Differentiation of normal and neoplastic bone tissue in dynamic gadolinium-enhanced magnetic resonance imaging: validation of a semiautomated technique.

ABSTRACT

PURPOSE: This study was undertaken to clinically validate the accuracy of a semiautomated software tool for analysing the enhancement curve in focal malignant bone lesions. MATERIALS AND METHODS: Twenty-three patients affected by cancer with malignant focal bone lesions underwent dynamic gadolinium-enhanced magnetic resonance (MR) imaging using the following protocol T1-weighted turbo spin-echo sequences (time to repeat [TR] 600 ms, time to echo [TE] 8.6 ms, field of view [FOV] 40x40 cm) before and after intravenous injection of gadolinium-containing contrast agent. Image postprocessing was performed using the software DyCoH. Each region of interest (5x5 pixels), drawn to include the area of the lesion with the highest values of the area under the curve map, was analysed to obtain time-intensity curves and relative perfusion parameters time to peak (TTP), peak intensity (PI), slope (60-s slope), intensity at 60 s after contrast agent injection (60-s I) and final intensity (FI). RESULTS: Data were obtained by analysing 86 malignant lesions and 86 apparently normal bone regions. PI, 60-s slope, 60-s I and FI were significantly different between neoplastic and apparently normal (p<0.001) samples. Sensitivity, specificity and accuracy were, respectively, 94%, 93% and 94% at a PI threshold of 100 (signal-to-noise ratio), with positive and negative predictive values of 93% and 94%. At a threshold value of 0.85 for 60-s slope, sensitivity and specificity values were both 91%. CONCLUSIONS: The semiautomated technique we report appears to be accurate for identifying neoplastic tissue and for mapping perfusion parameters, with the added value of a consistent measurement of perfusion parameters on colour-coded maps.