Computational forensic identification of deceased using 3D bone segmentation and registration.

Area of Science:

• Forensic Science
• Radiology
• Computer Science

Background:

• Current methods for identifying unknown deceased individuals via radiology are often subjective and lack statistical rigor.
• There is a need for more objective and statistically sound approaches in forensic identification.

Purpose of the Study:

• To develop a 3D computational approach for objective identification of unknown deceased individuals.
• To enhance the accuracy and reliability of radiologic identification methods.

Main Methods:

• A custom Python script was developed for automated 3D segmentation and registration of antemortem (AM) and postmortem (PM) CT scan data.
• The study utilized CT scans of sternal bones and the fifth thoracic (T5) vertebrae from anonymised AM and PM datasets.
• Identification accuracy was assessed using the Dice coefficient to measure the similarity between registered AM and PM bone data.

Main Results:

• Individual identification accuracy reached 86.7% for the sternal bone and 88.9% for the T5 vertebra.
• Challenges included insufficient CT quality and altered bone morphology due to surgical interventions.
• Combining both sternal bone and T5 vertebra data for identification increased accuracy to 97.8%.

Conclusions:

• The developed 3D computational tool offers a promising, objective approach for identifying unknown deceased individuals.
• The method's accuracy can be further enhanced by integrating data from multiple bone structures.
• The publicly available tool has the potential to be adapted for broader applications in forensic identification.