Craniofacial reconstruction evaluation by geodesic network.

Area of Science:

• Forensic Science
• Medical Imaging
• Computer Graphics

Background:

• Craniofacial reconstruction estimates facial models from skulls, crucial for forensics, archaeology, and cosmetic surgery.
• Objective evaluation methods for craniofacial reconstruction accuracy are lacking.
• Existing methods often rely on subjective assessments, limiting reproducibility.

Purpose of the Study:

• To propose and validate an objective, quantitative method for evaluating craniofacial reconstructions.
• To assess both global and local similarities between reconstructed and original facial models.
• To compare the proposed objective evaluation with subjective human assessments.

Main Methods:

• Constructing geodesic networks for both reconstructed and original craniofacial models using geodesics and isogeodesics.
• Defining facial features at network vertices using weighted averages of shape index values.
• Calculating holistic similarity via correlation coefficients of features between corresponding network vertices.
• Measuring local similarity by dividing models into six subregions (forehead, eyes, nose, mouth, cheeks, chin).

Main Results:

• The proposed geodesic network method provides an objective measure of craniofacial reconstruction accuracy.
• Holistic similarity quantifies overall resemblance, while local similarity assesses accuracy in specific facial areas.
• Experimental results with 100 datasets show strong consistency between the objective method and subjective evaluations by 35 individuals.

Conclusions:

• The developed objective method offers a reliable and reproducible way to evaluate craniofacial reconstructions.
• This approach can enhance the accuracy and trustworthiness of craniofacial reconstructions in various applications.
• The findings suggest that objective, network-based analysis can effectively complement subjective assessments in facial reconstruction evaluation.