Relaxed Statistical Shape Models for 3D Image Segmentation – Application to Mandible Bone in Cone-beam CT Data

Patient specific surface models of the jaw are beneficial for pre-operative planning and manufacturing of customized prosthesis. Such models can be generated on the basis of dental cone-beam CT images, but those suffer from a comparatively bad image quality with regard to the signal-to-noise ratio. Therefore, in this work, a statistical shape model (SSM) is used for robust segmentation of the mandible bone. While previous works with that application require manual interaction during SSM construction, we establish correspondence fully automatic by minimizing the description length of the model. Subsequently, the mandible bone is automatically localized and segmented using the SSM as shape constraint. The standard SSM constraint is known to be inherently limited insofar as patient specific anatomical details can often not be represented. To overcome this limitation, a new, mathematically sound, computationally fast, and intuitively interpretable, relaxed SSM constraint is derived, which can be applied without any user-provided parameter. Evaluation on clinical cone beam CT images yields an improvement of the Jaccard coefficient up to 45% compared to the standard SSM constraint. Our results are similar to that of alternative methods in the literature, indicating the general potential of the proposed relaxed SSM constraint for medical image segmentation.