Osteoporosis is a skeletal disorder that predominantly affects postmenopausal women. It is characterised by declining bone strength and can lead to bone fractures. Osteoporosis screening is routinely performed with Dual X-ray Absorptiometry (DXA), which measures areal Bone Mineral Density (BMD). DXA is a standardised technique and is widely used with high precision, but it has several limitations in monitoring interventions and predicting fractures. DXA could overestimate BMD in heavier subjects or those with diabetes, aortic calcifications or degenerative disease, among others¹. In addition, BMD only partly explains bone strength; the properties of bone that are not assessed with BMD are defined as bone quality^2,3. The different components of the bone (cortical shell that encapsulates trabecular bone) contribute to bone strength in different ways^{4, 5} and DXA, as a projection method, fails to take this into account. 3D imaging methods, such as Quantitative CT (QCT) and MRI, can be used to assess bone strength and the lack of ionizing radiation makes MRI attractive for serial measurements. As the spine and the proximal femur are more challenging sites to image with MRI (presence of hematopoietic marrow, lack of dedicated coils), research on bone MRI has focused mostly on peripheral sites (tibia, heel, and radius)⁶. The femoral neck, however, is a common site for osteoporotic fractures, so accurate assessment of bone quality in this site is important in clinical practice. In this abstract, we present a method for assessing trabecular bone density, which has been shown to predict the risk of fracture⁴, in vivo on the proximal femur using MRI.Of the 100 post-menopausal Chinese-Singaporean women that will be recruited in this study, data from 9 have been analysed to date (average age 59.6±5 years). All subjects underwent MRI (3T Siemens Prisma), QCT (Siemens mCT) scans of the proximal femur and whole body DXA (Hologic Discovery QDR 4500). A proton density (PD, TR=929ms, TE=54ms, 32 averages, acq time ~19min) sequence was used to acquire 6 coronal slices of the proximal femur of the non-dominant leg (voxel size 0.26x0.26x1.5mm), using an 18-channel body array coil (Siemens). Both hips were scanned with QCT, using a commercial phantom (Mindways QCT Pro). Images of a PD dataset and a QCT image slice are shown in Fig.1 and Fig. 2, respectively. Bias field correction was performed on the PD images⁷ and DXA equivalent volumes of interest (VOIs), femoral neck and the total proximal femur, were drawn manually on the PD images. Due to limitations in the resolution (trabecular size is 100-150um), PD voxels cannot contain pure trabecular bone, so calculating bone content per voxel (bone volume fraction - BVF) was a more realistic approach than classifying voxels as bone or marrow. BVF was calculated using the method proposed by Vasilic and Wehrli⁸. BVF maps of the total femur and femoral neck of the dataset in Fig. 1 are shown in Fig. 3 and Fig. 4, respectively. The PD VOIs were registered to the QCT images, and volumetric BMD (vBMD) was calculated for each VOI. Analysis was performed using in-house developed software (MATLAB R2013b, ITK) and dedicated software (Fiji⁹, ITK-Snap, elastix^{10, 11}).DXA measurements gave an average BMD of 0.61±0.12g/cm² for the femoral neck and 0.76±0.11 g/cm² for total femur of the non-dominant leg of the subjects. For each subject, the average BVF per region (MRI) was compared to the average vBMD (QCT) of the same VOI and the respective areal BMD (DXA). Average BVF can be considered a surrogate measure for trabecular density. Linear regression analysis showed strong correlations between PD and DXA BMD as well as PD and QCT vBMD measurements, see Table 1. The stronger correlation of average BVF to average QCT vBMD was expected, since the QCT measurements were performed on the same VOI as the MRI measurements, while a bigger volume, which also contained cortical bone, contributed to the DXA BMD measurements.In this abstract, we presented a method to assess the trabecular bone density of a trabecular VOI in the proximal femur using MRI. The strong correlations of trabecular BVF with DXA BMD and QCT vBMD measurements show that it could be used to assess bone quality, instead of DXA or QCT. Extensive evaluation is needed to prove that BVF could be a useful biomarker in clinical practice. Such a biomarker may predict fracture risk and be sensitive to bone quality changes at different time points, due to a nutritional or drug intervention.