Recognition of vertebral compression fractures in magnetic resonance images using statistics of height and width

Vertebral compression fractures (VCFs) present as partial collapses of vertebral bodies and may occur secondary to osteoporosis bone fragility and to metastatic cancer infiltration. The correct diagnosis of nontraumatic VCFs is therefore, fundamental for correct treatment. We aimed to classify VCFs using T1-weighted magnetic resonance images (MRI) of the lumbar spine acquired in the sagittal plane. Our study group comprised 63 patients (38 women and 25 men). From these patients 102 lumbar VCFs (53 benign and 49 malignant) and 89 normal vertebral bodies were manually segmented. The principal axis of each vertebral body region of interest was identified using moments. Statistical features of height and width measured perpendicular and parallel to the principal axis were computed. The k-nearest-neighbor method, a neural network with radial basis functions, and the naïve Bayes classifier were used with feature selection for classification. Areas under the receiver operating characteristic curve of 0.96 in the recognition of VCFs as compared with normal vertebral bodies and 0.73 for the classification of benign versus malignant VCFs were obtained. The proposed methods are promising for the recognition of VCFs, but additional features are needed to improve the classification of benign versus malignant VCFs.