Login / Signup

Comparative Assessment of a Novel Photo-Anthropometric Landmark-Positioning Approach for the Analysis of Facial Structures on Two-Dimensional Images.

Marta R P FloresCarlos E P MachadoMatteo D GallidabinoGustavo H M de ArrudaRicardo H A da SilvaFlávio B de VidalRodolfo F H Melani
Published in: Journal of forensic sciences (2018)
Positioning landmarks in facial photo-anthropometry (FPA) applications remains today a highly variable procedure, as traditional cephalometric definitions are used as guidelines. Herein, a novel landmark-positioning approach, specifically adapted for FPA applications, is introduced and, in particular, assessed against the conventional cephalometric definitions for the analysis of 16 landmarks on ten frontal images by two groups of examiners (with and without professional knowledge of anatomy). Results showed that positioning reproducibility was significantly better using the novel method. Indeed, in contrast to the classic approach, very low landmark dispersions were observed for both groups of examiners, which were usually below the strictest clinical standards (i.e., 0.575 mm). Furthermore, the comparison between the two groups of examiners highlighted higher dispersion consistencies, which supported a higher robustness. Thus, the use of an adapted landmark-positioning approach proved to be highly advantageous in FPA analysis and future work in this field should consider adopting similar methodologies.
Keyphrases
  • deep learning
  • convolutional neural network
  • healthcare
  • magnetic resonance
  • optical coherence tomography
  • body composition
  • magnetic resonance imaging
  • minimally invasive
  • high resolution
  • current status