Automated classification of skull deformity in 3D head CT images using regional shape information

We propose regional shape indices representing the degree of skull deformity for automatic classification and quantitative analysis of the skull deformities with sagittal and bicoronal synostosis on 3D head CT images. Regional shape indices are calculated by considering the degree of regional deformation of the deformed skull relative to the normal skull in each 3D bone segment separated by the sutures. In experimental results, the overall accuracy using multi-class support vector machine and random forest classifiers were estimated as 94.44% and 93.06%, respectively. our proposed regional shape indices can be used for early diagnosis and quantitative analysis of skull deformity with sagittal and bicoronal synostoses.