Animal models
Eighteen normal adult New Zealand white female rabbits weighing 3.5 to 4.5 kg (mean, 4.1 ± 0.3 kg) were used in this study. Radiographs of both femorotibial joints were taken and evaluated by two orthopedists to exclude animals with joint pathology. Six rabbits were treated as control. To induce OA, surgical transection of the anterior cruciate ligament (ACL) in the left femorotibial joint was performed under general anesthesia in twelve rabbits. Routine skin incision closure was performed. Antibiotics (penicillin 20,000 IU) were injected intramuscularly twice a day preoperatively and for 2 days postoperatively in the operated animals. Following surgery, free movement was allowed in separate cages for the duration of the experimental period. Experiments on the rabbits were approved by our institutional animal care and use committee and performed under the guidelines of the National Institutes of Health for the care of laboratory animals.
At 2, 4, and 6 weeks after surgery, six animals were euthanized, including two control rabbits and four rabbits with ACL transection. Each of the left knees was dissected and sectioned with a band saw to obtain the articular cartilage samples with pathological characteristics. 72 cartilage specimens were collected, including 18 medial femoralcondyles, 18 lateral femoral condyles, 18 medial tibial plateaus, and 18 lateral tibial plateaus. All the specimens were wrapped in wet gauze soaked with saline buffer and stored at -20°C until the ultrasound examination.
Ultrasound examination
UBM imaging was performed on these cartilage surfaces using a Vevo770 (VisualSonics, Toronto, ON, Canada) high-frequency UBM system. A transducer (RMV708) with a nominal central frequency of 55 MHz was used. This transducer has a focal length of 4.5 mm and an axial resolution of 30 μm.
Before ultrasound scan, each specimen was first assessed macroscopically by two radiologists [15]. Then the area on the cartilage surface with the most severe macroscopic lesion was selected to be scanned by ultrasound. The specimen was fixed using plasticine to ensure a horizontal cartilage surface of the ultrasound-scanned site (Figure 1). Gel was then put on the surface of the cartilage. A clamp that could be adjusted in the vertical and lateral directions was used to fix the transducer vertically to the cartilage surface. By adjusting the clamp, the transducer surface was immerged into gel and was placed approximately 4.5 mm above the cartilage surface with the focal zone of ultrasound beam located inside the cartilage layer to obtain the maximum echo amplitude. Two-dimensional UBM images were acquired with a 4 × 4 mm2 field of view at 46 frames/s. The data of the ultrasound radiofrequency (RF) signals were recorded and stored for the calculation of the acoustic parameters of the cartilage tissue.
Pathological grade
According to the grading standards of International Osteoarthritis Institute (OARSI), OA is graded as follows: grade 1 = uneven surface that can demonstrate superficial fibrillation; grade 2 = surface discontinuity that may be accompanied by cell proliferation, increased or decreased matrix staining in the mid zone; grade 3 = vertical fissures extending into the mid zone or erosion; grade 4 = denudation (the unmineralized hyaline cartilage is completely eroded); grade 5 = deformation. OA develops with three stages: early stage (grade 1-2), intermediate stage (grade 3) and advanced stage (grade 4-5) [15, 21].
Cartilage samples used for pathological observation were fixed in 10% neutral buffered formalin, decalcified with 14% ethylenediamine tetra-acetic acid, dehydrated through graded alcohols, cleared with toluene, and embedded in paraffin. Careful attention was paid to make the histologically assessed site consistent with the ultrasound-scanned site. Six-micrometer sections of articular cartilage were cut from the proximal part to the ultrasound-scanned site. The sections were stained with toluidine blue. Then the sections were observed and graded by three board-certified pathologists blinded with the purpose of the whole study.
Parameter extraction
In this study, three parameters are extracted to quantitatively describe the acoustic properties of articular cartilage, i.e. ultrasound roughness index (URI) of the cartilage surface, reflection coefficient (R) of the cartilage surface, reflection coefficient (Rbone) of the cartilage-bone interface. URI is used to describe the microtopography of the cartilage surface. R and Rbone are used to describe characteristics of cartilage tissue [11].
As shown in Figure 2, URI is obtained from the echo signals between the probe and the cartilage surface. It is calculated using equation 1.
(1)
where m is the number of the scan lines in the 4 mm sampling length. d i is the distance between the transducer and the solution- cartilage interface in the scan line i, and 〈d〉 is the average distance between the transducer and the surface.
R and Rbone are defined by equation 2 and equation 3, respectively.
(2)
(3)
where A
si
and A
bi
are the peak-to-peak amplitude of the ultrasound RF signals reflected from the cartilage surface and the cartilage-bone interface, respectively, in the scan line i. m is the number of scan lines. A
iref
is the reference peak-to-peak amplitude of the echoes reflected from the solution-air interface with the same distance as A
si
. In the parameter calculation, the ultrasound speed in the solution is 1500 m/s, and the average ultrasound speed in the cartilage tissue equals 1600 m/s [11]. The analysis of ultrasound echo signals and the extraction of the ultrasound parameters were carried out by a custom-designed Matlab program (MATLAB, Version 2009, The Math-Works, USA).
Statistical analysis
Statistical analyses were conducted with SPSS software (Version 17, SPSS Inc., USA). All values in the text are presented as mean ± standard deviation (SD). All specimens were evaluated to the normal, grade 1, grade 2, grade 3, grade 4, grade 5 groups. The analysis of variance in URI, R, Rbone between grading groups were performed using One-way ANOVA and LSD post-hoc tests. P < 0.05 was taken as statistically significant.