Searchable abstracts of presentations at key conferences in endocrinology
Endocrine Abstracts (2014) 36 OC3.5 | DOI: 10.1530/endoabs.36.OC3.5

BSPED2014 Oral Communications Oral Communications 3 (9 abstracts)

Statistical prediction of HRpQCT microstructural trabecular parameters using 1.5T skeletal MRI

Paul Dimitri 1 , Karim Lekadir 3 , Elspeth Whitby 2 , Paul Armitage 2 , Corné Hoogendoorn 3 & Alejandro Franji 2

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

Background: High resolution peripheral quantitative computed tomography (HRpQCT) can accurately determine three-dimensional in-vivo skeletal microstructure. However, HRpQCT is limited to the ultradistal radius and tibia (9 mm) imaging. MRI may be an alternative approach to cortical and trabecular bone analysis; to date there is limited information regarding the accurate quantification of trabecular bone.

Method: Ninety-three 13–16 years-old underwent ultra-distal HRpQCT and skeletal MRI (sMRI). Participants underwent 2/6 sMRI sequences (T1, T2, T2*, gradient-echo, fiesta, and ultrashort-time echo). sMRI sequences were delineated into cortical and trabecular compartments using segmentation software. Owing to the varying number of MRI slices available (one to four per subject), and the geometrical variation in trabecular area between participants, we calculated a number of image descriptors from trabecular signal intensities to obtain the same level of information for each case. Image descriptors included statistical measures (mean, S.D., entropy), geometrical measures (e.g. primitive emphasis primitive uniformity), and textural measures (e.g. homogeneity, contrast, and fractal dimension). Kernel partial least squares was used to find an optimal non-linear predictor model from the data relating sMRI to HRpQCT parameters.

Results: Leave-one-out cross-validation experiments were carried out to assess the mean prediction accuracy of HRpQCT from sMRI. To this end, the data used for testing the predictive model was removed from the construction of the statistical predictive model itself. We then calculated the relative errors of the predictions in percentage, i.e. error=(prediction−measurement)×100/measurement. sMRI predicted trabecular number, spacing, and thickness to within 7.52, 9.51, and 7.43% of HRpQCT respectively.

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

Volume 36

42nd Meeting of the British Society for Paediatric Endocrinology and Diabetes

British Society for Paediatric Endocrinology and Diabetes 

Browse other volumes

Article tools

My recent searches

No recent searches.