Diagnostic performance of tomosynthesis for evaluation of bone tumors and tumor-like lesions: a comparison with radiography.

Abstract

Even though radiologic diagnosis of bone tumors and tumor-like lesions is usually based on radiographs, radiographically faint imaging features sometimes remain challenging due to overlapping anatomical structures. To compare tomosynthesis with radiography for the evaluation of bone tumors and tumor-like lesions. Forty-seven bone tumors and tumor-like lesions were assessed with radiographs and tomosynthesis images. Two radiologists independently analyzed imaging features of lesions, including margin, periosteal reaction, cortical thinning, matrix mineralization, cortical destruction (such as pathologic fracture), and extraosseous soft-tissue extension. Computed tomography (CT) imaging was used as a reference method. Diagnostic performances of radiography and tomosynthesis were analyzed and compared based on sensitivity, specificity, accuracy, positive predictive value, and negative predictive value. Effective radiation dose was compared among the three imaging modalities by phantom studies. Inter-observer variability (kappa value) for imaging features was slight to moderate on radiography (0.167-0.588), whereas it was nearly perfect on tomosynthesis (0.898-1.000) except for extraosseous soft-tissue extension (0.647 vs. 0.647). Tomosynthesis showed significantly higher sensitivity than radiography in evaluating the margin for bone tumors or tumor-like lesions (1.00 vs. 0.85; P = 0.016), and significantly higher accuracy than radiography in evaluating the margin and matrix mineralization for those (1.00 vs. 0.85; P = 0.016 and 0.91 vs.0.77; P = 0.023, respectively). In phantom studies, mean effective radiation doses were highest in order of CT, tomography, and radiography. Tomosynthesis increases sensitivity and accuracy of the margin as well as accuracy of the matrix mineralization of bone tumors and tumor-like lesions compared to radiography.