Comparison of Digital Radiography, Computed Tomography, and Magnetic Resonance Imaging Features in Canine Spontaneous Degenerative Stifle Joint Osteoarthritis

Simple Summary Canine osteoarthritis (OA) is a common disease in aging dogs and mostly affects the synovial joint. Non-invasive imaging modalities, such as digital radiography (DR) and computed tomography (CT), are often used to diagnose and evaluate the severity of OA. Magnetic resonance imaging (MRI) is the preferred diagnostic method for the assessment of OA, is considered the gold standard in humans, and has been applied in several canine experimental studies. However, the value of MRI in diagnosing spontaneous canine OA and the comparison of different imaging modalities have seldom been addressed. This study compared multiple noninvasive imaging modalities in canine spontaneous stifle OA cases. The results showed that MRI provides the most comprehensive and superior lesion detection sensitivity for ligament, meniscus, cartilage, and synovial effusions. DR provides adequate bony structure information, while CT provides the most delicate images of bony structure lesions. These imaging findings may provide further understanding of the disease and help clinicians draft a more precise treatment plan. Abstract Canine stifle joint osteoarthritis (OA) is characterized by damage and degeneration of the articular cartilage and subchondral bone, bony hypertrophy at the margins, and synovial joint membrane changes. Non-invasive imaging modalities, such as digital radiography (DR), computed tomography (CT), and magnetic resonance imaging (MRI), can be used to describe these changes. However, the value of MRI in diagnosing spontaneous canine OA and the comparison of different imaging modalities have seldom been addressed. This study compared multiple noninvasive imaging modalities in canine spontaneous stifle OA cases. Four client-owned dogs with five spontaneously affected OA stifle joints were recruited and underwent DR, CT, and MRI. Information on osteophytes/enthesophytes, ligament/tendon lesions, synovial effusion and membrane thickening, subchondral bone lesions, and meniscal and cartilage lesions were scored and compared. The results showed that MRI provides the most comprehensive and superior lesion detection sensitivity for ligament, meniscus, cartilage, and synovial effusions. DR provides adequate bony structure information, while CT provides the most delicate images of bony structure lesions. These imaging findings may provide further understanding of the disease and help clinicians draft a more precise treatment plan.


Introduction
Canine osteoarthritis (OA) is a common clinical disorder that affects approximately 20% of skeletally mature dogs [1]. OA is characterized by damage and degeneration of articular cartilage and subchondral bone, bony hypertrophy at the margins, and synovial joint membrane changes, which lead to structural and functional decline of the joint and result in pain and lameness [1].
In clinical practice, noninvasive physical examination and imaging modalities, such as digital radiography (DR), computed tomography (CT), and magnetic resonance imaging (MRI), are commonly used to diagnose and evaluate the severity of OA. DR has the advantages of less space requirements, not necessitating specifically trained staff, less requirements of anesthesia, and lower cost, and hence is mostly applied in diagnosing OA. Common radiographic findings of OA include narrowing of the joint space, subchondral sclerosis, and osteophytosis [2]. However, information on ligament injuries and cartilage defects is limited. CT allows the evaluation of periarticular and articular tissues without superimposition on joint radiography [3]. These images can be constructed with the desired window levels or widths for specific joint anatomic components [2,4]. The disadvantage of CT is that the similar attenuation of complex intra-articular structures of the synovial joint may blur the details. MRI has recently been defined as the "gold standard" in human medicine for synovial joint examination because of the superior soft tissue contrast of the periarticular muscles, tendons, ligaments, and articular cartilage [2]. Signal changes within the bone marrow adjacent to osteoarthritic joints are common findings noted on MRI in humans and the subchondral lesions noted on MRI are considered crucial for determining treatment options for OA patients [5,6]. Nevertheless, the use of MRI in canine OA has not been fully explored. In addition, the settings of MRI for OA are challenging in dogs because the optimal sequences are often geared towards human use, and commercial customized coils for dog limbs are limited.
Previous studies have indicated that spontaneously occurring canine OA and human OA share a similar disease progression phenotype [7,8]. However, the MRI features of canine OA were reported in surgically induced experimental OA models rather than in spontaneous cases [9]. Moreover, a comparison of imaging findings from different imaging modalities in the same patients has also not been reported. Therefore, this study aimed to describe and compare the imaging features of DR, CT, and MRI in spontaneous canine stifle OA cases.

Cases Selection
Adult client-owned dogs with spontaneous chronic (>6 months) stifle OA and associated pain were enrolled in this study. Physical examination, synovial joint fluid analysis, and radiographic findings were used to diagnose osteoarthritis (OA). Dogs with neoplastic joint disease, immune-mediated joint disease, or dogs with a high anesthesia risk (>grade 4 of the American Society of Anesthesiologists) were excluded from this study.

Clinical Lameness Scoring System
Clinical impact of OA on the affected limbs was assessed using an objective clinical scoring system (CLSS, scores: 3-15) ( Table 1), which was adopted from a previous study [10]. The CLSS assessed lameness, pain on palpation, and weight-bearing status. Table 1. Clinical lameness scoring system. Adapted with the permission from Ref. [10]. 2013, the Korean Society of Veterinary Science.

Criterion Grade Clinical Evaluation
Lameness 1 Walks normally 2 Slightly lame when walking 3 Moderately lame when walking 4 Severely lame when walking 5 Reluctant to rise and will not walk more than five paces Pain on palpation 1 None 2 Mild signs; dog turns head in recognition 3 Moderate signs; dog pulls limb away 4 Severe signs; dog vocalizes or becomes aggressive 5 Dog will not allow palpation Weight-bearing 1 Equal on all limbs standing and walking 2 Normal standing; favors affect limb when walking 3 Partial weight-bearing standing and walking 4 Partial weight-bearing standing; non-weight-bearing walking 5 Non-weight-bearing standing and walking

Imaging Exam and Analysis
X-ray examinations were performed using an X-ray machine (MRAD-A50S, Toshiba Medical Systems Corporation, Nasu, Japan) with a digital receptor (CXDI-701c wireless digital radiography system, Canon, Japan). The digital radiographic (DR) images of each affected stifle included caudocranial (CdCr) and mediolateral views. The kVp and mA settings were adjusted according to the radiographic technique chart provided by the manufacturer.
Non-contrast CT was performed using a 64-slice positron emission tomography (PET)/computed tomography (PET/CT) (Discovery 64 Slice PET/CT Scanner, GE Healthcare, New York, NY, USA), with bone plus and standard desired reconstruction algorithm, 150 mAs and 120 kVp, and helical acquisitions with a table pitch factor of 0.516. The CT images were reformatted to soft tissue (window width: 400, window level: 40) and bone (window width: 1500, window level: 450). The dogs were placed in dorsal recumbency on V-shaped foam with a bilateral hindlimb stifle joint positioned at a 90-degree angle.
MRI was performed using a 1.5-T MRI system (Signa 1.5T; GE Healthcare, New York, NY, USA) with a flex coil (GE Signa 1.5 T Flex Coil, GE Healthcare, New York, NY, USA). The dogs were in a laterally recumbent position with the target stifle joint on the non-dependent side at an angle of 135 • -145 • . The contralateral hind limb was placed at a different angle from the target stifle joint. To reduce artifacts and noise signals, towels and large MRI foam sponges were used to maintain the position of the contralateral hindlimb while the drip lines and capnography sensor were kept away from the target stifle joint. The sequences performed included T1-weighted (T1W), T2-weighted (T2W), proton-density-weighted (PDW), proton-density-weighted fat saturation (PDW fs), and T2 mapping. Standard planes included the sagittal plane, which was oriented parallel to the medial surface of the medial femoral epicondyle; the transverse plane, which was oriented perpendicular to the patellar ligament and sagittal orientation line; and the dorsal plane, which was parallel to the patellar ligament. The scanning parameters are listed in Table 2.
All imaging procedures were performed by residents of veterinary radiology under general anesthesia. Each dog underwent the DR, CT, and MRI examination on the same day. Digital Imaging and Communications in Medicine (DICOM) format was evaluated using diagnostic imaging software (Osirix, Version 9.5.2, Bernex, Switzerland). The original DR, CT, and MRI images were assessed and analyzed by an experienced veterinarian with radiology expertise (L.-S.L.) and a chief resident of veterinary radiology (Y.-J.T.). The final decision was made by consensus.

Articular Lesions Evaluation
Articular lesions of osteophytes and enthesophytes, ligament lesions, synovial effusions, synovial membrane thickening, subchondral lesions, and meniscal lesions were evaluated by grading or presented as "present" or "absent".

Ligament/Tendon Lesions
Ligament/tendon lesions were assessed in 6 regions: medial collateral ligament (MCL), lateral collateral ligament (LCL), long digital extensor tendon (LDET), patellar ligament (PL), cranial cruciate ligament (CCL), and caudal cruciate ligament (CdCL). Because the ligament/tendon structure and the lesion cannot be well identified using DR and CT, the severity and incidence of ligament/tendon lesions were only recorded on MRI images and scaled from 0 to 3 (0: normal; 1: partial rupture; 2: partial rupture with thickening or increased peripheral effusion; 3: complete rupture), which was modified from a previous study [12].

Synovial Effusion and Membrane Thickening
The presence of synovial effusion in the infrapatellar region (IP) was recorded in the lateral view of the DR or sagittal planes of the CT and MRI images, and the maximum potential distention of the synovial cavity was scaled from 0 to 3 (0: normal; 1: ≤33%; 2: 33-66%; 3: ≥66%) [13]. The presence of synovial effusion extensions at the MF, LF, medial femoral trochlear ridges to the epicondyle region (MFTr), lateral femoral trochlear ridges to the epicondyle region (LFTr), and femoral trochlear groove (FTG) was recorded on MRI only ( Figure 2). In addition, synovial membrane thickening was observed on MRI.

Ligament/Tendon Lesions
Ligament/tendon lesions were assessed in 6 regions: medial collateral ligament (MCL), lateral collateral ligament (LCL), long digital extensor tendon (LDET), patellar ligament (PL), cranial cruciate ligament (CCL), and caudal cruciate ligament (CdCL). Because the ligament/tendon structure and the lesion cannot be well identified using DR and CT, the severity and incidence of ligament/tendon lesions were only recorded on MRI images and scaled from 0 to 3 (0: normal; 1: partial rupture; 2: partial rupture with thickening or increased peripheral effusion; 3: complete rupture), which was modified from a previous study [12].

Synovial Effusion and Membrane Thickening
The presence of synovial effusion in the infrapatellar region (IP) was recorded in the lateral view of the DR or sagittal planes of the CT and MRI images, and the maximum potential distention of the synovial cavity was scaled from 0 to 3 (0: normal; 1: ≤33%; 2: 33%-66%; 3: ≥66%) [13]. The presence of synovial effusion extensions at the MF, LF, medial femoral trochlear ridges to the epicondyle region (MFTr), lateral femoral trochlear ridges to the epicondyle region (LFTr), and femoral trochlear groove (FTG) was recorded on MRI only ( Figure 2). In addition, synovial membrane thickening was observed on MRI.

Subchondral Bone Lesions
Subchondral bone defects, cysts, and bone-marrow-edema-like lesions were defined as subchondral bone lesions and evaluated in five anatomical compartments: femoral trochlear (FT), femoral intercondylar fossa (ICF), femoral lateral condyle (LC), femoral medial condyle (MC), and tibial plateau (TP) (Figure 3). The presence of subchondral bone lesions was evaluated using DR, CT, and MRI. Severity was assessed using a modified

Meniscal and Cartilage Lesions
Because meniscal and cartilage lesions cannot be well identified in DR and CT images, these lesions were only evaluated using MRI. Meniscal lesions were evaluated at the medial and lateral menisci and recognized as linear or oval hyperintensity on T2W and PDW images. The severity of meniscal lesions was scaled from 0 to 3 (0, normal; 1, linear or oval core lesion; 2, linear or oval lesion extending to the articular surface; 3, complete rupture) [14]. Cartilage lesions were recognized as heterogeneous signals of cartilage, irregular cartilage contour, and cartilage defects/rupture. The presence of cartilage lesions was recorded in 13 anatomical compartments and numbered from C1-C13: two compartments of MC (C1, C2) and LC (C3, C4), and nine compartments of the tibial plateau (C5-C13) ( Figure 4). T2 relaxation times for evaluating cartilage lesions at the medial femoral condyle, lateral femoral condyle, medial tibial plateau, and lateral tibial plateau were calculated by manually drawing the region of interest (ROI) on color-coded T2 mapping images with a cartilage thickness > 0.5 mm (>1 pixel). The results were recorded and compared with those of a previous study [15]. C13) (Figure 4). T2 relaxation times for evaluating cartilage lesions at the medial femoral condyle, lateral femoral condyle, medial tibial plateau, and lateral tibial plateau were calculated by manually drawing the region of interest (ROI) on color-coded T2 mapping images with a cartilage thickness > 0.5 mm (>1 pixel). The results were recorded and compared with those of a previous study [15].

Statistical Analysis
Data were statistically analyzed using commercially available software (GraphPad Prism, ver. 7.00, San Diego, CA, USA). Descriptive statistics (mean, median, and percentage) were applied for continuous variables (age, body weight, and T2 relaxation time) and ordinal variables (CLSS and scores for each grading articular lesion). Categorical variables (signalments and affect joints) are presented in tabular form. Comparisons of scales

Statistical Analysis
Data were statistically analyzed using commercially available software (GraphPad Prism, ver. 7.00, San Diego, CA, USA). Descriptive statistics (mean, median, and percentage) were applied for continuous variables (age, body weight, and T2 relaxation time) and ordinal variables (CLSS and scores for each grading articular lesion). Categorical variables (signalments and affect joints) are presented in tabular form. Comparisons of scales obtained from DR, CT, and MRI images were performed using the Kruskal-Wallis test followed by Dunn's multiple comparisons test.

Animals
Four dogs (three Beagles and one Labrador retriever) with five stifle joints (three left and two right) met the inclusion criteria and were recruited. Three dogs were spayed females and one dog was an intact male. The median age of the dogs was 15 years (range, 10-17 years). The median weight was 11.7 kg (range, 7.8-28.0 kg). Hindlimb lameness was noted in all dogs, and the median CLSS score was 9.0 (range, 8-10).

Osteophytes/Enthesophytes
Representative images of the severity of osteophytes/enthesophytes in the stifle joint are shown in Figure 5. The incidences and scores of osteophytes and enthesophytes in each anatomical compartment are listed in Table 3. In the ME, LE, MC, LC, MT, and LT subregions, osteophytes/enthesophytes were identified in all joints by all modalities. In the FT subregion, osteophytes and enthesophytes could not be identified in only one joint on DR. The tibial tuberosity subregion was the most uncommon site for the osteophytes/enthesophytes, and they were only found in one joint (20%) by different modalities. In the medial and lateral fabella subregions, CT (100%) had a higher sensitivity than DR (60%) and MRI (60%) (Figure 6A-C). In the distal patella and proximal patella subregions, MRI (100%) had higher sensitivity than DR (60%) and CT (40%) (DP, PP) ( Figure 6D-F). No significant differences in osteophytes/enthesophytes lesion scores at different anatomic subregions among DR, CT, and MRI were noted. on DR. The tibial tuberosity subregion was the most uncommon site for the osteophytes/enthesophytes, and they were only found in one joint (20%) by different modalities. In the medial and lateral fabella subregions, CT (100%) had a higher sensitivity than DR (60%) and MRI (60%) (Figure 6A-C). In the distal patella and proximal patella subregions, MRI (100%) had higher sensitivity than DR (60%) and CT (40%) (DP, PP) ( Figure 6D-F). No significant differences in osteophytes/enthesophytes lesion scores at different anatomic subregions among DR, CT, and MRI were noted.

Subchondral Bone Lesions
Subchondral bone lesions were noted in all modalities (Figure 9). The incidence and scores of subchondral bone lesions in the different compartments on DR, CT, and MRI are listed in Table 4. ICF was the most common site of subchondral bone lesions, which could be noted in most joints, except for one joint on the DR images. In contrast, lesions were less commonly seen in FT, and only two joints were identified on DR images. Although CT and MRI provided similar sensitivities in detecting lesions, CT exhibited higher sensitivity in TP. In addition, no significant differences in subchondral bone lesion scores in

Subchondral Bone Lesions
Subchondral bone lesions were noted in all modalities (Figure 9). The incidence and scores of subchondral bone lesions in the different compartments on DR, CT, and MRI are listed in Table 4. ICF was the most common site of subchondral bone lesions, which could be noted in most joints, except for one joint on the DR images. In contrast, lesions were less commonly seen in FT, and only two joints were identified on DR images. Although CT and MRI provided similar sensitivities in detecting lesions, CT exhibited higher sensitivity in TP. In addition, no significant differences in subchondral bone lesion scores in each compartment were observed among the different modalities.

Meniscal and Cartilage Lesions
Medial meniscal lesions were noted in two joints (2/5 [40%]), with scores of 1 and 3, respectively. Except for linear or oval core lesions, a meniscal maceration lesion, which is a specific term used in MRI to describe chronic degenerative meniscal lesions causing meniscal wasting and becoming separated fragments, was noted in one joint ( Figure 10). A lateral meniscal lesion was only seen in one joint, which was encountered together with the medial meniscal lesion, with a score of 2. Cartilage lesions were noted in almost all subregions of all joints except for two subregions (C5 and C8) in one stifle joint. The T2 relaxation time of the cartilage lesions was generally lower than that of previously published data (Table 5). Table 4. Comparison of the subchondral bone lesions on DR, CT, and MRI images at different anatomic subregions.

Compartments
Modality Incidence Scores p-Value

Meniscal and Cartilage Lesions
Medial meniscal lesions were noted in two joints (2/5 [40%]), with scores of 1 and 3, respectively. Except for linear or oval core lesions, a meniscal maceration lesion, which is a specific term used in MRI to describe chronic degenerative meniscal lesions causing meniscal wasting and becoming separated fragments, was noted in one joint ( Figure 10). A lateral meniscal lesion was only seen in one joint, which was encountered together with the medial meniscal lesion, with a score of 2. Cartilage lesions were noted in almost all subregions of all joints except for two subregions (C5 and C8) in one stifle joint. The T2 relaxation time of the cartilage lesions was generally lower than that of previously published data (Table 5).

Discussion
This study recruited canine patients with naturally occurring chronic stifle OA with moderate lameness and evaluated DR, CT, and MRI imaging findings. Although similar clinical performances were noted in these patients, the individual associated soft and bony structural changes may differ, and the results could be altered using different modalities. Currently, various treatment options are available for canine OA, including surgery, rehabilitation, regeneration therapy, and nutritional supplementation. Detailed knowledge from these imaging modalities is essential for developing customized treatment for OA.
Our results indicated that osteophytes/enthesophytes could form in almost all anatomical compartments of the stifle joint in chronic OA, except for the tibial tuberosity. This is considered an acceptable result, since the tibial tuberosity is an extra-articular structure and may not be involved in the OA process. In addition, the osteophytes/enthesophytes at the femoral epicondyle (ME, LE), femoral condyle (MC, LC), and fabella (MT, LT) were the most evident and could be identified by all modalities. CT and MRI may be more sensitive than DR at femoral trochlear ridges (FT) because the superimposition of osteophytes, enthesophytes, and distal femur can be avoided.
In the fabella subregions (MF and LF), it may be difficult to identify osteophytes and enthesophytes by using DR and MRI. Severe osteophytes in the distal femur may superimpose on the fabella in the CdCr view, and bilateral fabellae may superimpose on each other in the lateral view. Moreover, multiple connective tissues may superimpose on each other at this site, such as the femoropatellar ligament and the origin of the gastrocnemius on DR images. These connective tissues also present hypointense signals on MRI; thus, they can sometimes be confounded by osteophytes and enthesophytes. In contrast, CT has the following advantages: multiple plane images; imaging processing by three-dimensional CT rendering and reconstruction; and the superior ability in differentiating the soft and bony tissues, allowing better identification of the osteophytes and enthesophytes better than the other modalities at the fabella site.
In the patellar region (DP, PP), MRI provided the highest sensitivity in detecting osteophytes and enthesophytes. The texture of osteophytes and enthesophytes at an early stage is similar to cartilage and appears radiolucent on DR or CT images, whereas the osteophytes and enthesophytes around the patella in the early stage may present hyperintense signals compared with cortical bone on MRI owing to incomplete mineralization [16,17]. This enables early detection of osteophytes and enthesophytes on MRI. In addition, the superimposition of incompletely mineralized osteophytes on DR images may provide better identification than CT.
It is challenging to evaluate ligament/tendon lesions using non-invasive imaging examinations, especially DR and CT. Nevertheless, MRI is useful for diagnosing ligamental/tendon lesions. In this study, various severities of ligament/tendon lesions were identified in the six ligaments and tendons around the stifle joints. CCL, CdCL, and LCL lesions were the major ligament lesions observed in all joints. CCL disease is the most common cause of hindlimb lameness in dogs, and progressive stifle OA will develop [18,19]. Although the exact pathogenesis of canine CCL disease remains controversial, aging changes resulting in degeneration of the CCL may be associated with programmed cell death, inflammatory cell infiltration, and extracellular matrix degeneration, and may contribute to the event [19][20][21][22]. Therefore, it is not surprising that CCL lesions were notable. In addition, the CdCL, which is close to the CCL in the same space, may share the same process of degeneration. This might explain the same incidence and severity of CCL and CdCL lesions here, even though injury of the CdCL was less commonly noted in dogs encountering traumatic processes. Interestingly, the LCL also revealed notable changes simultaneously with CCL and CdCL. Since the LCL can be injured when the tibia is pushed outward or excessively internally rotated, lesions of the LCL may result from excessive tension after CCL injury [23,24]. In contrast, the LDET and PL were the most common sites of ligament lesions. Unlike ligaments, tendons are responsible for the extension or flexion of joints, rather than the maintenance of joint stability. As a result, the lesions in the LDEL and PL were not notable and may not be related to the OA process.
Synovial effusion at the IP subregion, which was strongly correlated with chronic OA, was found in all modalities. Consistent results have also been reported in experimental canine OA models and working dogs [11,25]. Although all three modalities provided a high detection rate, CT and MRI were more sensitive than DR for detecting synovial effusion in the IP region. The extension of synovial fluid was mostly noted around the fabella in our cases, similar to the results reported in dogs undergoing experimental surgery after 26 weeks [11]. This may indicate that the fabella region, including the femorofabellar joint and joint capsule, is also important in OA processing. In contrast, synovial extension is least seen in the trochlear region (FTG), and it is thought that the presence of the patellar bone in the trochlear groove may limit the extension. In addition, thickening of the synovial joint capsule in the IP region was also identified in all joints on MRI. These results indicated that MRI exhibited high sensitivity for fluid signals and synovial membrane structures in stifle joints, which were not well identified on DR or CT.
The subchondral lesion or bone marrow lesion may indicate inflammation, edema, and ischemic condition of the bone marrow [26][27][28], which presents as T2W and PDW hyperintense lesions on MRI. Nevertheless, DR and CT provide details of bony structures, such as subchondral bone defects or cysts. Our results revealed that subchondral lesions at the FT were only noted on DR images. Combining the CT and MRI findings of the same joint, these lesions resulted from the superimposition of regional osteophytes with trochlear ridges rather than "true" subchondral lesions ( Figure 9). Therefore, the findings of DR imaging, used to detect subchondral lesions in the FT subregion, should be interpreted with caution. Although subchondral bone plays a crucial role in the initiation and progression of human OA [29], the clinical values still can not be identified here due to the chronic OA stage and single time point for assessment. However, a previous study demonstrated that the abnormal bone metabolism was detected before subchondral bone lesion formatted in human OA, and the sodium fluoride positron emission tomography/CT coregistered with MRI could serve as a feasible molecular imaging modality to assess early post-traumatic OA [30].
Meniscal lesions were observed less commonly in this study. Previous studies revealed that the degenerative change and cranial cruciate ligament rupture could be associated with meniscal lesions [31,32]. However, the underlying causes could not be identified due to limited information on illness history.
Cartilage lesions are common and extensive in canine stifle OA. Although a previous study showed that MRI could evaluate the morphology of articular cartilage [33], only the integrity of the cartilage could be documented in our study. Although a higher magnetic field MRI may be required to evaluate the details of stifle joint cartilage in small-to-middlebreed dogs, severe cartilage defects could still be detected using 1.5T MRI.
A recent study applied T2 relaxation time to determine changes in cartilage composition in normal beagles [15]. T2 relaxation time reflects the water composition change in articular cartilage and provides early detection of articular cartilage injury and subsequent mineralization. The increase in T2 relaxation time could be used to detect cartilage damage in dogs, whereas the decrease may be related to severe OA or articular cartilage mineralization in humans [34,35]. The T2 relaxation time in our study was generally lower than previous results in normal beagle stifle joints [15]. The profound cartilage lesions with severe OA observed in our cases may explain these changes.
Our study had several limitations. The primary limitation is the lack of comparison between imaging findings with arthroscopy and histological examination because all cases were client-owned clinical cases. Another limitation of our study is its cross-sectional design with a small sample size of clinical OA cases, making it difficult to investigate the etiology of OA lesions. Further longitudinal studies including more dogs and patients with different severities of OA are warranted.

Conclusions
In conclusion, comparing the imaging modalities for canine stifle OA, DR provides adequate bony structure information, the shortest examination time, and the least budget consumption; however, most soft tissue and cartilage lesions could not be seen. In addition, the superimposition of abundant osteophytes and enthesophytes may have misled the results. CT provides the most delicate bony structure lesion information with a moderate examination time. However, articular soft tissue lesions cannot be obtained without invasive techniques (e.g., contrast arthrography), and a higher cost than DR is required. MRI provides images of the superior ligament, meniscus, cartilage, and synovial effusion with adequate bony structure information. Although the disadvantages of long anesthesia time during the examination and high cost are still of concern, MRI provides the most comprehensive joint images compared to other non-invasive imaging modalities used in this study, and these findings provide detailed information on stifle OA.

Institutional Review Board Statement:
The study protocol was approved by the Institutional Animal Care and Use Committee of the National Pingtung University of Science and Technology (NPUST-108-059, 2019/10).

Informed Consent Statement:
Informed consent was obtained from the owners of all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available upon request from the corresponding author.