Deriving pQCT-Type Spatial Density from DXA Images.

INTRODUCTION

Musculoskeletal overuse injuries are a serious problem in the military, particularly in basic combat training, resulting in hundreds of millions of dollars lost because of limited duty days, medical treatment, and high rates of reinjury. Injury risk models have been developed using peripheral computed tomography (pQCT)-based injury correlates. However, pQCT image capture on large number of recruits is not practical for military settings. Thus, this article presents a method to derive spatial density pQCT images from much lower resolution but more accessible dual-energy X-ray absorptiometry (DXA).

MATERIALS AND

METHODS

Whole-body DXA images and lower leg pQCT images for 51 male military recruits were collected before a 40-day School of Infantry. An artificial neural network model was constructed to relate the DXA density profiles to spatial pQCT density at the 38% and 66% tibial locations.

RESULTS

Strong correlation, R2 = 0.993 and R2 = 0.990 for the 38% and 66% pQCT slices, respectively, was shown between spatial density predicted by the artificial neural network model and raw data.

CONCLUSIONS

High potential exists to create a practical protocol using DXA in place of pQCT to assess stress fracture risk and aid in mitigating musculoskeletal injuries seen in military recruits.