Quantitative characterization of metastatic disease in the spine. Part II. Histogram-based analyses.

Radiological imaging is essential to the appropriate management of patients with bone metastasis; however, there have been no widely accepted guidelines as to the optimal method for quantifying the potential impact of skeletal lesions or to evaluate response to treatment. The current inability to rapidly quantify the response of bone metastases excludes patients with cancer and bone disease from participating in clinical trials of many new treatments as these studies frequently require patients with so-called measurable disease. Computed tomography (CT) can provide excellent skeletal detail with a sensitivity for the diagnosis of bone metastases. The purpose of this study was to establish an objective method to quantitatively characterize disease in the bony spine using CT-based segmentations. It was hypothesized that histogram analysis of CT vertebral density distributions would enable standardized segmentation of tumor tissue and consequently allow quantification of disease in the metastatic spine. Thirty two healthy vertebral CT scans were first studied to establish a baseline characterization. The histograms of the trabecular centrums were found to be Gaussian distributions (average root-mean-square difference=30 voxel counts), as expected for a uniform material. Intrapatient vertebral level similarity was also observed as the means were not significantly different (p > 0.8). Thus, a patient-specific healthy vertebral body histogram is able to characterize healthy trabecular bone throughout that individual's thoracolumbar spine. Eleven metastatically involved vertebrae were analyzed to determine the characteristics of the lytic and blastic bone voxels relative to the healthy bone. Lytic and blastic tumors were segmented as connected areas with voxel intensities between specified thresholds. The tested thresholds were mu-1.0 sigma, mu - 1.5 sigma, and mu - 2.0 sigma, for lytic and mu + 2.0 sigma, mu+3.0 siema, and mu + 3.5 sigma for blastic tissue where mu and sigma were taken from the Gaussian characterization of a healthy level within the same patient. The ideal lytic and blastic segmentation thresholds were determined to be mu-sigma and mu + 2 sigma, respectively. Using the optimized thresholds to segment tumor tissue, a quantitative characterization of disease is possible to calculate tumor volumes, disease severity, and temporal progression or treatment effect. Our proposed histogram-based method for characterizing spinal metastases shows great potential in extending the quantitative capacity of CT-based radiographic evaluations.