Abstract:
Dental caries, or tooth decay, is the most prevalent chronic condition and significantly contributes to the $165 billion dental expenditure in the United States annually [1-3]. Caries lesions develop due to an imbalance between demineralization from acid generated by a bacterial biofilm and remineralization from saliva. The earliest form of lesions has a characteristic structure with a highly mineralized surface zone (SZ) that prevents the remineralization of the less mineralized body of the lesion. Our long-term goal is to advance the prevention and treatment of early caries by elucidating the mechanism by which the SZ forms and how its structure and composition evolve. Given the importance of fluoride in caries prevention, we are particularly interested in its impact on these processes.
We have previously demonstrated that lesions with SZs analogous to those in humans form in the “gold standard” rodent caries model [4]. Herein, we report on our characterization of lesions and SZs formed in the presence of dietary fluoride in a second study. Our study included 8 animals per group with a two-factor design (dietary fluoride: low or high; cariogenic stimulus: none or high) and was evaluated at two time points (14d or 28d post-weaning) [5]. 64 synchrotron microcomputed tomography (SMCT) recontructions of M1 molars were collected at an isotropic voxel size of 1.73 μm. We will outline an automated workflow for unbiased and
reproducible segmentation utilizing a convolutional neural network (CNN), followed by the extraction of quantitative metrics that describe lesions and healthy enamel, and a comparison between experimental groups. As expected, lesions only form under high cariogenic stimuli. Further, our pipeline confirmed that fluoride in bulk reduced caries. We saw this through a significant reduction of lesion volumes and a significant increase in LAC (linear attenuation coefficient) values in the high dietary fluoride groups compared to the low. At the microscale, we determined differences in lesion structure through average mineral density profiles and surface zone metrics. We will discuss the impact of dietary fluoride in a multifaceted way. We believe that CNN-based segmentation and automated quantification are useful methods to advance our understanding of SZ evolution to allow for the advancement and analysis of new treatment options.