Differentiation of giant cell tumours of bone, primary aneurysmal bone cysts, and aneurysmal bone cysts secondary to giant cell tumour of bone: using whole-tumour CT texture analysis parameters as quantitative biomarkers

Abstract

Aim: To determine whether computed tomography (CT) texture analysis parameters can be used as quantitative biomarkers to help differentiate giant cell tumour of bones (GCTs), primary aneurysmal bone cysts (PABCs), and aneurysmal bone cysts (ABCs) secondary to giant cell tumours of bone (GABCs).

Materials and methods: One hundred and seven patients with 63 GCTs, 31 PABCs, and 13 GABCs were analysed retrospectively. All patients underwent preoperative CT. Two radiologists independently evaluated the qualitative features of the CT images and extracted texture parameters. Patient demographics, qualitative features, and texture parameters among GCTs, PABCs, and GABCs were compared statistically. Differences in these parameters between ABCs and GCTs were also assessed. ROC curves were obtained to determine optimal parameter values.

Results: The best preoperative CT parameters to differentiate GCTs, PABCs, and GABCs included one qualitative feature (location around the knee) and four texture parameters (95^th percentile, maximum intensity, skewness, and kurtosis). Age and three texture parameters (5^th percentile, inhomogeneity, and kurtosis) enabled statistically significant differentiation between GCTs and ABCs. Combination of the above four parameters generated the largest area under the ROC curve (AUC) for the differentiation of GCTs and ABCs.

Conclusion: CT texture analysis parameters can be used as quantitative biomarkers for preoperative differentiation among GCTs, PABCs, and GABCs.