Magnetic resonance imaging for multi-factorial age estimation using Bayes' rule: a validation study in two...

Area of Science:

• Forensic anthropology
• Radiology
• Biometrics

Background:

• Multi-factorial age estimation (MFA) models combine MRI data from third molars (T), wrist (W), and clavicles (C).
• Previous internal validation showed promising performance, but external validation is crucial for practical application.

Purpose of the Study:

• To externally validate MFA models using T, W, and C MRI data in independent Caucasian samples.
• To assess the accuracy and reliability of age estimation in forensic and legal contexts.

Main Methods:

• Prospective MRI scans of wrist and clavicles (W+C) were conducted in 108 Ghent volunteers (16-21 years).
• MRI scans of third molars, wrist, and clavicles (T+W+C) were performed on 335 Graz males (13-24 years).
• Developmental staging and intra-observer reliability were assessed; age estimation performance was evaluated.

Main Results:

• Clavicle staging exhibited the highest intra-observer variability.
• The W+C model achieved mean absolute errors of 1.49-1.66 years across samples and sexes.
• The T+W+C model in the Graz sample yielded a mean absolute error of 1.41 years, with high correct categorization rates for adults and minors.

Conclusions:

• Both T+W+C and W+C MFA models demonstrated comparable age estimation accuracy in external validation.
• A higher risk of misclassifying minors as adults was observed compared to internal validation.
• Emphasized the critical importance of prediction intervals for reliable age estimation in practical forensic applications.