< Back to previous page
Project
Dental age estimation in sub-adults: striving for an optimal approach.
Increasing global human migration, raises management concerns in the countries where immigrants seek shelter. A special protective status must be given to immigrating unaccompanied children. Therefore, most nationallaws enforce specialized medical investigations to get proof of the ageof unaccompanied youngsters with no, or lacking official identificationdocuments and claiming to be minors. Dental age estimation in this particular age group relies on the only dental age predictor(s) available, namely the developing third molar(s). Hence, scientific correct dental age estimations in sub-adults, especially when originating from distant countries and diverse biological origin are requested. The general research aim was to optimize dental age estimation based on third molar development.
Panoramic radiographs were retrospectively and cross-sectional sampled to collect data registering third molar development. For that registration, two techniques were described. The sequence of third molar development was divided in succeeding stages, and the observed third molar development was classified in the corresponding stage. Otherwise, the dimensions of third molars increase during its maturation and measures of the observed third molar sizes were registered. In the current thesis,both registration techniques were compared. Third molar stages (categorical data) were best related to age and provided the most accurate age predictions compared to all collected tooth measurements and ratios of tooth measurements (continuous data). Combining the scored third molar stages with tooth measurements or ratios did not contribute to a clinical relevant information gain for age prediction.
Multiple tooth development staging techniques were reported, based on the described and considered borderlines between succeeding stages the quantity of stages covering the third molar development process differs between techniques. Therefore, it was studied if the number of stages used in a staging technique is influencing the age prediction performances. The number of stages utilized in the third molar registration technique slightly influencedthe age predictions. The choice of third molar development registrationtechnique has to depend on its stages described for the developmental period of interest and should not compromise the feasibility of correctlyregistering all these stages.
The classical approach for age estimation uses regression analysis to model the collected reference data. Drawbacks of this technique concern the age distribution of the residuals, the high correlation between the independent variables, often observed missing values of the independent variables, and a systematic bias in the age predictions. Therefore, a Bayesian approach of age estimation on third molar development was established in the current study. The age prediction performances of both approaches were compared. Both models provided similar accuracy, precision and coverage in age estimation outcome. The Bayesian approach reduced the bias that is typically present inthe regression models. The age of juveniles was less overestimated, yielding a better discrimination between subjects older or younger than 18 years. Moreover, the Bayesian model integrated all available third molar information.
Sub-adult age estimations are mostly requested to discriminate a child from an adult during migration and asylum procedures. Due to themigration aspect, frequently the age of an applicant with a particular geographical and biologic origin was estimated using methods or models developed on a reference sample, including subjects with unlike origin. It was investigated whether differences in third molar development between populations with different geographic and biological origin exist. Therefore, third molar development was analysed and compared on 13 country-specific samples using a factor analysis. Differences in third molar development between countries exist, but they were not constant over age and varied in an unordered way. Because the magnitude of the differences turned out to be small there was no evidence for important differences in third molar development between the countries.
Age estimation models developed on a particular country-specificreference sample were validated on their age prediction performances using a validation sample from a different geographic and biological origin as the reference sample. Validated on 13 country-specific databases using information from Belgium, or all countries pooled together changes the difference between observed and predicted age obtained on country-specific information only slightly. For the adult-juvenile discrimination, the Belgium reference model provided a maximal advantage of the doubt toinvestigated unaccompanied minor fugitives. The reference model based on all pooled countries, substituted the country-specific reference modelmost accurately in sub-adults.
The age prediction performances of age estimation models constructed on a single age-related variable are possibly ameliorated, adding age-related information of one or more variables present in the considered period of life. Therefore, reference samples registering at a specific moment third molar development, as well as tooth morphological or skeletal age-related variables, were collected, modelled and validated. Due to the inherent image quality of panoramic radiographs tooth morphological measurements based on secondary dentine apposition, could only be achieved on a restricted sample. Clinically the gain in age prediction accuracy was negligible when adding the time consuming additional tooth morphological measurementsto the staged third molar development. On cephalometric radiographs theskeletal age predicting variable(s) and related registration systems providing the most information on age were cervical vertebrae scoring systems. Combining the information from cervical vertebrae and third molars improved the age predictions drastically in the period of early third molar development. In sub-adults no, or a negligible, gain in accuracy of age prediction was obtained. In an optimal approach, dental age estimations in sub-adults are based on the radiologicallyobserved developmental status or the absence, of each third molar. The observed information is registered according to a staging technique. Thecollected reference information is modeled as dependent ordinal data ina multivariate Bayesian approach. Despite detected country-specific differences in third molar development, the clinical impact on age estimation is minimal. Based on an ongoing country-specific data collection, a quantification of the maximal difference in prediction error using not county-specific reference information is established. Further on, the Belgium reference information classifies more juveniles compared to country-specific information and is recommended in lack of a country-specific reference model for age estimations of young unaccompanied fugitives.
Panoramic radiographs were retrospectively and cross-sectional sampled to collect data registering third molar development. For that registration, two techniques were described. The sequence of third molar development was divided in succeeding stages, and the observed third molar development was classified in the corresponding stage. Otherwise, the dimensions of third molars increase during its maturation and measures of the observed third molar sizes were registered. In the current thesis,both registration techniques were compared. Third molar stages (categorical data) were best related to age and provided the most accurate age predictions compared to all collected tooth measurements and ratios of tooth measurements (continuous data). Combining the scored third molar stages with tooth measurements or ratios did not contribute to a clinical relevant information gain for age prediction.
Multiple tooth development staging techniques were reported, based on the described and considered borderlines between succeeding stages the quantity of stages covering the third molar development process differs between techniques. Therefore, it was studied if the number of stages used in a staging technique is influencing the age prediction performances. The number of stages utilized in the third molar registration technique slightly influencedthe age predictions. The choice of third molar development registrationtechnique has to depend on its stages described for the developmental period of interest and should not compromise the feasibility of correctlyregistering all these stages.
The classical approach for age estimation uses regression analysis to model the collected reference data. Drawbacks of this technique concern the age distribution of the residuals, the high correlation between the independent variables, often observed missing values of the independent variables, and a systematic bias in the age predictions. Therefore, a Bayesian approach of age estimation on third molar development was established in the current study. The age prediction performances of both approaches were compared. Both models provided similar accuracy, precision and coverage in age estimation outcome. The Bayesian approach reduced the bias that is typically present inthe regression models. The age of juveniles was less overestimated, yielding a better discrimination between subjects older or younger than 18 years. Moreover, the Bayesian model integrated all available third molar information.
Sub-adult age estimations are mostly requested to discriminate a child from an adult during migration and asylum procedures. Due to themigration aspect, frequently the age of an applicant with a particular geographical and biologic origin was estimated using methods or models developed on a reference sample, including subjects with unlike origin. It was investigated whether differences in third molar development between populations with different geographic and biological origin exist. Therefore, third molar development was analysed and compared on 13 country-specific samples using a factor analysis. Differences in third molar development between countries exist, but they were not constant over age and varied in an unordered way. Because the magnitude of the differences turned out to be small there was no evidence for important differences in third molar development between the countries.
Age estimation models developed on a particular country-specificreference sample were validated on their age prediction performances using a validation sample from a different geographic and biological origin as the reference sample. Validated on 13 country-specific databases using information from Belgium, or all countries pooled together changes the difference between observed and predicted age obtained on country-specific information only slightly. For the adult-juvenile discrimination, the Belgium reference model provided a maximal advantage of the doubt toinvestigated unaccompanied minor fugitives. The reference model based on all pooled countries, substituted the country-specific reference modelmost accurately in sub-adults.
The age prediction performances of age estimation models constructed on a single age-related variable are possibly ameliorated, adding age-related information of one or more variables present in the considered period of life. Therefore, reference samples registering at a specific moment third molar development, as well as tooth morphological or skeletal age-related variables, were collected, modelled and validated. Due to the inherent image quality of panoramic radiographs tooth morphological measurements based on secondary dentine apposition, could only be achieved on a restricted sample. Clinically the gain in age prediction accuracy was negligible when adding the time consuming additional tooth morphological measurementsto the staged third molar development. On cephalometric radiographs theskeletal age predicting variable(s) and related registration systems providing the most information on age were cervical vertebrae scoring systems. Combining the information from cervical vertebrae and third molars improved the age predictions drastically in the period of early third molar development. In sub-adults no, or a negligible, gain in accuracy of age prediction was obtained. In an optimal approach, dental age estimations in sub-adults are based on the radiologicallyobserved developmental status or the absence, of each third molar. The observed information is registered according to a staging technique. Thecollected reference information is modeled as dependent ordinal data ina multivariate Bayesian approach. Despite detected country-specific differences in third molar development, the clinical impact on age estimation is minimal. Based on an ongoing country-specific data collection, a quantification of the maximal difference in prediction error using not county-specific reference information is established. Further on, the Belgium reference information classifies more juveniles compared to country-specific information and is recommended in lack of a country-specific reference model for age estimations of young unaccompanied fugitives.
Date:1 Feb 2009 → 2 Jul 2013
Keywords:forensic
Disciplines:Other biological sciences
Project type:PhD project