We tested the hypothesis that sagittal DESSwe at 3 T may overcome some of the current limitations of quantitative analysis of cartilage morphology by permitting one to obtain accurate and highly reproducible quantification of cartilage morphology in all knee cartilage plates from one single (sagittal) dataset at reasonable acquisition times. In the femorotibial joint (tibia and central femoral condyles), we found that measures of cartilage morphology derived from DESSwe were consistent with those from FLASHwe, and that precision errors for unpaired analysis were similar between both protocols. Precision errors in the posterior femoral condyles (sagittal DESSwe) tended to be similar or lower than in the central femoral condyles and were substantially less than the intersubject variability, indicating that sagittal DESSwe can be used to reliably determine cartilage morphology in the posterior aspects of the femur. In the patella and trochlea, in contrast, precision errors of sagittal DESSwe were considerably higher than previously reported for unpaired analysis of 1.5 mm sagittal acquisitions with FLASHwe at 1.5 T.7,14
The difference should, however, be interpreted with care, as the results do not originate from the same study.
In contrast to most previous studies, repeat datasets were not segmented as pairs within the same analysis session, but in random order with blinding to subject identification. Two previous studies reported considerably higher errors for unpaired (resegmentation) conditions (different sessions) than for paired analysis under short term conditions (same analysis session) for sagittal and axial FLASHwe sequences, respectively.7,14
The precision errors reported here are likely to overestimate the precision error relevant to a longitudinal study, in which baseline and follow up datasets are read in parallel. A previous study with cor FLASHwe at 3 T (1.5 mm slice thickness) in which repeat datasets were processed in pairs,4,10
reported RMS CV% values of
2.5% for cartilage volume and thickness in MT and LT, of
3% in cMF and cLF, and values of 1.1–1.4% for areas.4,10
In the current study with unpaired analysis, the precision errors for volume and thickness were
4% in MT,
6% in LT and cLF, and
7% in cMF, and those for areas
6.3%. These results emphasise the higher reproducibility of quantitative cartilage measures in paired data analyses, because segmentation regions (for instance, number of slices) can be kept more consistent.
Precision errors for corMPR DESSwe were similar to cor FLASHwe despite the lower in‐plane resolution of corMPR DESSwe. One exception was the smaller errors in surface area measurements in cMF and cLF by corMPR DESSwe. One explanation is that with an MPR, the double oblique orientation could be made more consistently, whereas the cor FLASHwe orientation could not be optimised post hoc. In fact, the average deviation in rotational alignment relative to the posterior tips of the femoral condyles within pairs was only 0.1 slices for corMPR DESSwe, but 0.6 for cor FLASHwe. The precision errors for sag DESSwe were similar to the cor FLASHwe, except for the surface areas in LT. The larger errors were in part due to larger differences in the number of slices segmented in sag DESSwe pairs (2.3) than in corMPR DESSwe (0.9) and cor FLASHwe (0.6).
Precision errors in the femoropatellar joint (sag DESSwe) were relatively large (3.7–7.7% for volume and thickness, and 3.4–8.5% for areas) compared with previous studies4
with sagittal and axial FLASHwe at 1.5 T and 1.5 mm slice thickness. These results correspond with the subjective difficulty of separating P and TrF in the sag DESSwe owing to low cartilage‐cartilage contrast in the contact zone. Because the thinner slice thickness (0.7 mm) and higher field strength (3 T) should have advantages over acquisitions at 1.5 mm thickness and 1.5 T,7,14
our findings indicate that sagittal DESSwe may be less suitable than FLASHwe for analysing femoropatellar cartilage. A face to face comparison between axial FLASHwe, sag FLASHwe, sag DESSwe, and axial DESSwe should be undertaken to clarify this issue.
For cross calibration of corMPR DESSwe with cor FLASHwe at 3T, correlations were high (>0.95), except for VCtAB and ThCtAB at cLF (0.91 and 0.88). The same applied for sag DESSwe (>0.90), except for VC and ThCtAB of cLF (0.89 and 0.88), despite the differences in section orientation and spatial resolution. Unfortunately, measurements in the femoropatellar joint could not be validated, because no sagittal or axial FLASHwe images were available for this purpose as part of this pilot study.
This study poses several other limitations: only a limited number of participants (n
19) were evaluated and only two repeat scans were acquired for each protocol. The aggregate MRI examination time of almost 2 hours precluded increasing the number of repeat scans. Previous studies have shown that, when a FLASHwe sequence is used, changes of cartilage morphology can be measured with changes in mechanical loading conditions16,17
and in OA18,19,20
over relatively short period of time. The sensitivity of DESSwe to longitudinal changes in cartilage morphology remains to be tested. Because the OAI will obtain image data from 5000 subjects at yearly intervals over a period of 5 years, a basis for such a study is under way.
When discussing the potential advantages and limitations of DESSwe, the following points are worth noting: (a) DESSwe provides more heterogeneous signal characteristics of the cartilage, the joint space, and other periarticular tissues, which may be of considerable interest for clinically evaluating or scoring OA. The high cartilage to fluid contrast may be of particular advantage in detecting focal cartilage lesions, whereas the present study focused only on integral quantitative measures of cartilage morphology, such as volume and thickness. (b) When acquiring coronal images with a FLASHwe sequence, relatively large areas of the femur (trochlea, posterior condyles) are not accessible to analysis. With sagittal DESSwe it is possible to image the entire knee with high resolution (0.7 mm slice thickness) in about 10 minutes. The acquisition time for the coronal FLASHwe was shorter (8 minutes) but that for sagittal FLASHwe at comparable resolution would be >20 minutes. (c) At 0.7 mm slice thickness, segmentation time for sagittal images is twice as long as for 1.5 mm cor FLASHwe (or corMPR DESSwe). Additional segmentation time is also required for DESSwe contrast because of the more heterogeneous signal and, subjectively, the experienced readers found DESSwe more challenging than FLASHwe. Some of these limitations may be overcome by more automated segmentation algorithms, but the complex signal characteristics of DESSwe may render this goal an even larger challenge than for FLASHwe. (d) DESSwe is currently available from only one MR system vendor, which may limit its applicability in multicentre studies.
In conclusion, DESSwe imaging at 3 T permits quantitative measures of cartilage morphology in the femorotibial joint to be obtained with a high level of accuracy and precision in subjects with and without OA. The performance (reproducibility) of sag DESSwe in the femoropatellar joint is, however, less promising. Future studies are needed to investigate the accuracy of DESSwe versus FLASHwe in the femoropatellar joint and the sensitivity of DESSwe to longitudinal change of cartilage morphology.