Forensic Applications of 3D Whole-Body Scans: Visual evaluation
| Authors | |
|---|---|
| Year of publication | 2025 |
| Type | Article in Periodical |
| Magazine / Source | Forensic Imaging |
| MU Faculty or unit | |
| Citation | |
| web | https://doi.org/10.1016/j.fri.2025.200659 |
| Doi | https://doi.org/10.1016/j.fri.2025.200659 |
| Keywords | Identification; visual traits; 3D models; whole-body models; ROC analysis |
| Description | This study investigates the potential of using three-dimensional digital models of the human body, captured with a [TC]2 whole-body scanner, to assess visual traits for forensic personal identification. The study sample contained 309 volunteers (164 males, 145 females). Fifteen visual traits were assessed (13 unisex and one sex-specific per sex). Intra-observer error was quantified to assess repeatability. Relationships between traits and participants were examined using correspondence analysis. Differences between individuals were expressed as Hamming distances, evaluated by ROC analysis for identification performance (AUC, sensitivity, specificity). Closest-match analysis was used to test rank-1 identification accuracy. The intra-observer error revealed moderate inconsistencies for three traits: prominence of scapulae (??=?0.42), chest shape (??=?0.40), and muscle development (??=?0.43). Correspondence analysis explained 90% of total variance, with the first three axes accounting for 8.1%, 5.6%, and 4.8%, reflecting variation in muscle development and shoulder morphology. ROC analysis demonstrated suitable classification accuracy (AUC?=?0.919). Males showed higher accuracy (AUC?=?0.937) than females (AUC?=?0.901). Closest-match analysis showed that 47.7% of individuals were correctly identified at rank 1, with better results for females (48.3%) than males (41.2%). |
| Related projects: |