595 Impact of Superimposition on 3D Stereophotogrammetric Facial Landmark Reliability

Thursday, March 22, 2012: 3:30 p.m. - 4:45 p.m.
Presentation Type: Poster Session
D. BLANCHETTE1, D.V. DAWSON1, C. KUMMET2, E. WANG3, D. DEFAY4, S.F. MILLER2, and L. MORENO5, 1Dows Institute, University of Iowa, Iowa City, IA, 2Dows Institute for Dental Research, University of Iowa, Iowa City, IA, 3College of Dentistry, University of Iowa, Iowa City, IA, 4Rampton and Defay Orthodontics, Westerville, OH, 5University of Iowa, Iowa City, IA
Objectives:  To evaluate the impact of two approaches to 3D image superimposition on the reliability of soft tissue facial landmark selection as measured by the intraclass correlation coefficient.

Methods:  3D stereophotogrammetric images of 15 subjects were landmarked by a calibrated rater using 3dMd software (Atlanta, GA) for 22 standard anthropometric soft tissue landmarks; a second landmarking was performed by the same rater within two weeks of the first.  The dual landmark coordinates were then exported for intra-rater reliability assessments using intraclass correlation coefficients (ICCs) for each dimension of each landmark. Reliability was initially determined without superimposition, then following superimposition using one of two methods.  Method 1: the landmark subnasale (SN) was translated to the origin for each landmarking of each subject. Method 2: a mean form was calculated from the two landmark configurations. The centroid of this mean form was translated to the origin for each subject for both measures. The impact of superimposition was assessed by comparison of ICCs and Euclidean distances between duplicate landmark measurements.

Results:  Without superimposition, reliability was excellent. The ICC ranged from 0.9971 to 1.0000 (p<0.0001). Subnasale (SN) translation decreased reliability for all 22 landmarks except for the Z coordinate of Gnathion (DICC = +0.001). The ICC ranged from 0.6789 to 0.9996 (p<0.0001).  This translation changed the distance between the repeated measures, changed within-landmark variances and prevented reliability assessment for subnasale.  Mean form centroid translation decreased reliability to a lesser extent for all 22 landmarks except for one coordinate in right tragion, left tragion and gnathion respectively. The ICC values ranged from 0.8326 to 0.9999 (p<0.0001). This translation did not change Euclidean distances, did not alter within-landmark variances and did not prevent reliability assessment for subnasale.

Conclusion:  Failure to superimpose facial landmark configurations can overestimate the reliability of landmark selection in 3D stereophotogrammetric techniques.

This abstract is based on research that was funded entirely or partially by an outside source: Supported by CDC R01DD000295 and by NIDCR T32DE014678

Keywords: Cephalometric analysis, Methodology and Reliability
See more of: Craniofacial Anatomy
See more of: Craniofacial Biology