Category: Central DXA: (DXA, TBS)

30 - Quantification of the effects of soft tissues on the experimental variogram, the basic algorithm of TBS

Purposes: TBS is derived from the initial slope of the log-log experimental variogram of the DXA bone image, and the TBS computation process includes a soft tissue compensation. The aim of this study is to quantify the influence of soft tissues on the variogram, and provide background information on how tissues could influence TBS, to better implement corrective adjustments if necessary.

Material &
Method:
We used a set of 15 dried lumbar or low thoracic human vertebrae, set in AP position. Soft tissues were simulated by a mix of homogeneous layers of HDPE and PVC, achieving tissue thickness and fat percent compatible with normal human variation. First, all vertebrae were scanned at a constant thickness of 17cm, with fat percent varying from 4% to 45%. Then, all vertebrae were scanned at a constant fat percent of 30%, with thickness varying from 13cm to 25cm. Vertebrae were scanned on GE-Lunar Prodigy device, using Standard mode. We called rawTBS the initial slope of the log-log variogram, computed directly without any soft tissue compensation algorithm. For each vertebra, rawTBS was computed over the largest rectangular region inside the vertebral body. RawTBS values for all vertebrae were averaged for each tissue configuration and plotted against the relevant variable.


Results:
A positive relationship was observed between rawTBS and fat percent, with an almost flat slope of 0.0012 per percent of fat (r²=0.97). A negative relationship was observed between TBS and tissue thickness, with a slope of -0.055 per cm of tissue (r²=0.99). These results are presented in Figure 1.

Discussion /
Conclusion:
This study showed that the effect of soft tissues on rawTBS depends mainly on the thickness of soft tissues and to a smaller extent on the tissue fat percent. Indeed, a change of 45% in fat percent appeared to have an equivalent absolute impact on rawTBS as a change of 1cm of tissue thickness. This makes tissue thickness the most relevant tissue characteristic, and implies that the correlation between rawTBS and soft tissues is mostly negative. In addition, since noise is known to influence the slope of the log-log variogram and depends to some extend on tissue thickness, we also studied their relationship, to be published soon. All these findings are already integrated in TBS iNsight, as the TBS computation process includes soft tissue compensation techniques. These compensations are notably dependent on the DXA device brand and model, as well as on the scan mode.

Didier Hans

Professor
Center of Bone diseases, Bone & Joint Department, Lausanne University Hospital, Lausanne, Switzerland
Lausanne, Vaud, Switzerland

Franck Michelet

Medical imaging scientist
Medimaps SASU, Merignac, France
Merignac, Aquitaine, France