Changes in viscoelastic properties of articular cartilage in early stage of osteoarthritis, as determined by optical coherence tomography-based strain rate tomography

Female New Zealand white rabbits aged 8–9 months and weighing 3.6–4.0 kg with closed epiphyses (SLC, Inc. Japan) were used for this study. Closed epiphyses were confirmed via radiography. Experimental OA was induced via anterior cruciate ligament transection (ACLT) in the left knee joint. All rabbits were anesthetized with subcutaneous injection of ketamine (40 mg/kg) and xylazine (5 mg/kg). The medial parapatellar incision was made, and the patella was dislocated laterally. The ACL was visualized and transected with a No. 11 blade. An anterior drawing test was performed to confirm that the ACL was transected completely. The joint was irrigated with sterile saline, and the capsule and skin were closed. Postoperatively, the animals with free activity allowed were housed singly in cages in an animal room with controlled temperature and humidity and periodic light cycles. The animals were sacrificed at 2, 4, and 8 weeks after surgery (n = 5/time point) via intravenous injection of sodium pentobarbital (3 ml). Three animals who did not undergo ACLT were sacrificed for the control group (n = 6). After euthanasia, distal femurs were amputated at the supracondylar site using an electric saw. Distal femurs were then prepared for mechanical tests and histological analysis. All animal experiments were strictly performed in accordance with the regulations of the Institutional Animal Care and Use Committee, Osaka City University Medical School (Approval number: 15012).

The SR-OCSA system is shown in Fig. 1a and b. The SR-OCSA system was constructed with the Swept Source OCT (IVS-2000; Santec Inc., Aichi, Japan) system by taking the video rate of tissue tomogram and an indenter controlled by a piezoelectric (PZT) actuator (nPFocus100; nPoint Inc., Middleton, MI, USA). Mechanical measurement locations were defined as the top point of the femoral condyle. The distal femurs were glued onto a dish using cyanoacrylate adhesive gel and immersed in phosphate-buffered saline. The dish was fixed to the stage using a handmade fixture. The cartilage image was acquired by using OCT, and the cartilage thickness was measured from the acquired OCT image (Fig. 2). The actuator was driven from the state where the indenter was in close contact with the cartilage at the same location as the OCT image experiment was conducted. In the stress relaxation test, the amount of cartilage compression was set to 10% of the thickness of articular cartilage. All samples were tested using a stress-relaxation protocol (10% pre-strain, 10% strain, 2.5% strain per second speed, and 60 s relaxation time). The deformed cartilage was visualized using the OCT probe installed on the upper side of the revolution arm while deforming with the indenter. The system had a depth (z) and width (x) resolution of 7.1 μm and 7.5 μm, respectively. The OCT image size acquired was set as a 2-dimensional xz tomographic image of 1 mm × 1.51 mm (128 × 192 pixel), and images were continuously acquired at 79.05 frames per second (a total of 4743 frames). By obtaining the image acquisition signal of the OCT and the drive signal of the PZT actuator, the compression relaxation behavior and OCT tomographic image can be synchronized. In this approach, OCT images are captured continuously during the stress relaxation test using a PZT actuated indenter, and SR-OCSA can provide the tomography of the deformation vector and strain rate.

OCSA is a new method to detect the mechanical properties at the microscale using OCT, and it is mainly constructed using a digital image correlation method to tomographically detect the displacement vector distribution inside tissues [14]. Specifically, the OCSA is applied to the OCT tomographic images obtained under mechanical load to the tissue before and after deformation. OCSA directly detects the displacement vector distribution of the tissue and was continuously performed in this study on the acquired continuous OCT images to tomographically detect the strain rate distribution inside the tissue at the microscale. The flowchart of the OCSA algorithm is shown in Fig. 3. The tissue is deformed by the mechanical load from the actuator, and the OCT tomographic images before and after deformation are acquired (Fig. 4a). The OCSA algorithm uses the speckle cross-correlation analysis recursively to obtain synthetic OCT images with a quarter reduction of the subset window size. However, OCT images have speckle noise and contain less valid information than other images commonly used in digital image correlation (DIC) analysis. Furthermore, an OCT image typically has the lower number of pixels than a normal DIC image; such low quality and quantity cause erroneous vectors in cross-correlation analysis. As shown in Fig. 3, the cross-correlation method and adjacent cross-correlation multiplication technique are introduced recursively to correct vectors and enhance the vector resolution by suppressing the erroneous vectors and error propagation. Furthermore, the amount of high-precision subpixel deformation is calculated using the sub-pixel analysis method (the upstream gradient method and image deformation method). OCSA can provide data on the distribution of highly accurate speed vectors, taking account of tissue deformation, as an instantaneous tomographic image (Fig. 4b). The time sequential data of the vector field are smoothed by the time-space moving least square method as the approximated speed vectors. The tomographic distribution of the strain rate tensor is calculated from the differential coefficient of the approximate expression (Fig. 4c).

The viscoelastic body shows stress relaxation behavior when the whole body is held and the stress gradually decreases. The micro deformation inside the viscoelastic body is observed with stress relaxation. Therefore, we adopted the SR-OCSA (i.e., continuous OCSA) during the stress relaxation test to detect sequential tomographic changes in the strain rate. Linear approximation of α (x, z, t) t + β (x, z, t) was applied to the detected compressive strain rate in the depth direction (z) to visually evaluate the viscoelastic properties of the tissue as attenuation coefficient of strain rate (ACSR). ACSR is an index of the viscoelastic properties during stress relaxation (Fig. 4d). The viscoelastic properties in the object were determined from the ACSR averaged in the cartilage region determined from the OCT image.

After the SR-OCSA experiment, the validity of the ACSR obtained from SR-OCSA was verified via an indentation test using a universal tabletop mechanical tester (EZ-LX; Shimadzu, Kyoto, Japan) under the same conditions as those in the SR-OCSA mechanical test (10% pre-strain, 10% strain, 2.5% strain/sec speed, and 60 s relaxation time). The compressive forces were measured instantaneously via the load cell, and the force relaxation curves were acquired during the indentation test and were approximated using an exponential decay function. Relaxation time (τ) was defined as the time at which the initial stress becomes 1/e in stress relaxation and represents the ratio of the viscosity and elasticity of the viscoelastic body, i.e., the balance of elasticity and viscosity.

The samples were fixed with 4% paraformaldehyde and decalcified by soaking in Morse’s solution for 2 weeks. The samples were then dehydrated and embedded in paraffin and were cut in 4-μm microsections in the sagittal plane of the medial femoral condyle. The sections were deparaffinized using xylene and ethanol and stained with hematoxylin/eosin and safranin O-fast green to evaluate OA changes. Histological changes of the articular cartilage in the medial femoral condyle were assessed using the Osteoarthritis Research Society International (OARSI) histochemical/histological scoring system [15]. This grading system assigns scores ranging from 0 to 24 based on the safranin O staining: staining (0–6), structure (0–11), chondrocyte density (0–4), and cluster formation (0–3).

One-way analysis of variance test with Tukey’s post hoc analysis to compare the averages of continuous variables among the four groups was used. The correlations between τ and ACSR, OARSI score and τ, and OARSI score and ACSR were tested using the Spearman correlation analysis test. Statistical analyses were performed with GraphPad Prism v.7.0. (GraphPad Software, La Jolla, Ca). P < 0.05 was considered significant. Post hoc power analysis verified that the experimental protocol could distinguish between groups with a statistical power of > 0.80. The post hoc power analysis was performed using G*power software (Faul, Erdfelder, Lang & Buchner, 2007).

At 2 weeks after ACLT, there were slight OA signs, intact surface, and reduction in safranin O staining only in the surface layer (Fig. 5a). At 4 weeks after ACLT, there were apparent OA signs, including superficial diffuse loss of safranin O and fibrillation. At 8 weeks after ACLT, there was diffuse loss of safranin O in the whole layer and fibrillation extending to the middle zone. The OARSI histological scores were 3.4 ± 0.5 at 2 weeks after ACLT (n = 5, mean ± standard deviation), 4.8 ± 0.4 at 4 weeks after ACLT (n = 5), and 9.2 ± 2.8 at 8 weeks after ACLT (n = 5; Fig. 5b). There were no histological OA signs in the control group. The OARSI histological score in the control group was 1.0 ± 0 (n = 6). The OARSI histological scores at 2 weeks after ACLT were higher than in the control group, but the difference was not significant (control vs 2 weeks: p = 0.06; Fig. 5b). The OARSI score significantly increased in a time-dependent manner, beginning at 4 weeks after ACLT (control vs 4 weeks: p < 0.01, control vs 8 weeks: p < 0.01, 2 weeks vs 4 weeks: p = 0.62, 2 weeks vs 8 weeks: p < 0.01, and 4 weeks vs 8 weeks: p < 0.01; Fig. 5b).

The mechanical properties of the cartilage measured via the universal tabletop mechanical tester are summarized in Fig. 6a. The relaxation time (τ) in the control group showed the highest value (12.7 ± 2.3 s, n = 6) (mean ± S.D.). At 2 weeks after ACLT, τ significantly decreased, although the histological change was not prominent (2 weeks: 8.7 ± 1.4 s, n = 5), and τ in the ACLT groups tended to decrease in a time-dependent manner as histological changes became more prominent (4 weeks: 9.2 ± 2.0 s, n = 5; 8 weeks: 7.0 ± 0.7 s, n = 5) (control vs 2 weeks: p = 0.02, control vs 4 weeks: p = 0.04, control vs 8 weeks: p < 0.01). Furthermore, τ was negatively correlated with the OARSI score (r = − 0.65, p < 0.01; Fig. 6b).

Similar to the relaxation time, at 2 weeks after ACLT, the averaged ACSR in the cartilage significantly increased much earlier than when the histological change became prominent (control vs 2 weeks: p = 0.03; Fig. 6c). ACSR in the ACLT groups tended to increase in a time-dependent manner as histological changes became more prominent (control vs 4 weeks: p = 0.04; control vs 8 weeks: p < 0.01). Furthermore, ACSR had a strong positive correlation with the OARSI score (r = 0.80, p < 0.01; Fig. 6d) and a strong negative correlation with τ (r = − 0.70, p < 0.01; Fig. 6e).

A representative OCT image and the distribution of ACSR in each group are shown in Fig. 7a and Fig. 7b. The elevation of ACSR is expressed via becoming darker in the color tone. Each group showed mainly elevated ACSR around the tidemark, rather than in the superficial layer. Compared with the control group, the group at the 2 weeks after ACLT showed an expansion of the range where ACSR in the deep layer increase. The group at 4 and 8 weeks after ACLT showed further increase in ACSR in the deep layer. The degree of elevation in ACSR showed a tendency to be larger in a time-dependent manner.