A community detection analysis of malocclusion classes from orthodontics and upper airway data

Abstract

Objective The interaction between skeletal class and upper airway has been extensively studied. Nevertheless, this relationship has not been clearly elucidated, with the heterogeneity of results suggesting the existence of different patterns for patients' classification, which has been elusive so far, probably due to oversimplified approaches. Hence, a network analysis was applied to test whether different patterns in patients' grouping exist. Settings and sample population Ninety young adult patients with no obvious signs of respiratory diseases and no previous adeno‐tonsillectomy procedures, with thirty patients characterized as Class I (0 < ANB < 4); 30 Class II (ANB > 4); and 30 as Class III (ANB < 0). Materials and methods A community detection approach was applied on a graph obtained from a previously analysed sample: thirty‐two measurements (nineteen cephalometric and thirteen upper airways data) were considered. Results An airway‐orthodontic complex network has been obtained by cross‐correlating patients. Before entering the correlation, data were controlled for age and gender using linear regression and standardized. By including or not the upper airway measurements as independent variables, two different community structures were obtained. Each contained five modules, though with different patients' assignments. Conclusion The community detection algorithm found the existence of more than the three classical skeletal classifications. These results support the development of alternative tools to classify subjects according to their craniofacial morphology. This approach could offer a powerful tool for implementing novel strategies for clinical and research in orthodontics.

Artificial intelligence and machine learning in orthodontics

“…29 Di Carlo et al applied a network analysis to verify the possible association between the orthodontic features and upper airway characteristics and found the most closely connected features. 30 The decision on whether or not to perform extraction is one of the most debated issues in clinical orthodontics. This decision is not made simply based on morphological dental and/or skeletal problems, rather it is influenced by various factors.…”

“…Auconi et al applied case‐based reasoning in predicting the progression of Class III malocclusion 29 . Di Carlo et al applied a network analysis to verify the possible association between the orthodontic features and upper airway characteristics and found the most closely connected features 30 …”

Network analysis of three-dimensional hard-soft tissue relationships in the lower 1/3 of the face: Skeletal Class I-normodivergent malocclusion versus Class II-hyperdivergent malocclusion

“…Network analysis, as an advanced data mining method based on graph theory, has been widely applied in epidemics [15], genomics and proteomics, such as tumor strati cation [16] and discovery of biomarker genes for diseases [17,18]. In clinical orthodontic studies, network analysis has also been proven to be a promising methodology [19,20,21,22]. The main idea is that a system of elements that interact with each other can be represented by a mathematical object called a 'network' (or a 'graph', Fig.…”

Complexity and data mining in dental research: A network medicine perspective on interceptive orthodontics

“…The networks are obtained by fixing a threshold value T: if the Pearson correlation between two vertices (features) is larger than T, they are considered connected (“linked”) 17 . Several topological network metrics are typically considered to convey biological meaning to cephalometric correlations (Figure 5A), as ‘betweenness’ 20 (a node centrality index) or modularity (the capacity of nodes to form communities characterized by similar properties) 77 . Once the graph is computed, the topological structure of the input is encoded in simpler structures (eg vectors of reals), and then the ML algorithms can be run.…”