickness of the cartilage layer for the same samples before and after trypsin degradation was reported and no significant change in thickness after trypsin degradation was observed (Wang et al. 2010). Thus, the change in thickness of the cartilage layer cannot explain the increased attenuation after trypsin digestion. Increased attenuation is also consistent with earlier studies reporting that the chemically induced degradation of cartilage proteoglycans increases the frequency-dependent attenuation in the cartilage tissue (Joiner et al. 2001; Nieminen et al. 2002). The increase of attenuation may result from the changes in the organization of the collagen fibrils when the proteoglycans are not anymore opposing the tensile forces of the collagen network (Joiner et al. 2001). However, in true OA changes may be different as one study has reported decreased attenuation in spontaneously degenerated tissue (Nieminen et al. 2004), possibly relating to simultaneous increase of water content in clinical OA. Furthermore, clinical OA includes also changes in the subchondral bone that probably affect acoustic properties of the cartilage-bone interface (Buckwalter

120

Ultrasound in Medicine and Biology

and Martin 1995; Day et al. 2004), which might in turn affect the center frequency downshift and the bandwidth decrease. Current results from the collagenase digestion experiments suggest that the attenuation of high-frequency ultrasound in the cartilage layer does not change significantly after collagenase degradation. This is also consistent with the earlier finding that enzymatic degradation of the superficial cartilage collagens has no effect on the ultrasound attenuation (Nieminen et al. 2002). In general, the degeneration models by trypsin or collagenase are highly specific, compared with the actual changes existing in natural OA and, therefore, the current results cannot be generalized to spontaneously degenerated cartilage. To clarify the diagnostic potential of the frequency domain characterization of the ultrasound reflection from the cartilage-bone interface, further studies should be conducted with the spontaneously degenerated, preferably human cartilage tissue. Comparison of the present results of series II with the values of ultrasound parameters reported earlier (IBS and AIBbone) revealed that as the amplitude of the backscattered echo from the cartilage-bone interface is large the shift in center frequency or bandwidth is simultaneously small, corresponding to a lower attenuation in the cartilage tissue. This also confirms the hypothesis that attenuation in the cartilage layer is the dominant factor causing the center frequency downshift. However, it should be noted that since AIBbone is calculated from the same ultrasound echo signal from the cartilage-bone interface as the relative change in the center frequency or bandwidth, it is expectable that they might have a priori interrelationships at least to some extent. With regard to IBS, the relations were more conflicting. When the backscatter in the middle part of the cartilage increases, the shift in center frequency or bandwidth simultaneously decreases, corresponding to the lower frequency-dependent attenuation in the cartilage tissue (Fig. 5a and b). One possible explanation for this surprising relation may be the significant attenuating effect of the superficial layer. When the backscattering increases in the middle layer, which was the case after enzymatic degradations (Wang et al. 2010), simultaneously the attenuation in the superficial layer (including the proximal surface reflection) may decrease more, effectively lowering also the frequency-dependent attenuation in the whole cartilage tissue. As mentioned earlier, these relations should be further clarified using separated cartilage layers at different depths. In the present study, we also analyzed the change in the center frequency and bandwidth for the reflection at the proximal cartilage surface (data not reported). However, no significant changes in the bandwidth and center frequency occurred in ultrasound reflection from the proximal cartilage surface after collagenase and

Volume 37, Number 1, 2011

trypsin degradation. In contrast, a significant decrease in ultrasound reflection amplitude from the proximal cartilage surface has been reported after collagenase degradation for the same samples (Wang et al. 2010). Thus, it seems that after enzymatic degradations ultrasound reflection amplitude from the proximal surface decreases but no significant changes occur in the frequency content. SUMMARY In this study, the downshift of the center frequency and the reduction in the bandwidth of broadband highfrequency (40 MHz) ultrasound signal from the cartilage-bone interface were investigated. In addition, the diagnostic potential of these frequency domain parameters was clarified by analyzing the ultrasound reflection in the frequency domain before and after enzymatic degradation of cartilage tissue proteoglycans and collagen network. The original hypothesis, suggesting that the center frequency downshift is caused by both the variable acoustic properties of the cartilage-bone interface and by the attenuation of the overlying cartilage layer, was partly confirmed in the study. However, the results suggested that attenuation in the cartilage layer is the dominant factor causing the large downshift of the center frequency and the acoustic properties of the subchondral bone play only a minor role. The results also indicated that the measurement of ultrasound bandwidth includes a larger variability than that of the center frequency and the change in the bandwidth cannot be regarded as a sensitive indicator of attenuation properties as the change in the center frequency. Results from current enzymatic digestion tests suggest that the attenuation of the cartilage layer do not change after the chemical degradation of collagen fibrils while a significant increase in the frequency-dependent attenuation might exist after the chemical degradation of proteoglycans with trypsin. This could be a significant finding as, traditionally, ultrasound measurements have been quantitatively analyzed only from the proximal cartilage surface and it has been reported that degradation of cartilage tissue proteoglycans does not change the ultrasound reflection amplitude from the proximal cartilage surface (Pellaumail et al. 2002; Saarakkala et al. 2004), while the collagen degradation does (Saarakkala et al. 2004). This suggests that the proteoglycan loss, typical to OA, may be possible to detect via the downshift in the center frequency. However, to confirm the diagnostic potential of center frequency downshift further studies should be conducted with spontaneous OA tissue. Acknowledgments—The financial support from Hong Kong Research Grant Council (PolyU5318/05E, PolyU5245/03E); The Hong Kong Polytechnic University (J-BB69); Academy of Finland (project 127198);

Ultrasound spectral analysis of cartilage-bone interface d S. SAARAKKALA et al. and Ministry of Education, Finland (project 5741, University of Eastern Finland grant, University of Kuopio, Finland) is acknowledged.

REFERENCES Agemura DH, O’Brien WD Jr, Olerud JE, Chun LE, Eyre DE. Ultrasonic propagation properties of articular cartilage at 100 MHz. J Acoust Soc Am 1990;87:1786–1791. Brown CP, Hughes SW, Crawford RW, Oloyede A. Ultrasound assessment of articular cartilage: Analysis of the frequency profile of reflected signals from naturally and artificially degraded samples. Connect Tissue Res 2007;48:277–285. Buckwalter JA, Martin J. Degenerative joint disease. Clin Symp 1995; 47:1–32. Cherin E, Saied A, Laugier P, Netter P, Berger G. Evaluation of acoustical parameter sensitivity to age-related and osteoarthritic changes in articular cartilage using 50-MHz ultrasound. Ultrasound Med Biol 1998;24:341–354. Day JS, Van Der Linden JC, Bank RA, Ding M, Hvid I, Sumner DR, Weinans H. Adaptation of subchondral bone in osteoarthritis. Biorheology 2004;41:359–368. Disler DG, Raymond E, May DA, Wayne JS, McCauley TR. Articular cartilage defects: In vitro evaluation of accuracy and interobserver reliability for detection and grading with US. Radiology 2000;215: 846–851. Harris ED Jr, Parker HG, Radin EL, Krane SM. Effects of proteolytic enzymes on structural and mechanical properties of cartilage. Arthritis Rheum 1972;15:497–503. Hattori K, Takakura Y, Ishimura M, Habata T, Uematsu K, Ikeuch K. Quantitative arthroscopic ultrasound evaluation of living human cartilage. Clin Biomech 2004;19:213–216. Jaffre B, Watrin A, Loeuille D, Gillet P, Netter P, Laugier P, Saied A. Effects of anti-inflammatory drugs on arthritic cartilage: A highfrequency quantitative ultrasound study in rats. Arthritis Rheum 2003;48:1594–1601. Joiner GA, Bogoch ER, Pritzker KP, Buschmann MD, Chevrier A, Foster FS. High frequency acoustic parameters of human and bovine articular cartilage following experimentally-induced matrix degradation. Ultrason Imaging 2001;23:106–116. Laasanen MS, Saarakkala S, Toyras J, Rieppo J, Jurvelin JS. Sitespecific ultrasound reflection properties and superficial collagen content of bovine knee articular cartilage. Phys Med Biol 2005;50: 3221–3233. Laasanen MS, Toyras J, Hirvonen J, Saarakkala S, Korhonen RK, Nieminen MT, Kiviranta I, Jurvelin JS. Novel mechano-acoustic technique and instrument for diagnosis of cartilage degeneration. Physiol Meas 2002;23:491–503. Laasanen MS, Toyras J, Vasara A, Saarakkala S, Hyttinen MM, Kiviranta I, Jurvelin JS. Quantitative ultrasound imaging of spontaneous repair of porcine cartilage. Osteoarthritis Cartilage 2006;14: 258–263. Laasanen MS, Toyras J, Vasara AI, Hyttinen MM, Saarakkala S, Hirvonen J, Jurvelin JS, Kiviranta I. Mechano-acoustic diagnosis of cartilage degeneration and repair. J Bone Joint Surg Am 2003; 85(A Suppl 2):78–84.

121

Minns RJ, Steven FS, Hardinge K. Osteoarthrotic articular cartilage lesions of the femoral head observed in the scanning electron microscope. J Pathol 1977;122:63–70. Narayana PA, Ophir J. A closed form method for the measurement of attenuation in nonlinearly dispersive media. Ultrason Imaging 1983;5:17–21. Narayana PA, Ophir J, Maklad NF. The attenuation of ultrasound in biological fluids. J Acoust Soc Am 1984;76:1–4. Nieminen HJ, Saarakkala S, Laasanen MS, Hirvonen J, Jurvelin JS, Toyras J. Ultrasound attenuation in normal and spontaneously degenerated articular cartilage. Ultrasound Med Biol 2004;30: 493–500. Nieminen HJ, Toyras J, Rieppo J, Nieminen MT, Hirvonen J, Korhonen R, Jurvelin JS. Real-time ultrasound analysis of articular cartilage degradation in vitro. Ultrasound Med Biol 2002;28:519–525. Ophir J, Jaeger P. Spectral shifts of ultrasonic propagation through media with nonlinear dispersive attenuation. Ultrason Imaging 1982;4:282–289. Pellaumail B, Watrin A, Loeuille D, Netter P, Berger G, Laugier P, Saied A. Effect of articular cartilage proteoglycan depletion on high frequency ultrasound backscatter. Osteoarthritis Cartilage 2002;10:535–541. Saarakkala S, Laasanen MS, Jurvelin JS, Toyras J. Quantitative ultrasound imaging detects degenerative changes in articular cartilage surface and subchondral bone. Phys Med Biol 2006;51: 5333–5346. Saarakkala S, Toyras J, Hirvonen J, Laasanen MS, Lappalainen R, Jurvelin JS. Ultrasonic quantitation of superficial degradation of articular cartilage. Ultrasound Med Biol 2004;30:783–792. Saied A, Cherin E, Gaucher H, Laugier P, Gillet P, Floquet J, Netter P, Berger G. Assessment of articular cartilage and subchondral bone: Subtle and progressive changes in experimental osteoarthritis using 50 MHz echography in vitro. J Bone Miner Res 1997;12: 1378–1386. Senzig DA, Forster FK, Olerud JE. Ultrasonic attenuation in articular cartilage. J Acoust Soc Am 1992;92:676–681. Shingleton WD, Hodges DJ, Brick P, Cawston TE. Collagenase: A key enzyme in collagen turnover. Biochem Cell Biol 1996;74: 759–775. Spriet MP, Girard CA, Foster SF, Harasiewicz K, Holdsworth DW, Laverty S. Validation of a 40 MHz B-scan ultrasound biomicroscope for the evaluation of osteoarthritis lesions in an animal model. Osteoarthritis Cartilage 2005;13:171–179. Treece G, Prager R, Gee A. Ultrasound attenuation measurement in the presence of scatterer variation for reduction of shadowing and enhancement. IEEE Trans Ultrason Ferroelectr Freq Control 2005; 52:2346–2360. Wang Q, Zheng YP, Qin L, Huang QH, Lam WL, Leung G, Guo X, Lu HB. Real-time ultrasonic assessment of progressive proteoglycan depletion in articular cartilage. Ultrasound Med Biol 2008;34: 1085–1092. Wang SZ, Huang YP, Saarakkala S, Zheng YP. Quantitative assessment of articular cartilage with morphologic, acoustic and mechanical properties obtained using high frequency ultrasound. Ultrasound Med Biol 2010;36:512–527.