Publication Details

Proceedings SPS-DARTS 2006, The second annual IEEE Benelux/DSP Valley Signal Processing Symposium

Contribution To Book Anthology


Despite the rapid proliferation of digital subtraction radiography (DSR) procedures in general medicine, dental caries diagnosis seems to be largely deprived from these recent developments. This is all the more remarkable given that the clinical assessment of incipient caries could greatly benefit from the application of DSR with its much improved performance for revealing subtle changes in mineralization. The apparent reluctance in embracing DSR may to a large extent be attributed to the particularly challenging problem of securing a constant exposure geometry between successive intraoral radiographic exposures, combined with the sharp density contrasts at the caries initiation sites, which makes that even the slightest geometric misalignment may cause undesirable artifacts that can mimic or even obfuscate true changes in radiolucency. Another more fundamental obstacle, however, resides in the very nature of the currently used methods for retrospective geometric standardization of intraoral radiographs acquired with a nonconsistent exposure geometry. Largely depending on the manual assignment of corresponding features between pairs of images, these methods require considerable precision and dexterity on the part of the user. It is rather doubtful that procedures depending on such an extensive user involvement will ever find widespread acceptance as a clinical tool for the diagnosis of caries. In this presentation, we disclose results from an ongoing study, in which the prospects for a fully automated application of DSR in dental caries diagnosis are explored. The success of such concept crucially relies on the seamless integration of direct digital image acquisition with an automatic retrospective geometric standardization of radiographs. In our study, we specifically compared a more conventional pyramidal edge based registration method, with a more advanced approach based on the concept of Mutual Information (MI). The performance of the MI-based method was found to degrade less rapidly under increasing angular disparity.