Automatic Teeth Recognition in Multi-Slice CT Images

摘要

Teeth are unique in identification of deceased persons where other biometric features are not applicable. In many cases, e.g., fire victims, the conventional biometric features such as face, fingerprint, iris, etc. may not be available. In this research, we present a technique for automatic recognition of teeth in multi-slices CT images. Teeth classification is performed by proposing appropriate segmentation, feature extraction and classification techniques. We segment a tooth by a hybrid approach including anatomical based histogram thresholding, panaromic resampling, and Level-Set techniques. We perform feature extraction by employing the following techniques; (1) calculating the Eigen-Values of each tooth by Dirichlet Laplacian technique; (2) providing the spectral features by Fourier and wavelet descriptors; and (3) determining statistical moments utilizing the teeth intensity range. Experimental results reveal that the technique is successful to automatically classify teeth in more than 93% of the cases in the lower and upper jaws. Our method is independent of anatomical information such as the sequence and locality of the teeth in jaws.