Adult Skeletal Age-at-Death Estimation through Deep Random Neural Networks: A New Method and Its Computational...

Area of Science:

• Forensic Anthropology
• Bioarchaeology
• Computational Biology

Background:

• Accurate age-at-death estimation from skeletal remains is vital for human identification.
• High variability in adult senescence processes complicates traditional age estimation methods.
• Existing methods often lack sufficient accuracy, particularly for older individuals.

Purpose of the Study:

• To develop and validate a novel, accurate method for age-at-death estimation in adult skeletal remains.
• To improve the precision of age estimation, especially for the elderly population.
• To provide a computational tool for forensic anthropologists and researchers.

Main Methods:

• A multifactorial macroscopic analysis incorporating 64 skeletal traits.
• Application of deep random neural network (DRNN) models for age estimation.
• Utilizing a reference dataset of 500 identified skeletons (19-101 years old).
• Employing a regression approach for point and prediction interval estimates.

Main Results:

• Achieved accurate age estimation across the entire adult age span with a mean absolute error of approximately 6 years.
• Demonstrated the ability to obtain informative estimates and prediction intervals for the elderly.
• Validated the method through cross-validation and computational experiments.

Conclusions:

• The proposed deep random neural network approach significantly enhances the accuracy of age-at-death estimation from skeletal remains.
• The DRNNAGE software tool offers a valuable resource for the scientific community.
• This method provides reliable age estimates and prediction intervals, addressing limitations of previous techniques.