Cranial malformations classification caused by primary craniosynotosis using nonlinear kernels

Abstract

Single-suture craniosynostosis (SSC) is the pathologic condition ofpremature fusion of a calvarial suture. Premature fusion produces significant cranial deformities and is associated with an increased risk of cognitive deficits and neurobehavioral impairments. For these reasons, SSC represents an important area of research that requires effective methods for characterizing cranial morphology. In this paper we evaluate a new approach that combines the use of nonlinear kernels, co-occurrences of skull shape features, a new feature selection process and standard nonlinear dimensionality reduction techniques, as a means to classify cranial malformations due to SSC using computed tomography (CT) imaging. CT images were obtained from CT studies of 102 sagittal synostosis crania, 42 metopic synostosis crania, 12 unicoronal synostosis crania and 65 nonsynostotic skulls. We validate our approach with an extensive series of experiments and show that our proposed approach outperforms the classification performance of previously published techniques, achieving classification rates above 95%.