Beyond the Nasal Index: Anthropometric Compression Loss and Uncertainty-Aware Sex Estimation for Disaster Victim Identification

Disaster victim identification (DVI) often requires rapid biological profiling when primary identifiers are unavailable or delayed. Nasal morphometry may assist sex estimation, but the conventional nasal index (NI) compresses nasal width and height into one ratio and may discard useful biological information. We re-analysed a de-identified dataset of 200 adults from Ranchi District, India (100 male, 100 female), independently reconstructed NI, audited derived fields, and evaluated whether preserving primary dimensions improved discrimination. One technically implausible height value was excluded from the prespecified primary analysis, leaving 199 observations. We introduce Anthropometric Compression Loss (ACL), defined as the cross-validated AUC difference between nasal width plus height and NI alone. Under repeated stratified 10-fold cross-validation with 10 repeats, NI achieved AUC 0.537, whereas width plus height achieved AUC 0.809, yielding ACL 0.272 (95% CI 0.179–0.365). Sensitivity analysis retaining all 200 records produced ACL 0.250. Linear models matched or exceeded nonlinear ensembles, indicating that the gain arose from information preservation rather than algorithmic complexity. A reject option returned ambiguous cases as indeterminate. These findings support uncertainty-aware nasal morphometry only for supplementary DVI profiling and candidate prioritization, not individual identification, pending independent external validation across populations, settings, and modalities.