Diagnostic performance of in vivo 3-T MRI for articular cartilage abnormalities in human osteoarthritic knees using histology as standard of reference

Abstract

The purpose of this study was (1) to evaluate the sensitivity, specificity and accuracy of sagittal in vivo 3-T intermediate-weighted fast spin-echo (iwFSE) sequences in the assessment of knee cartilage pathologies using histology as the reference standard in patients undergoing total knee replacement, and (2) to correlate MR imaging findings typically associated with osteoarthritis such as bone marrow edema pattern (BMEP) and cartilage swelling with histological findings. Tibial plateaus and femoral condyles of eight knees of seven patients were resected during surgery, and sagittal histological sections were prepared for histology. Preoperative MRI findings were compared to the corresponding region in histological sections for thickness, surface integrity and signal pattern of cartilage, and histological findings in areas of BMEP and swelling were documented. The overall sensitivity, specificity and accuracy were 72%, 69% and 70% for thickness, 69%, 74% and 73% for surface and 36%, 62% and 45% for intracartilaginous signal pattern. For all cases of BMEP on MRI subchondral ingrowth of fibrovascular tissue and increased bone remodeling were observed. MRI using fat-saturated iwFSE sequences showed good performance in assessing cartilage thickness and surface lesions, while signal changes of cartilage were not suited to characterize the severity of cartilage degeneration as validated by histology.

Introduction

Osteoarthritis (OA), characterized by progressive loss of hyaline articular cartilage, is an extraordinarily common disease and the primary cause for disability in the United States [1], with 33% of persons aged 60 years and older having radiographic evidence of knee osteoarthritis [2]. Detection of cartilage abnormalities using magnetic resonance imaging (MRI) plays a substantial role in determination of the correct surgical, therapeutic and pharmacological intervention to alter or retard the course of disease in OA [3].

A number of different clinically available pulse sequences have been used in MRI of cartilage, each taking advantage of differing contrast characteristics of cartilage and adjacent tissues. The most widely used of these cartilage-specific sequences includes fat-saturated spoiled gradient-recalled echo (SPGR) and T2-weighted or intermediate-weighted fast spin-echo (FSE) imaging. FSE imaging techniques allow for the acquisition of high-resolution images in a relatively short amount of time while imparting increased contrast to the cartilage abnormalities. Also, of clinical importance, FSE imaging techniques are valuable pulse sequences in the diagnostic evaluation of other intraarticular structures including menisci, ligaments and subchondral bone [4–6]. As such, intermediate- and T2-weighted fast spin-echo imaging sequences, with and without fat suppression, have been advocated in the assessment of articular cartilage integrity [7–10].

With intermediate-weighted FSE sequences, cartilage appears intermediate in signal intensity, and joint fluid appears bright. Cartilage abnormalities can be seen as areas of reduced thickness, morphologic surface irregularities, including fraying and surface fibrillation, or as regions of relatively increased signal intensity within cartilage [11, 12], most likely reflective of increased intracartilaginous free water and the disruption of the collagen network [13, 14].

FSE imaging has been shown to have high sensitivity in detection of arthroscopically validated cartilage damage [7]. However, the sensitivity, specificity and accuracy of these pulse sequences performed in vivo at 3 T in detection and staging of articular cartilage lesions have not been validated in comparison to histology as the gold standard technique for assessing the severity of cartilage lesions.

The purpose of this study was therefore (1) to evaluate the sensitivity, specificity, and accuracy of sagittal in vivo 3-T intermediate-weighted FSE (iw FSE) sequences in the assessment of cartilage pathologies of the knee by using histology as the reference standard and (2) to correlate MR imaging findings typically found in OA such as bone marrow edema pattern (BMEP) and cartilage swelling with histological findings in patients undergoing total knee replacement.

Methods

The study was approved by the university’s Institutional Review Board. Patient privacy was strictly respected in compliance with the Health Insurance Portability and Accountability Act (HIPAA). Written informed consent was obtained from all patients. All authors declare no conflicts of interest.

Patients

Seven consecutive patients (three male and four female; average age 65.6 years) with diagnosed advanced osteoarthritis of the knee joint, preoperative MRI at 3 T and completed total knee replacement were recruited for this study. One patient underwent bilateral joint replacement surgery, increasing the number of knees included in the study to eight. Inclusion criteria were: (1) advanced osteoarthritis of the knee (Kellgren-Lawrence score 3 and 4) with limited function requiring total joint replacement, (2) referral for total knee replacement surgery. Exclusion criteria for patient recruitment were: (1) contraindications for MR imaging, (2) prior knee surgery and (3) secondary osteoarthritis of the knee due to trauma or inflammatory arthritis (such as hemophiliac arthropathy and rheumatoid arthritis).

MR imaging

Within a month before surgery, the patients’ symptomatic knee joints were examined at 3 T (Signa, GE, Medical Systems, Waukesha, WI) using an eight-channel transmit/receive phased array knee coil. Two sagittal cartilage dedicated MR imaging sequences were obtained: (1) a fat-saturated spoiled gradient echo (SPGR) sequence with the following parameters: repetition time (TR): 20 ms, echo time (TE): 7.5 ms, flip angle of 12°, matrix of 512×512 pixels, in-plane resolution of 0.293×0.293 mm², FOV of 16 cm, and slice thickness of 1 mm. Number of excitations was set to 0.75 to reduce acquisition time to 7 min 37 s. Partial data were acquired in Ky direction, and the data were synthesized to a full data set before image reconstruction. (2) An intermediate-weighted fat-saturated FSE sequence (TR/TE 4,300/51 ms, echo train length of 9, number of excitations 2 and acquisition time of 12 min 42 s, 45 sections, FOVof 16 cm (matrix 512×256) with an in-plane spatial resolution of 0.293×0.293 mm², a slice thickness of 2 mm and a section gap of 0.5 mm).

Histology

Tibial plateaus and femoral condyles of the knees (n=8) were resected during surgery and marked by the surgeon immediately with colored markers to allow correct anatomical comparison with preoperative MRI. Following resection, the specimens were fixed in 10% buffered formalin, slowly decalcified and embedded in paraffin. Maximum length for each embedded piece was 2 cm. Sagittal histologic sections (3–4 mm thick) from each cartilage specimen were obtained in the same orientation as in the sagittal preoperative MR scans and stained with Safranin-O (specific for proteoglycans), hematoxylin and eosin (H&E) for histological analysis [15]. Magnifications used were 4×, 10× and 20× objectives.

The location of the histological section was exactly recorded by anatomical measurements from lateral and medial joint borders as well as anatomical and pathological landmarks such as tibial spine and osteophytes. In all specimens only less severely affected compartments of the knee were analyzed, mostly from mid-part of the lateral joint compartment, in order to focus on correlations between histology and MRI findings in mild and moderate cartilage pathology. Depending on preoperative radiographic and MR imaging and intraoperative findings, the lateral joint compartments were chosen in varus osteoarthritis and the medial joint compartments were chosen in valgus osteoarthritis cases.

Image analysis

Great care was taken to match MR sections and histological slides based on intraoperative information, edge-distance measurements and morphological features. This was performed in consensus by four investigators, a trained bone pathologist (BJ), a radiologist (TML) and two investigators who were present during surgery, specimen preparation and histology (ES and JC) in consensus. For better comparability of histology and MR images, three to six 0.5–1-cm-wide (sagittal diameter) “observation units” (OUs) were defined on sagittal sections. Each of these was measured on the obtained histologic cartilage samples and the corresponding MR images. OUs were also matched using anatomical and pathological as well as morphological features. A total of 103 OU were compared side by side.

For comparison of MR images to the corresponding histological slides, a grading system was used based on previous classifications taking into account the severity of cartilage lesions with regard to thinning, surface abnormalities and intracartilaginous signal pattern (Table 1).

Table 1.

Grading system for comparison of MRI findings to histological features in osteoarthritic articular cartilage

	Thickness
Score	MRI	Histology
−1	Swelling	Swelling
0	Normal	Normal
1	<50% loss	Mild reduction
2	>50% loss	Severe reduction
3	Complete loss	Complete loss
	Surface		Signal pattern
Score	MRI	Histology	MRI	Histology
0	Normal	Normal	Homogeneous	Normal   staining/   slight   reduction
1	Fraying	Slight   fibrillation	Inhomogeneous	Severe   reduction
2	Fibrillation	Severe   fibrillation

Classification of cartilage thickness defects was performed as previously described [16]. In brief, defects were classified as full thickness, less and more than 50% loss as well as increased cartilage thickness (swelling). Concerning surface lesions, our classification was based on a similar classification previously published by Jones et al. using a modified Mankin score [17]; using confocal microscopy these investigators differentiated surface irregularities and clefts that extended into the transitional zone of the cartilage (less than 50% cartilage thickness) and deeper clefts extending into the radial and calcified zone (more than 50% cartilage thickness). Using this grading system we defined two separate types of “surface” lesions: (1) mild fibrillation and surface irregularities involving less than 50% of the cartilage and (2) deeper clefts involving the radial and/or calcified zone affecting more than 50% of the cartilage layer. The MRI findings of each OU were then compared to the corresponding region in the histological section for thickness and surface integrity using this grading system. The MRI signal pattern was compared to the proteoglycan content of the corresponding OU in histology as determined semiquantitaively by a modified Mankin scoring system based on Safranin-O staining [18] (Table 1). Abnormal signal pattern was defined as cartilage that was heterogeneous in signal and either too bright or too dark compared to normal cartilage and paired with moderate (~25–50% loss of staining) to severe (~>50% loss of staining) loss of proteoglycans based on the Mankin score. Cartilage with normal signal was paired with normal or mild loss of proteoglycans (~<25% loss of staining).

At the time of the analysis the radiologist was blinded to the microscopic, histological findings and the pathologist was blinded to results of the radiological analysis.

Statistical analysis

Statistical methods used included calculation of sensitivity, specificity and accuracy. The segmentation of cartilage within each knee created a set of measurements that were correlated with up to 18 measurements contributed by a single knee. For this reason all proportions were calculated using a generalized estimating equations (GEE) approach [19] to remove bias due to intra-knee correlation, as first suggested by Smith and Hadgu [20] and Leisenring, Pepe and Longton [21]. Statistical significance was determined using the GEE approach of logistic regression models. The significant level was set as 5%. We implemented these models using the Genmod procedure in SAS version 8.2 (SAS Institute Inc., Cary, NC).

Results

Overall diagnostic performance

The overall sensitivity, specificity and accuracy for the analysis of thickness, surface and signal pattern are reported in Table 2. The iw FSE images correctly revealed pathologically proven thinning of cartilage with 72% sensitivity, but specificity was lower (69%). Figure 1 shows an example of a correctly diagnosed loss of <50% of cartilage thickness. The outcome was reversed with regards to the cartilage surface abnormalities: The iw FSE images were less sensitive to cartilage surface abnormalities such as fraying and clefts (69% sensitivity), but MR diagnosis of normal cartilage surface corresponded to normal histology more frequently (74% specificity). Figure 1 and Figure 2 demonstrate histologically verified cartilage surface fraying that was also visualized in the MR images. The accuracies of diagnosis of pathologically proven thinning of cartilage or surface irregularities using the iw FSE images were 70% and 73%, respectively. Cartilage MR signal changes did not show strong correlation with degeneration as graded semiquantitatively by Safranin-O staining. However, Fig. 3 shows an example of increased cartilage signal in the FSE images that corresponded to a loss of proteoglycans histologically. Both sensitivity and specificity were substantially lower than those of thickness and surface measurements.

Table 2.

Overall sensitivity and specificity (with 95% confidence intervals) for the thickness, surface and signal pattern assessments of osteoarthritic cartilage assessed via MRI and compared to histological findings

	Thickness	Surface	Signal pattern
Sensitivity	72% [47%–88%]	69% [49%–84%]	36% [21%–54%]
Specificity	69% [55%–80%]	74% [45%–91%]	62% [34%–84%]
Accuracy	70% [56%–81%]	73% [57%–85%]	45% [36%–54%]

Fig. 1.

Sagittal fat-saturated FSE (a) and SPGR (b) images of the medial joint compartment and corresponding tibial histological section (H&E) (c) in a 54-year-old female patient. * in (c) indicates sectioning and folding artifacts. Magnified histological areas demonstrate osteophyte (d), fibrocartilage overlying the osteophyte (e) and synovial (pannus) tissue overlying normal cartilage. An area of bone marrow edema pattern is identified at the anterior part of the tibia (large arrow) in the region of the osteophyte (a), which corresponds to an area of fibrovascular tissue in the histological section (c, d). Increased fibrovascular tissue ingrowth (arrow in d) within the bone marrow of the osteophyte is observed. Fibrocartilage overlying the osteophyte (e) corresponds to an area of increased signal in (** in a). Black arrows in (a, b, c) indicate less than 50% thinning and surface fraying of the cartilage in the central tibia. Synovial (pannus) tissue in (c, e) corresponds to an area of bright signal in the FSE image (# in a and c)

Fig. 2.

Sagittal fat-saturated FSE (a) and SPGR (b) images of the lateral joint compartment and corresponding tibial histological section (H&E) (c) in a 51-year-old male patient. On the histological image (c) fraying with fibrillation involving less than 50% of the cartilage surface is shown centrally and posteriorly at the tibia. Magnification shows minor synovial tissue ingrowth in the posterior part, degrading the superficial cartilage layer. Corresponding findings on the FSE MR image (a), however, are relatively subtle. The superficial cartilage surface irregularity at the posterior tibial cartilage (highlighted and magnified region) and less severe centrally are demonstrated. Surface irregularities are not appreciated in the SPGR images (b)

Fig. 3.

Sagittal fat-saturated FSE (a) and SPGR (b) images of the lateral joint compartment and corresponding tibial histological section (Saf-O) (c) in a 74-year-old patient. Cutting artifacts (*) are demonstrated in the anterior part of the tibia (c). The FSE image (a) shows a T2 signal increase (arrow) and height reduction in the central region of the tibial cartilage that relates to cartilage thinning, fibrillation and proteoglycan loss by a 50% reduction in red-staining in the Saf-O staining (c, arrow). SPGR image does not well demonstrate these findings, yet shows some signal inhomogeneity. Signal inhomogeneity at the surface of the posterior tibia shows pannus tissue on the cartilage surface (** in a, b, c)

Regional diagnostic performance

Sensitivity, specificity and accuracy of the diagnosis of pathological cartilage thinning were higher for the anterior aspect of the femoral condyle and tibial plateau compared to the posterior aspect (differences between the two compartments were 17% for sensitivity, 7.5% for specificity and 12.3% for accuracy); however, differences were not statistically significant (P>0.05). Interestingly, surface lesions were revealed with higher sensitivity, specificity and accuracy in the posterior aspects of the femoral condyle and tibial plateau vs. the anterior aspect (differences were 46.5% for sensitivity, 28.3% for specificity and 46.6% for accuracy). However, though percent differences were larger, again no statistically significant differences were found (P>0.05). There were only very minor differences in the sensitivity, specificity and accuracy between the mesial and distal aspects of the femoral condyles or the tibial plateaux, which were not statistically significant.

Error analysis

Thickness assessment

Of the 32 histologically proven cartilage thickness defects, 21 (65.6%) were graded identically on histological evaluation and MR imaging. The 11 defects that were not graded identically differed by one grade in ten cases (90.1%), by two grades in only one case (9.9%) and by more than two in no cases. Grade 1 and 2 defects were not distinguished well from each other: only 11 of 21 histologically proven grade 1 defects (52.4%) were identified on the MRI as such; three were identified as grade 2 and seven were identified as normal (grade 0) by MRI. Of the two histologically proven grade 2 defects, one was scored identically by MRI and the other was scored as normal (grade 0). All grade 3 defects (n=7), however, were correctly identified on the FSE images. Over-diagnosis of cartilage thickness pathologies (25 of 103 observation units examined, 24.3%) was more common than under-diagnosis (8 of 103 observation units examined, 7.7%) (Table 3).

Table 3.

Error analysis for assessment of articular cartilage thickness via MRI as compared to histology as standard of reference

MRI
		Grade −1	Grade 0	Grade 1	Grade 2	Grade 3	Total
Histology	Grade −1	2	0	0	0	0	2
	Grade 0	6	49	12	4	0	71
	Grade 1	0	7	11	3	0	21
	Grade 2	0	1	0	1	0	2
	Grade 3	0	0	0	0	7	7
	Total	8	57	23	8	7	103

Surface assessment

Regarding surface pathologies, of the 67 histologically proven lesions, 37 (55.2%) were graded identically on histological evaluation and MR imaging. The 30 lesions that were not graded identically differed by one grade in 27 cases (90%) and by two grades in 3 cases (10%). Under-diagnosis of cartilage lesions (28 of 96 observation units examined, 29.1%) was more common than over-diagnosis (10 of 96 observation units examined, 10.4%) (Table 4). Figure 4 demonstrates examples of histologically verified cartilage lesions in a complete histological reconstruction of one sagittal medial knee section, some of which are not appreciated on the corresponding MR images.

Table 4.

Error analysis for assessment of articular cartilage surface abnormalities via MRI as compared to histology as standard of reference

MRI
		Grade 0	Grade 1	Grade 2	Total
Histology	Grade 0	21	8	0	29
	Grade 1	16	32	2	50
	Grade 2	3	9	5	17
	Total	40	49	7	96

Fig. 4.

Sagittal fat-saturated FSE (a, c, d) and SPGR (b) MR images of a 68 year-old female patient and complete histological reconstruction of the corresponding sagittal knee section (e, f, i=H&E staining; g, h= Safranin-O staining) at the medial joint compartment (Single asterisk sectioning and folding artifacts). Cartilage thickness and surface are relatively well maintained in MR and histological images, yet a number of pathologies are depicted in the histological section that are only partially visualized in the MR images: (I) Deep (>50%) fissures/clefts in the cartilage (two small black arrows) in (e) are not visualized in corresponding FSE (a) and SPGR (b) images (two small arrows). (II) Osteophyte (arrow, pound/number sign in a and e, f) with bone marrow edema pattern in (a) and fibrovascular tissue with multiple large vascular sinuses within the marrow (e, f). (III) In e and g (Double asterisk) a small area of thick fibro-cartilage, with reduced Safranin-O red staining indicating reduced proteoglycan content is shown. This is not visualized in the corresponding MR images (a, b), but in a neighboring section an area of increased cartilage signal intensity was demonstrated (arrow and Double asterisk in c). (IV) In (i) necrotic chondrocytes with myxoid degeneration of the surrounding matrix (Weichselbaum's lacunae) are demonstrated in the vicinity of a cartilage lesion (h), however, there is no corresponding pathology identified on the MR images. (V) In (h) horizontal cartilage delamination with reduced PG (Saf-O) content in the separated cartilage layer is shown (triple asterisk in e). The corresponding finding in theMR image (d) is a subtle line of increased signal intensity (triple asterisk) at the posterior femoral condyle

Correlation of bone marrow edema pattern and cartilage swelling with histological findings

MRI findings consistent with BMEP were visualized in three cases in the bone marrow of one patient where sufficient subchondral bone was available for histo-pathological analysis in the harvested specimens. In general patients with advanced Kellgren-Lawrence grade 3 and 4 OA also had cartilage damage at the less affected compartment together with additional BMEP. In all cases, subchondral ingrowth of fibrovascular tissue and increased bone remodeling was observed at the exact anatomical location of the bone marrow edema pattern in MRI (Fig. 1 and Fig. 4).

Cartilage swelling was observed in two MR studies. In one of these the swelling was correlated to pannus overgrowth consisting of proliferating synovial tissue on the articular cartilage surface (Fig. 1, Fig. 4 and Fig. 5); the other was correlated to severe surface fibrillation. In three cases, pannus was identified on the histological sections, but no evidence for them was found in the MR images.

Fig. 5.

Sagittal fat-saturated FSE (a) of the medial joint compartment and corresponding tibial histological section (Saf-O) (b) in a 54 year-old female patient. Cartilage thickening/swelling at the posterior aspect of the tibia with high signal intensity region (highlighted area in (a)). The corresponding Saf-O staining (b) shows an area of cartilage thickening due to synovial overgrowth (pannus tissue) with proteoglycan loss

Discussion

In this study, intermediate-weighted fat-saturated FSE sequences were clinically evaluated in the less severely affected knee compartments of patients with osteoarthritis who were scheduled for total knee replacement surgery. Although data on diagnostic performance of iw FSE sequences are available with arthroscopic validation [7–10, 22], our study provides new information on the diagnostic performance of this sequence by using histology as the standard of reference.

The choice of arthroscopy as the gold standard in any MR investigation of cartilage abnormalities has several important limitations, and its accuracy in the evaluation of cartilage disorders has been questioned [23]. Since arthroscopy is only capable of visualizing surfaces if the subchondral bone is exposed, it is inherently inaccurate in estimating cartilage thickness and the depth of lesions. The use of histology, on the other hand, allows for more exact matching of the MRI images to anatomical findings, permits the accurate and objective recording of the changes in cartilage thickness separately from the cartilage surface abnormalities and provides a microscopic view of the cartilage surface as well as subchondral bone changes associated with OA.

Using histological images as a reference, we were also able to pinpoint the histological findings correlating to the bone marrow edema patterns on MRI. In all three cases, the MRI evidence of BMEP correlated to fibrovascular tissue ingrowth of the bone and not edema per se. We suspect the increased blood flow to the area due to the new ingrowth of fibrovascular tissue to contribute to the hyperintense signal on intermediate-weighted images. Zanetti et al. [24] have found that among all BME patterns on MRI images, only a few correlate to pathology, and among those that do, only a minor fraction (2%) correlate to edema. Abnormal trabeculae or bone marrow fibrosis was more commonly correlated to BMEPs in their study.

In this study, we have used the less-severely affected compartments of the knee, mostly from the lateral joint compartment, in order to focus on correlations between histology and MRI findings of mild and moderate cartilage pathology. This fact reduced investigator bias for both methods. Our results indicate that iw FSE sequences are relatively well suited for the diagnosis of mild and moderate cartilage pathologies. Loss of cartilage thickness was predicted with 72% sensitivity and 69% specificity; however, sensitivity for cartilage surface integrity was lower and specificity was higher (69% and 74%, respectively). While others have assessed the diagnostic performance of MRI for cartilage lesions using histology [25–30], few [9, 10, 22] have determined the diagnostic performance of FSE sequences with comparison to histology. Our outcomes are comparable to some, but slightly lower than the findings of other investigators in sensitivity and specificity. Sonin et al. [10] reported a sensitivity of 59–73% and specificity of 87–90% for axial and coronal FSE images in detection of cartilage lesions. In a recent study by Yoshioka et al. [22], iw FSE sequences were found to have a sensitivity of 100% and specificity of 68%. Potter et al. [9] found 87% sensitivity and 92% specificity for the ability of FSE in detection of lesions. The higher specificity of FSE sequences reported in other studies is in large part attributable to the difference in methods of verifying cartilage abnormalities (arthroscopy vs. histology). Much subtler abnormalities of both cartilage thickness and surface are visualized using histology compared to arthroscopy. Lesions found and staged with arthroscopy are more likely to be detected using MRI (due to larger size) than those found using histology. The relatively lower sensitivity to lesions in our study might also be in part due to our exclusive use of sagittal MRI images for detection and evaluation of cartilage lesions, whereas others have used multiple planes to locate and stage abnormalities [8, 26, 27].

Interestingly, surface pathologogiaclly proven lesions were visualized with better sensitivity, specificity and accuracy at the posterior part of the femoral condyle and the tibial plateau versus the anterior region, though differences were non-significant. The better visualization of surface abnormalities may be due to more pronounced chemical shift artifacts at the anterior region of the femoral condyle and more complex anatomy of the anterior part of tibia, especially in the more mesial aspect of the tibial cartilage, yet numbers are too small for definite conclusions.

The intrasubstance cartilage signal pattern was not found sensitive or specific enough to characterize the proteoglycan content of the tissue as assessed semiquantitatively by Safranin-O staining, indicating the limited ability of iw FSE sequences to show early molecular changes in cartilage [31, 32] and the need for other MR techniques such as T1rho, T2 and dGEMRIC imaging in this regard [33, 34]. On the other hand it should also be considered that semiquantitative assessment with Safranin-O staining may also have limitations in predicting proteoglycan loss. Also, internal cartilage derangement is a more complex process involving changes in collagen structure and content [35]. However, histological analysis using H&E staining and Safranin staining does not provide sufficient information on collagen structure and orientation.

The iw FSE sequence has important advantages and shortcomings compared to the other clinically applied cartilage-dedicated sequences such as SPGR. In SPGR images, cartilage is only bright, and signal abnormalities are generally not observed, whereas iw FSE sequences do demonstrate changes in cartilage signal, believed to be caused by increased intracartilaginous water content as a result of loss of proteoglycans in OA [36], which, however, is not supported by the findings of our study. Further, SPGR imaging does not provide cartilage-fluid contrast, lowering its ability to delineate small cartilage defects and surface lesions as compared to FSE imaging; in a study by McGibbon et al. [27], for example, SPGR imaging at 1.5 T significantly under-predicted the actual depth of defects in cartilage, when compared to histology. However, the FSE sequence suffers from limited through-plane resolution relative to SPGR imaging. Currently, new 3D FSE sequences are being developed and investigated that provide isotropic voxels with a very thin slice thickness. Previous studies have shown the potential of these sequences at 1.5 T and 3.0 T for musculoskeletal imaging [37, 38]; however, clinical validation studies comparing these sequences with standard 2D FSE sequences concerning cartilage pathology have not yet been performed.

Our study has some limitations: the first is the low number of patients and the potential for clustering of data due to the isolation of multiple observation units in each patient’s knee in this study. However, the recruitment of patients in a consecutive manner and our lack of specific inclusion criteria beyond referral for total knee replacement surgery alleviate the potential bias in our study. This, together with the use of generalized estimating equations to account for data clustering, leads us to believe that the outcome measures would likely remain similar in a more expansive study. Additionally, the patients recruited for this study all suffered from advanced and painful osteoarthritis requiring surgery, and motion artifacts were sometimes introduced in the imaging as a result, potentially reducing the diagnostic ability of our imaging for smaller cartilage lesions. However, this limitation will be present for any imaging study of a patient’s symptomatic knee with suspected cartilage damage. Another limitation is the lack of inter- and intra-observer reproducibility because only one musculoskeletal radiologist evaluated the MR images. However, good reproducibility in the assessment of cartilage abnormalities has been reported previously [12].

In conclusion, this study demonstrated, using histology as a standard of reference, that fat-saturated iw FSE sequences at 3.0 T showed good performance in assessing cartilage thickness and surface lesions, while signal changes of the cartilage were not suited to characterize the amount of cartilage degeneration. Interestingly, all cases of BMEP identified showed ingrowth of fibrovascular tissue on histological examination, which so far has not been documented in the previous literature. Also we found that pannus overgrowth is related to osteoarthritis and may lead to apparent cartilage swelling on MRI.