Osteoarthritis (OA) is a debilitating whole joint disease that has a higher prevalence in individual’s post-traumatic knee injury [1]. Despite being a disease that affects the whole joint, most MRI studies focus on analyzing tissues separately. While some studies have shown that the underlying subchondral bone affects cartilage health, the relationship between bone and cartilage OA changes remains unclear [2,3]. New integrated PET/MR systems allow simultaneous, quantitative assessment of the bone and cartilage morphology, biochemistry, and metabolic activity [4-6]. Comprehensive 3D visualization and quantification of bone metabolic activity linked to cartilage matrix changes may help identify early OA and help track progression. This work aims to develop a tool that combines projection maps and cluster analysis [7,8] to compare elevated bone metabolic activity between injured vs. contralateral knees as well as to spatially corresponding cartilage quantitative parameters.

Acquisition and Parametric Mapping: A 3T PET-MR hybrid system (GE Healthcare, Milwaukee, WI) was used for the simultaneous imaging of six subjects with previous unilateral knee injuries. All scans were done under IRB approval. Using a 16-channel flex coil, each knee was scanned with a 30 minute MRI protocol which included quantitative Double-Echo in Steady-State (DESS) [9] for T2 mapping and thickness measurements, T1rho prepared 3D-fast spin echo with variable flip angle refocusing (CubeQuant) [10] for T1rho mapping and a T2-weighted Fast Spin Echo sequence (FSE) for OA feature identification. PET data was acquired simultaneous with MRI following an injection of 2.5-5mCi ¹⁸F-NaF. PET standard uptake value (SUV) maps were generated by normalizing uptake for patient weight, injection tracer dose and decay time.

Image Processing: Using the quantitative DESS images, slice-by-slice segmentation was performed to delineate the cartilage and bone (Figure 1a). 2D projection maps of cartilage and bone quantitative data were created using a previously developed method (Figure 1b) [8]. Two experienced radiologists identified OA features such as bone marrow lesions (BMLs) and osteophytes on MRI. Five projection maps showing PET ¹⁸F-NaF maximum pixel SUV (SUV_max), the location of bone OA features, cartilage quantitative T2 and T1rho relaxation times and thickness measurements were created.

Cluster Analysis: Cluster analysis was used to track areas of PET uptake across knees and tissues. Clusters were identified in each projection SUV_max map using an intensity threshold of SUV_max greater than 4.0 and a contiguous area of 3.8mm². The amount of the cartilage plate covered by these clusters was defined as the percent cluster area (%CA). Eight locations of OA features in the injured knees that corresponded with identified clusters were used to compare SUV_max uptake and mean T1rho cartilage relaxation times in the injured and contralateral knees. A Wilcoxon signed-rank test was used to compare the percent cluster areas between knees and the bone and cartilage quantitative values within the eight OA feature clusters.

There was significant SUV_max increase (p<0.01) between the eight OA feature clusters of the injured and contralateral knees. Figure 2 shows representative PET uptake differences between knees in two subjects. Within the injured knees, high ¹⁸F-NaF uptake in the bone generally corresponds with structural defects such as BMLs and osteophytes (Figure 3 – white arrows) or elevated T2 and T1rho quantitative values (Figure 3 – black arrows). The mean (+ std) %CA’s in the injured knee of the five subjects was 4.9 (+ 8.9) compared to 0.078 (+ 0.18) in the contralateral knee. Additionally, there was an increasing trend in both the SUV_max values and the mean T1rho values of the eight OA feature clusters between contralateral and injured knees (Figure 4). The significant differences in SUV_max uptake and the higher mean T1rho values for the injured knee vs. the contralateral knee demonstrate the potential of a combined outcome measure that quantifies bone and cartilage OA abnormalities. Some areas of high uptake correspond to higher quantitative values within the cartilage while others that did not show high T2 or T1rho relaxation times correspond to thinner cartilage. These observations suggest that apart from the uptake of ¹⁸F-NaF correlating with later stages of degeneration, it may provide complementary information on earlier OA changes and could improve overall sensitivity when combined with quantitative MR parameters. We demonstrated the full visualization of quantitative bone and cartilage data using the projection maps and identified significant differences in PET uptake between knees as well as high bone metabolic activity that corresponded with elevated T2 and T1rho relaxation times. This tool could help determine the relationship between bone metabolic activity and the development of cartilage and bone damage and may eventually help evaluate disease modifying OA drugs.