Quantitative Imaging Biomarkers of Knee Cartilage Composition

For a long time, radiography and subsequently conventional magnetic resonance imaging (MRI) were used as imaging biomarkers for evaluating cartilage morphological disease state in osteoarthritis (OA). Because research is switching its focus towards disease modification or even prevention to target OA at an early stage, imaging techniques that measure cartilage composition rather than its morphology became of interest. Several MRI and computed tomography (CT) based quantitative imaging biomarkers for cartilage composition were developed. These techniques were advocated to allow a quantitative measure of the sulphated glycosaminoglycan (sGAG) content, an important composite of the cartilage extracellular matrix. The main aims of this thesis is based have been divided between MRI and CT based quantitative imaging biomarkers since their different stage of application in research. MRI has already been applied in human OA research, whereas CT was still to be translated and implemented in clinical research.

The first part of this thesis focused on MRI based techniques and aimed at optimization of image post processing, assessing reproducibility, comparison of different MRI sequences and application in clinical OA research.

Since accurate image post processing is of utmost importance to generate reliable and robust quantitative MRI outcomes, an imaging post processing tool was developed and described in chapter 2. This tool corrects for intra-sequence patient motion during acquisition of quantitative MR images, by applying image registration reducing errors and incorrect outcomes. This resulted in 6-14% improvement in accuracy of delayed gadolinium-enhanced MRI of cartilage (dGEMRIC) T1 relaxation time. Using image registration, the tool also allows assessment of the same cartilage region throughout multiple MRI acquisitions, which makes analyses less time consuming. Finally, the algorithm also involves a fitting technique which corrects for unreliable quantitative MRI biomarker data by calculating a weighted mean outcome for all voxels in a specific cartilage region based on the inaccuracy of each voxel. Because of these abilities and the fact that this tool could be used in any quantitative MRI biomarker, e.g. T1rho-mapping or T2-mapping, the image post processing tool was used in all chapters in this thesis where MRI based measures were used for cartilage sGAG content.

Along with robust image processing tools, the outcomes of the MRI exam itself should also be reproducible in order to be able to apply the particular technique in cross-sectional or longitudinal study designs. Therefore, chapter 3 described a reproducibility study of dGEMRIC acquired at 3 Tesla in early stage knee OA patient. It was shown that dGEMRIC is highly reproducible in terms of results in large cartilage regions, as well as for differentiating between spatial distributions of diverse cartilage quality within a single slice. dGEMRIC can therefore be used as an imaging biomarker in cross-sectional and longitudinal study designs. In addition, a threshold for defining significant changes in dGEMRIC results for longitudinal follow-up was determined.

T1rho-mapping has been proposed as a non-contrast-enhanced alternative to dGEMRIC for sGAG quantification in clinical studies. However, no thorough validation has been performed comparing both techniques within the same OA patients using a reference standard for cartilage sGAG. Therefore, in chapter 4 an in vivo comparison and validation study assessing the capability of dGEMRIC and T1rho-mapping was performed. In knee OA patients, dGEMRIC results strongly correlate with cartilage sGAG content, whereas T1rho-mapping did not. Therefore, it appears that T1rho-mapping cannot be regarded as an alternative for dGEMRIC to measure cartilage sGAG content in clinical OA research. It was also shown that results of dGEMRIC are also partially influenced by the collagen content of cartilage. Despite the fact that dGEMRIC cannot be regarded as entirely sGAG specific, the difference in strength in correlation between dGEMRIC and sGAG (r=0.73) and collagen content (r=0.40) suggests that dGEMRIC may be used as an imaging biomarker for sGAG content. The outcomes of T1rho-mapping did also not correlate with cartilage collagen content. Therefore, the use of T1rho-mapping as imaging biomarker in OA is questionable.

Since dGEMRIC has been shown to be a robust and valid sGAG imaging biomarker in longitudinal studies, in Chapter 5 we used dGEMRIC to assess the effects of a potential disease-modifying OA drug, i.e. intra-articular injections with hyaluronic acid, in patients with early stage knee. This therapy has been advocated to improve cartilage composition by increasing the cartilage sGAG content. There was, however, no effect of the injections on cartilage sGAG content measured with dGEMRIC and using the threshold for change as defined in chapter 3. Although promising results were and are still being generated in in vitro and in vivo animal research with regard to improving cartilage composition using disease-modifying therapies, large human OA treatment trials have yet to be set up. The use of quantitative image biomarkers for cartilage composition is, in addition to clinical examination and questionnaires, a desirable outcome measure in such studies.

OA is nowadays regarded as a whole joint disease rather than a disease of only cartilage, therefore protocols imaging several relevant tissues within the knee joint became of interest in clinical research. Since articular cartilage and the meniscus are both partially composed of sGAG, the ability of dGEMRIC to serve as imaging biomarker for both tissues in one MR examination was assessed in chapter 6. It was shown that T1 relaxation times of degenerated menisci on conventional MRI were borderline significantly lower compared to healthy menisci. These differences in outcomes of morphologically healthy and degenerated menisci were highly reproducible independent of the anatomical location of the assessed menisci, showing the potential of contrast-enhanced MRI as a quantitative imaging biomarker for assessment of cartilage and meniscal composition. However, future research is needed to assess the role of contrast-enhanced MRI for multiple OA tissues within one examination.

The second part of this thesis focused on CT based techniques and aimed at translating CT arthrography from a microscopic to a macroscopic setup and making suitable to be used in clinical OA research.

Similar to dGEMRIC, contrast-enhanced CT can be used as a quantitative imaging biomarker for cartilage composition. Previously, microCT arthrography (μCTa) was successfully used as sGAG biomarker in in vivo animal OA research. Chapter 7 describes an ex vivo study in human cadaveric knee joints in which results of CT arthrography (CTa) are shown to be strongly correlated with cartilage sGAG content measured using a reference standard for sGAG. This study shows the potential of CTa as an imaging biomarker in human OA research.

Compared to MRI, CT has the disadvantage of using ionizing radiation. In chapter 8 it was shown that when only 10% of the radiation dose used in chapter 7 was used, it is still possible to measure cartilage sGAG content in cadaveric human knee joints. Using such a low radiation dose could be of interest especially if CTa would be used in longitudinal study designs or if the technique would be used in relatively young patients. Moreover, chapter 8 showed that CTa can determine spatial differences in sGAG content in adjacent cartilage regions. However, this capability is reduced at lower radiation dose and the amount of radiation used to acquire CTa should therefore be adjusted to the research question.

After translating μCTa to a clinically applicable CTa protocol, chapter 9 describes an in vivo validation study for CTa to measure cartilage sGAG content in humans with knee OA. It was shown that outcomes of CTa have a strong correlation with cartilage sGAG content in human OA tibial and femoral cartilage. In agreement with chapter 4, it was again shown that contrast distribution throughout the cartilage is also partially influenced by its collagen content. The correlation of CTa with the collagen content of cartilage (r=0.56), however, less strong compared to the correlation between CTa and its sGAG content (r=76). Because of its high relative high resolution, low coasts and speed of acquisition, CTa may be used as an alternative to dGEMRIC in future human OA research. This is also true for patients with relative or absolute contra-indications for MRI.

The research on which this thesis is based forms a contribution towards the application of quantitative imaging biomarkers of cartilage composition in OA. The most important results can be summarized as follows: -Until now, dGEMRIC is the only validated and robust imaging biomarker for quantifying cartilage sGAG content in human OA subjects in clinical research.  -dGEMRIC and dGEMRIM may be capable of assessing cartilage and meniscus sGAG content within one contrast-enhanced MRI examination. -T1rho-mapping appears not suitable as an alternative to dGEMRIC for quantifying cartilage sGAG content in articular cartilage and seems therefore not applicable as an imaging biomarker for OA research. -CT arthrography may be suitable as an imaging biomarker for quantitative measurement of cartilage sGAG content in clinical OA research and therefore could be considered an alternative to MRI. -Sophisticated imaging post processing tools are of utmost importance to generate reliable outcomes of quantitative MRI biomarkers of cartilage composition