Individuals whose teething period deviates from the expected
timeframe or whose biological and physiological ages diverge from their
chronological age pose challenges for dental care planning. This
discrepancy complicates the provision of effective dental healthcare,
underscoring the importance of early diagnosis. Early identification not only
addresses dental health issues but also enables timely intervention in other
disorders causing misalignment between tooth age and chronological age.
Additionally, dental age and gender estimation find applications in
anthropology and forensic science. Estimating age and gender from
panoramic radiographs typically demands expert evaluation, relying on
personal skills and expertise, which can lead to potential errors. Presently,
artificial intelligence systems have been introduced to alleviate challenges
associated with expert evaluation. However, a lingering concern is the
trustworthiness of the model's decisions.
We also demonstrate that the DeepToothDuo model produces
improved interpretability results using SHAP. The researchers trained the
language model for application in developing web applications aimed at
answering users' dental questions. Our proposed model achieves a gender
prediction accuracy of 87.38% and exhibits an age estimation error of 1.96 years.
Model interpretability studies reveal that the model considers anatomical
features consistent with existing studies on dental and human anatomy.
