Prognostic approach to Class III malocclusion through case-based reasoning.

OBJECTIVE: This investigation evaluates the evidence of case-based reasoning (CBR) in providing additional information on the prediction of future Class III craniofacial growth. SETTINGS AND SAMPLE POPULATION: The craniofacial characteristics of 104 untreated Class III subjects (7-17 years of age), monitored with two lateral cephalograms obtained during the growth process, were evaluated. MATERIALS AND METHODS: Data were compared with the skeletal characteristics of subjects who showed a high degree of skeletal imbalance ('prototypes') obtained from a large data set of 1263 Class III cross-sectional subjects (7-17 years of age). RESULTS: The degree of similarity of longitudinal subjects with the most unbalanced prototypes allowed the identification of subjects who would develop a subsequent unfavourable skeletal growth (accuracy: 81%). The angle between the palatal plane and the sella-nasion line (PP-SN angle) and the Wits appraisal were two additional craniofacial features involved in the early prediction of the adverse progression of the Class III skeletal imbalance. CONCLUSIONS: Case-based reasoning methodology, which uses a personalized inference method, may bring additional information to approximate the skeletal progression of Class III malocclusion.

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

Procedures and models of computerized data analysis are becoming researchers' and practitioners' thinking partners by transforming the reasoning underlying biomedicine. Complexity theory, Network analysis and Artificial Intelligence are already approaching this discipline, intending to provide support for patient's diagnosis, prognosis and treatments. At the same time, due to the sparsity, noisiness and time-dependency of medical data, such procedures are raising many unprecedented problems related to the mismatch between the human mind's reasoning and the outputs of computational models. Thanks to these computational, non-anthropocentric models, a patient's clinical situation can be elucidated in the orthodontic discipline, and the growth outcome can be approximated. However, to have confidence in these procedures, orthodontists should be warned of the related benefits and risks. Here we want to present how these innovative approaches can derive better patients' characterization, also offering a different point of view about patient's classification, prognosis and treatment.

Computer-aided heuristics in orthodontics.

INTRODUCTION: During the decision-making process, physicians rely on heuristics that consist of simple, useful procedures for solving problems, intuitive shortcuts that produce reliable decisions based on limited information. In clinical situations characterized by a high degree of uncertainty such as those encountered in orthodontics, cognitive biases and judgment errors related to heuristics are not uncommon. This study aimed at promoting trust in the effective interface between the intuitive reasoning of the orthodontic practitioner and the computational heuristics emerging from simple statistical models. METHODS: We propose an integrative model based on the interaction between clinical reasoning and 2 computational tools, cluster analysis and fast-and-frugal trees, to extract a structured craniofacial representation of untreated subjects with Class III malocclusion and to forecast the worsening of the malocclusion over time. RESULTS: Cluster analysis of cephalometric values from 144 growing subjects with Class III malocclusion followed longitudinally (T1: mean age, 10.2 ± 1.9 years; T2: mean age, 13.8 ± 2.7 years) produced 3 morphologic subgroups with predominant sagittal, vertical, and slight maxillomandibular imbalances. Fast-and-frugal trees applied to different subgroups extracted heuristics that improved the prediction of key features associated with adverse craniofacial growth. CONCLUSIONS: Provided that cephalometric values are placed in the appropriate framework, the matching between simple and fast computational approaches and clinical reasoning could help the intuitive logic, perception, and cognitive inferences of orthodontic practitioners on the outcome of patients affected by Class III disharmony, decreasing errors associated with flawed judgments and improving the accuracy of decision making.

Understanding interactions among cephalometrics variables during growth in untreated Class III subjects.

OBJECTIVE: The aim of the present study was to apply a computational method commonly used in data mining discipline, classification trees (CTs), to evaluate the growth features in untreated Class III subjects. MATERIALS AND METHODS: CT was applied to data from 91 untreated Class III subjects (48 females and 43 males) and compared with the results of discriminant analysis (DA). For all subjects, lateral cephalograms were available at T1 (mean age 10.4 ± 2.0 years) and at T2 (mean age 15.4 ± 1.9 years). A cephalometric analysis comprising 11 variables was performed. The subjects were divided into two subgroups, unfavourable ('Bad') and favourable ('Good') growers, according to the quality of the skeletal growth rate in comparison with the normal craniofacial growth. RESULTS: CTs showed that the most informative attribute for the prediction of favourable/unfavourable skeletal growth was the SNA angle. Subjects with SNA values lower than 79.1 degrees showed a risk of 94 per cent of growing unfavourably. DA was able to select palatal plane to mandibular plane angle as predictors. DA, however, showed a statistically significant higher rate of misclassification when compared with CTs (40.7 per cent versus 12.1 per cent, binomial exact test: odds ratio = 6.20; P < 0.0001). CONCLUSIONS: CTs provided a valid measure of elucidating the effective contribution of craniofacial characteristics in predicting favourable/unfavourable growth in untreated Class III subjects.

Prediction of Class III treatment outcomes through orthodontic data mining.

OBJECTIVE: To determine whether it is possible to predict Class III treatment outcomes on the basis of a model derived from a combination of computational analyses derived from complexity science, such as fuzzy clustering repartition and network analysis. METHODS: Cephalometric data of 54 Class III patients (32 females, 22 males) taken before (T1, mean age 8.2 ± 1.6 years) and after (T2, mean age 14.6 ± 1.8 years) early rapid maxillary expansion and facemask therapy followed by fixed appliances were analysed. Patients were classified at T1 on the basis of high membership grade into three main dentoskeletal fuzzy cluster phenotypes: hyperdivergent (HD), hypermandibular (HM), and balanced (Bal) phenotypes. The prevalence rate of successful and unsuccessful cases at T2 was calculated for the three clusters and compared by means of Fisher's exact test corrected for multiple testing (Holm-Bonferroni method). RESULTS: Unsuccessful cases were 9 out of 54 patients (16.7%). Once patients were framed into their cluster membership, the individualized pre-treatment prediction of unsuccessful cases was largely differentiated: HD and HM patients showed a significantly greater prevalence rate of unsuccessful cases than Bal patients (0% in Bal cluster, 28.6% in HM cluster, and 33.3% in HD cluster). Network analysis captured some noticeable interdependencies of Class III patients, showing a more connected interactive structure of cephalometric data sets in HM and HD patients compared with Bal patients. The results were confirmed after minimizing the geometrical connections between cephalometric variables in the model. CONCLUSIONS: Fuzzy clustering repartition can be usefully used to estimate an individualized risk of unsuccessful treatment outcome in Class III patients.

Forecasting craniofacial growth in individuals with class III malocclusion by computational modelling.

AIM: To develop a mathematical model that adequately represented the pattern of craniofacial growth in class III subject consistently, with the goal of using this information to make growth predictions that could be amenable to longitudinal verification and clinical use. MATERIALS AND METHODS: A combination of computational techniques (i.e. Fuzzy clustering and Network analysis) was applied to cephalometric data derived from 429 untreated growing female patients with class III malocclusion to visualize craniofacial growth dynamics and correlations. Four age groups of subjects were examined individually: from 7 to 9 years of age, from 10 to 12 years, from 13 to 14 years, and from 15 to 17 years. RESULTS: The connections between pathway components of class III craniofacial growth can be visualized from Network profiles. Fuzzy clustering analysis was able to define further growth patterns and coherences of the traditionally reported dentoskeletal characteristics of this structural imbalance. Craniofacial growth can be visualized as a biological, space-constraint-based optimization process; the prediction of individual growth trajectories depends on the rate of membership to a specific 'winner' cluster, i.e. on a specific individual growth strategy. The reliability of the information thus gained was tested to forecast craniofacial growth of 28 untreated female class III subjects followed longitudinally. CONCLUSION: The combination of Fuzzy clustering and Network algorithms allowed the development of principles for combining multiple auxological cephalometric features into a joint global model and to predict the individual risk of the facial pattern imbalance during growth.

The Validity of Machine Learning Procedures in Orthodontics: What Is Still Missing?

Artificial intelligence (AI) models and procedures hold remarkable predictive efficiency in the medical domain through their ability to discover hidden, non-obvious clinical patterns in data. However, due to the sparsity, noise, and time-dependency of medical data, AI procedures are raising unprecedented issues related to the mismatch between doctors' mentalreasoning and the statistical answers provided by algorithms. Electronic systems can reproduce or even amplify noise hidden in the data, especially when the diagnosis of the subjects in the training data set is inaccurate or incomplete. In this paper we describe the conditions that need to be met for AI instruments to be truly useful in the orthodontic domain. We report some examples of computational procedures that are capable of extracting orthodontic knowledge through ever deeper patient representation. To have confidence in these procedures, orthodontic practitioners should recognize the benefits, shortcomings, and unintended consequences of AI models, as algorithms that learn from human decisions likewise learn mistakes and biases.

Exploiting the interplay between cross-sectional and longitudinal data in Class III malocclusion patients.

The aim of the study was to investigate how to improve the forecasting of craniofacial unbalance risk during growth among patients affected by Class III malocclusion. To this purpose we used computational methodologies such as Transductive Learning (TL), Boosting (B), and Feature Engineering (FE) instead of the traditional statistical analysis based on Classification trees and logistic models. Such techniques have been applied to cephalometric data from 728 cross-sectional untreated Class III subjects (6-14 years of age) and from 91 untreated Class III subjects followed longitudinally during the growth process. A cephalometric analysis comprising 11 variables has also been performed. The subjects followed longitudinally were divided into two subgroups: favourable and unfavourable growth, in comparison with normal craniofacial growth. With respect to traditional statistical predictive analytics, TL increased the accuracy in identifying subjects at risk of unfavourable growth. TL algorithm was useful in diffusion of information from longitudinal to cross-sectional subjects. The accuracy in identifying high-risk subjects to growth worsening increased from 63% to 78%. Finally, a further increase in identification accuracy, up to 83%, was produced by FE. A ranking of important variables in identifying subjects at risk of growth worsening, therefore, has been obtained.

Bayesian Networks Analysis of Malocclusion Data.

In this paper we use Bayesian networks to determine and visualise the interactions among various Class III malocclusion maxillofacial features during growth and treatment. We start from a sample of 143 patients characterised through a series of a maximum of 21 different craniofacial features. We estimate a network model from these data and we test its consistency by verifying some commonly accepted hypotheses on the evolution of these disharmonies by means of Bayesian statistics. We show that untreated subjects develop different Class III craniofacial growth patterns as compared to patients submitted to orthodontic treatment with rapid maxillary expansion and facemask therapy. Among treated patients the CoA segment (the maxillary length) and the ANB angle (the antero-posterior relation of the maxilla to the mandible) seem to be the skeletal subspaces that receive the main effect of the treatment.

Using networks to understand medical data: the case of Class III malocclusions.

A system of elements that interact or regulate each other can be represented by a mathematical object called a network. While network analysis has been successfully applied to high-throughput biological systems, less has been done regarding their application in more applied fields of medicine; here we show an application based on standard medical diagnostic data. We apply network analysis to Class III malocclusion, one of the most difficult to understand and treat orofacial anomaly. We hypothesize that different interactions of the skeletal components can contribute to pathological disequilibrium; in order to test this hypothesis, we apply network analysis to 532 Class III young female patients. The topology of the Class III malocclusion obtained by network analysis shows a strong co-occurrence of abnormal skeletal features. The pattern of these occurrences influences the vertical and horizontal balance of disharmony in skeletal form and position. Patients with more unbalanced orthodontic phenotypes show preponderance of the pathological skeletal nodes and minor relevance of adaptive dentoalveolar equilibrating nodes. Furthermore, by applying Power Graphs analysis we identify some functional modules among orthodontic nodes. These modules correspond to groups of tightly inter-related features and presumably constitute the key regulators of plasticity and the sites of unbalance of the growing dentofacial Class III system. The data of the present study show that, in their most basic abstraction level, the orofacial characteristics can be represented as graphs using nodes to represent orthodontic characteristics, and edges to represent their various types of interactions. The applications of this mathematical model could improve the interpretation of the quantitative, patient-specific information, and help to better targeting therapy. Last but not least, the methodology we have applied in analyzing orthodontic features can be applied easily to other fields of the medical science.