A new system is presented that is capable of automatically estimating a person’s age and sex from dental X-rays
The new technology, based on Artificial Intelligence techniques such as deep learning, is the result of a PhD thesis carried out at CiTIUS (Singular Research Centre in Intelligent Technologies at the University of Santiago de Compostela - USC).
The second phase consisted in designing a fully automatic system to detect the mandibular contour in X‑rays and then characterise it in terms of shape and size. The results of this new method showed that lifetime changes in the shape of the jaw are useful for establishing a subject’s chronological age, while its size is essential for distinguishing between men and women. Professor Inmaculada Tomás points out that “the new system is compatible with any other dental practice that requires characterising the jaw, such as the quantification of mandibular asymmetry or the planning of orthognathic surgeries, among others.”
Finally, the research team developed a new methodology to detect the teeth present in the X‑ray, with the aim of using them in the estimation process. Researcher María José Carreira explains the approach: “after estimating the age and sex of each tooth separately, the resulting individual predictions are combined to obtain a single global estimate.” “Thanks to this integration we have a highly explainable system, since it is always possible to know which tooth can be trusted more and which less,” stresses the CiTIUS researcher. In addition, to further strengthen this interpretability, the system provides additional heat maps designed to assess the areas of each tooth that most influence the final prediction.
The results obtained throughout this thesis outperformed previous work by a wide margin, which confirms the usefulness of the developed methodology. The researchers highlight three key benefits of the new technologies: “they are highly accurate when determining age and sex, they dramatically reduce the time and subjectivity of manual estimates, and they are highly explainable, which facilitates their deployment in real‑world settings,” they conclude.
The second phase consisted in designing a fully automatic system to detect the mandibular contour in X‑rays and then characterise it in terms of shape and size. The results of this new method showed that lifetime changes in the shape of the jaw are useful for establishing a subject’s chronological age, while its size is essential for distinguishing between men and women. Professor Inmaculada Tomás points out that “the new system is compatible with any other dental practice that requires characterising the jaw, such as the quantification of mandibular asymmetry or the planning of orthognathic surgeries, among others.”
Finally, the research team developed a new methodology to detect the teeth present in the X‑ray, with the aim of using them in the estimation process. Researcher María José Carreira explains the approach: “after estimating the age and sex of each tooth separately, the resulting individual predictions are combined to obtain a single global estimate.” “Thanks to this integration we have a highly explainable system, since it is always possible to know which tooth can be trusted more and which less,” stresses the CiTIUS researcher. In addition, to further strengthen this interpretability, the system provides additional heat maps designed to assess the areas of each tooth that most influence the final prediction.
The results obtained throughout this thesis outperformed previous work by a wide margin, which confirms the usefulness of the developed methodology. The researchers highlight three key benefits of the new technologies: “they are highly accurate when determining age and sex, they dramatically reduce the time and subjectivity of manual estimates, and they are highly explainable, which facilitates their deployment in real‑world settings,” they conclude.