Towards deep learning reliable gender estimation from dental panoramic radiographs

Abstract

Gender estimation is a crucial task in many forensic procedures. Currently, it is carried out by taking several measurements in the body and compare them to population-specific reference tables. Among others, the oral cavity is widely used because it contains highly-dimorphic bones, such as the mandible or the teeth. However, these features provide less accurate gender estimations in subjects younger than 20. In this work, three fully automatic approaches based on deep learning architectures have been compared in a database of 3400 dental panoramic images. The results are in line with the literature in patients older than 20, yielding accuracies between 90 and 96%. Although the performance decreases in younger people, the proposed methodology provides accuracies over 84% in subjects older than 10, demonstrating the usefulness of automatic approaches in sex prediction.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 8 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.