Bone Abstracts (2015) 4 OC2 | DOI: 10.1530/boneabs.4.OC2

Feature-based recognition of trabecular microstructure using 1.5T magnetic resonance imaging: a new methodology

Paul Dimitri1,2, Karim Lekadir3, Corne Hoogendoorn3, Paul Armitage1, Elspeth Whitby1, David King1 & Alejandro Frangi1


1The University of Sheffield, Sheffield, South Yorkshire, UK; 2Sheffield Children’s NHS Foundation Trust, Sheffield, South Yorkshire, UK; 3Universitat Pompeu Fabra, Barcelona, Spain.


Background: Magnetic resonance imaging (MRI) is used clinically to assess bone marrow, muscle and joints. The assessment of cortical and trabecular structure using MRI may provide further insight into the muscle–bone–bone marrow unit. Previous studies using MRI to evaluate microarchitecture are confined to research due to the need for specially adapted coils and navigator software to limit motion artifact. We present a novel statistical method using HRpQCT to determine the accuracy of a clinical 1.5 Tesla (T) MRI in quantifying trabecular microstructure.

Methods: We recruited 96 healthy 13–16 years old to undergo HRpQCT and 1.5T MRI of the non-dominant ultradistal tibia. Participants underwent two of the following axial MRI sequences: T1-weighted Fast Spin Echo (T1w), T2-weighted Fast Spin Echo (T2w), T2*-weighted Gradient Echo (T2*w), FIESTA, Ultrashort Time Echo (UTE), Ultrashort Time Echo High Resolution (UTE-HR). By relating trabecular parameters derived from the HRpQCT images, contextual image features contained within a defined region of interest within low resolution MRI sequences were used to develop a statistical prediction model designed to predict trabecular microstructural parameters. Image descriptors included statistical variability (mean intensity, standard deviation, skewness, and kurtosis), pattern repeatability (using grey level co-occurrence matrices), and pattern complexity (using run-length analysis and fractal dimension). Kernel partial least squares was used to find an optimal non-linear predictor model from the data relating MRI sequences to HRpQCT parameters. Prediction errors for the trabecular indices (trabecular thickness, spacing and number) were determined by using the different MRI sequences as the input of the prediction model.

Results: The FIESTA and UTE-HR image sequences demonstrated the highest accuracy in predicting all three trabecular parameters (12.01±3.44%, 12.08±4.48% respectively). T1w provided the highest predictive value in quantifying trabecular spacing (7.42% average error). T2w and T2*w most accurately predicted trabecular thickness (9.51%) and trabecular number (7.51%), respectively. Combining MRI sequences in the model to predict individual trabecular components did not improve the accuracy

Conclusions: Using the established predictive model, 1.5T MRI sequences can predict trabecular number, spacing, and thickness to within 10% of the values derived from HRpQCT. This study demonstrates the future potential of clinical MRI in assessing trabecular bone.

Disclosure: The authors declared no competing interests.

Article tools

My recent searches

No recent searches.