Artificial neural networks reconstruct missing perikymata in worn teeth.

Area of Science:

• Paleoanthropology
• Developmental Biology
• Bioarchaeology

Background:

• Dental evolutionary studies are crucial for understanding hominin growth and development.
• Tooth enamel's incremental growth records chronological information via perikymata, but early lines are lost to wear.
• Previous methods relied on polynomial regressions, with limitations in accuracy and applicability to worn teeth.

Purpose of the Study:

• To develop a novel method for reconstructing early, wear-lost perikymata in hominin teeth.
• To improve the accuracy of dental developmental timing and growth rate estimations.
• To enable the inclusion of worn teeth in dental evolutionary research.

Main Methods:

• Utilized an artificial neural network (ANN) approach on a modern human dataset.
• Reconstructed earliest perikymata from visible, later-developed perikymata.
• Applied the method to all tooth types (incisors, canines, premolars, molars).

Main Results:

• The ANN method accurately predicts perikymata counts within 2 units in the initial crown height deciles.
• This method enhances the reliability of estimating dental growth periodicity.
• Worn teeth, previously excluded, can now be incorporated into research.

Conclusions:

• The ANN method offers a significant advancement for reconstructing dental growth lines in fossil hominins.
• This tool expands the dataset potential for studies on dental development and evolutionary biology.
• The open-source R package 'teethR' facilitates broader application by researchers worldwide.