Department of Materials Science and Engineering, University of Washington, USA.
Department of Materials Science and Engineering, University of Washington, USA; Department of Restorative Dentistry, School of Dentistry, University of Washington, USA; Department of Oral Health Sciences, School of Dentistry, University of Washington, USA.
J Mech Behav Biomed Mater. 2022 May;129:105147. doi: 10.1016/j.jmbbm.2022.105147. Epub 2022 Mar 2.
Understanding aging of tooth tissues is the first step to developing robust treatments that support lifelong oral health. In this study selected nanomechanical, compositional and structural parameters of human enamel were characterized to assess the effects of aging on its durability in terms of the apparent fracture toughness (K) and brittleness (B). The interdependencies between aging and the enamel properties were assessed using a combination of traditional Pearson's correlation coefficient matrices and self-organizing maps (SOMs) via unsupervised machine learning. To consider age effects, the enamel of three age groups of donor teeth was studied, including primary (donor age ≤10), young (20 age ≤ age ≤50), and old (55 ≤ age) and differences in properties and correlations were identified. Results showed that K was negatively correlated to the E, H, degree of crystallinity, and fluoridation, but positively correlated with carbonate content; the opposite trends were observed in B. Interestingly, the SOMs showed that the outer enamel of the old group underwent a degradation in durability (decrease in K and increase in B) that was related to multiple contributions, whereas the inner enamel did not undergo this change. Application of K-means clustering on the trained SOMs offered novel insights into the contributions of enamel durability with aging, unique visualization of high-dimensional data onto 2D plots and identified new research directions that would not have otherwise been discovered. Overall, the findings demonstrate the opportunities for understanding aging of enamel using machine learning techniques to pursue age-targeted oral health care.
了解牙齿组织的衰老过程是开发支持终生口腔健康的强大治疗方法的第一步。在这项研究中，选择了人类牙釉质的纳米力学、组成和结构参数进行了表征，以评估衰老对其耐用性的影响，其耐用性体现在表观断裂韧性（K）和脆性（B）上。使用传统的皮尔逊相关系数矩阵和无监督机器学习的自组织图（SOM）的组合，评估了衰老与牙釉质特性之间的相互依赖关系。为了考虑年龄的影响，研究了来自三个供体牙齿年龄组的牙釉质，包括乳牙（供体年龄≤10 岁）、年轻牙（20 岁≤年龄≤50 岁）和老年牙（55 岁≤年龄），并确定了特性和相关性的差异。结果表明，K 与 E、H、结晶度和氟化程度呈负相关，但与碳酸盐含量呈正相关；B 则呈现相反的趋势。有趣的是，SOM 表明，老年组的牙釉质外层的耐用性（K 值降低和 B 值增加）发生了退化，这与多种因素有关，而内层牙釉质没有发生这种变化。在经过训练的 SOM 上应用 K-均值聚类提供了对牙釉质随年龄变化的耐用性的新见解，将高维数据独特地可视化到 2D 图上，并确定了原本无法发现的新的研究方向。总的来说，这些发现表明，使用机器学习技术有机会了解牙釉质的衰老过程，以追求针对年龄的口腔保健。
