Quantitative Analysis and Classification of Skull Deformities on 3D Head CT Images Using Shape Descriptors

We propose a quantitative analysis of skull deformities with typical and mild cases of sagittal and bicoronal synostoses on 3D head computed tomography images using shape descriptors. First, cranial reference plane and two representative planes were generated. Second, five indices including cranial index (CI), cranial radius (CR) index, cranial partial slope (CPS) index, cranial extreme spot (CES) index and near cranial extreme spot (NCES) index were calculated using shape descriptors that reflect the characteristics of the deformed skull on the two representative planes. Finally, the skulls were classified using Support Vector Machine (SVM) classifier into three classes, sagittal synostosis, bicoronal synostosis and normal. In the experiments, classification performance was evaluated using 45 deformity subjects with sagittal and bicoronal synostoses and 45 normal subjects. The average accuracy levels of sagittal and bicoronal synostoses and normal were 91.69%, 98.89% and 90.06%, respectively. Our experimental results showed that the proposed method effectively classified the deformed skulls with sagittal and bicoronal synostoses and normal skulls. Our proposed shape descriptors can be used for early diagnosis, surgical planning of craniosynostosis as well as quantitative analysis of skull deformities.