Related Experiment Video
Updated: Jan 8, 2026
07:57
Sampling and Pretreatment of Tooth Enamel Carbonate for Stable Carbon and Oxygen Isotope Analysis
Published on: August 15, 2018
14.6K
Estimating weaning duration from incremental dentine δ15N and δ13C using a sequence-based LSTM neural network: A deep
Elissavet Ganiatsou1,2, Angelos Souleles1,2, Christina Papageorgopoulou1,2
1Laboratory of Biological Anthropology, Department of Humanities, Democritus University of Thrace, Komotini, Greece.
Plos One
|December 17, 2025
Summary
A new machine learning model accurately estimates infant weaning duration using dental isotopes (δ15N and δ13C). This approach offers a robust tool for understanding past population health and demography.
Area of Science:
- Bioarchaeology
- Paleoanthropology
- Computational Biology
Background:
- Estimating weaning duration in past populations is crucial for understanding health, nutrition, and demographic trends.
- Traditional methods often rely on limited isotopic data, potentially overlooking complex developmental patterns.
Purpose of the Study:
- To develop and validate a novel machine learning approach for estimating weaning duration from serial dentine isotopic data (δ15N and δ13C).
- To introduce a Long Short-Term Memory (LSTM) neural network model capable of analyzing sequential isotopic patterns and tooth-specific variations.
Main Methods:
- Utilized a sequence-based LSTM neural network trained on published serial isotopic data from 279 individuals.
- Incorporated temporal features and tooth type encoding to enhance predictive accuracy.
- Evaluated model performance using RMSE, MAE, and R2 metrics, with uncertainty quantified via Monte Carlo dropout.
Main Results:
- The LSTM model achieved high predictive accuracy with an RMSE of 0.46 years and an R2 of 0.82.
- Demonstrated superior performance compared to established weaning estimation methods (WEAN, ChangeR) in a proof-of-concept study.
- Successfully leveraged sequential patterns in both δ15N and δ13C for inferring weaning processes.
Conclusions:
- Introduced the first machine learning framework for estimating weaning duration from serial isotopic data.
- The model provides a scalable and robust tool for bioarchaeological research, enabling more accurate reconstruction of past weaning practices.
- Highlights the potential of advanced computational methods in analyzing complex biological trajectories.
