Nose-skull relationships for craniofacial identification: computed tomography (CT) assessment of nose width, nasal...

The nose is one of the primary facial features assessed from skulls when undertaking craniofacial identification. While several nose estimators have been published, out-of-group tests are relatively few. In this study, we use high-resolution CT scans and a pool of 92 Australian adults to assess three previously published nose estimation models: (1) nose width estimated from the maximum width of the nasal aperture; (2) nasal tip curve (inferior oblique view) estimated from the superior nasal aperture shape; and (3) pronasale position estimated from two-tangents according to recent revisions. Tests of Method 1 in 63 subjects reveal standard errors of the estimate (SEE) of 4.0 mm. Tests of Method 2 in 31 carefully screened subjects produced a Procrustes distance (d) of 0.24. Tests of Method 3 in 43 subjects, yielded a SEE of 10 mm for Euclidean distances between the estimated and ground truth pronasale position. Cartesian coordinate error (x/y) for the latter method, revealed pronasale was placed too far antero-inferiorly in most subjects (x = +2.2 mm and y = -6.2 mm). These results indicate that current published nose estimation models often yield noses with substantially incorrect dimensions for all three factors: nose width, nose tip curve and position of the nose tip. To improve the established methods, we propose a correction factor (sign-reversed mean error) that delivers a calibration-in-the-large for the estimation models helping to increase their accuracy.