Bayesian inversion method for 3D dental X-ray imaging

Diagnostic and operational tasks in dentistry require three-dimensional (3D) information about tissue. A novel type of low dose dental 3D X-ray imaging is considered. Given projection images taken from a few sparsely distributed directions using the dentist’s regular X-ray equipment, the 3D X-ray attenuation function is reconstructed. This is an ill-posed inverse problem, and Bayesian inversion is a well suited framework for reconstruction from such incomplete data. The reconstruction problem is formulated in a well-posed probabilistic form in which a priori information is used to compensate for the incomplete data. A parallelized Bayesian method (implemented for a Beowulf cluster computer) for 3D reconstruction in dental radiology is presented (the method was originally presented in (Kolehmainen et al., 2006)). The prior model for dental structures consist of a weighted ` and total variation (TV)-prior together with the positivity prior. The inverse problem is stated as finding the maximum a posterior (MAP) estimate. The method is tested with in vivo patient data and shown to outperform the reference method (tomosynthesis).