Identification and quantification of caries from CBCT segmented images

Abstract

Interproximal caries is a bacterial infection that occurs in the oral cavity, causing structural lesions between teeth. Diagnosis typically involves using radiographic techniques to capture images, but the use of Cone Beam Computed Tomography (CBCT) is still under-explored. This study explores CBCT, which acquires three-dimensional radiographic images, and employs two different image acquisition protocols to identify potential lesions. We developed a set of image processing techniques to segment three dental structures and accurately identify interproximal caries. Our results using classical metrics indicate an AUC of 0.928, a sensitivity of 87.33%, a precision of 88.50%, and a Jaccard Index of 0.7037. Our method effectively identifies lesions in dental structures, with the potential for practical assistance in diagnosing this disease.