Forensic personal identification using two-dimensional computed tomographic imaging of the frontal sinus: a comparative analysis of feature matching algorithms.

BACKGROUND

Personal identification plays a crucial role in forensic investigations. The frontal sinus (FS), due to its unique anatomical characteristics, is often used for human identification with radiographic imaging. This study evaluates the feasibility of personal identification using computed tomographic (CT) images of the FS with feature matching programs, the Accelerated-KAZE (AKAZE) and Oriented Features from Accelerated Segmented Test and Rotated Binary Robust Independent Elementary Features (ORB).

METHODS

CT images of 180 deceased individuals (135 males, 45 females) were analyzed. Antemortem (AM) and postmortem (PM) CT images of the FS were processed and compared using the AKAZE and ORB algorithms. The similarity between the paired AMCT and PMCT images was calculated using brute-force matching. Welch's t-test was used to assess differences in similarity scores between the matched and mismatched groups. Receiver operating characteristic (ROC) analysis was used to evaluate the accuracy of the identification.

RESULTS

The mean similarity scores of the matched groups were significantly lower than those of the mismatched groups across all subgroups (p < 0.001). ROC analysis showed that AKAZE had a higher true positive rate (TPR = 0.950) and a lower false positive rate (FPR = 0.129) than ORB (TPR = 0.883, FPR = 0.172) for all samples.

CONCLUSIONS

Feature-matching algorithms applied to CT images demonstrated high reliability for personal identification. AKAZE performed better than ORB in distinguishing between the matched and mismatched groups. These findings support the potential application of this method to forensic investigations, particularly where DNA, fingerprints, and dental materials are not available.