3128 Precision of 3D Craniofacial Landmark Data Using Magnetic Resonance Scans
L.H.S. CEVIDANES1, B. BUSBY1, C. CHEEK1, W.R. PROFFIT1, G. GERIG1, A.A. FRANCO2, and J.W. VIGORITO2, 1University of North Carolina, Chapel Hill, USA, 2Methodist University of Sao Paulo, Sao Bernardo, Brazil

Magnetic Resonance (MR) data for clinical and basic research in craniofacial growth requires that landmarks in regions of interest of the image be accurately located. Objectives: This study evaluated the repeatability, measurement error and validity of locating anatomical landmarks in high-resolution three-dimensional MRI data with 1mm isotropic voxel resolution. Methods: We identified and defined 10 anatomical landmarks in the coronal, axial and sagittal planes of four head MRIs. The different head scans were automatically aligned with the MIRIT software resulting in a registration with subvoxel accuracy. Each identification was repeated five times by three observers using a tool for interactive visualization and manipulation of 3D data (UNC-IRIS 2000). Results: At least 80% (p<0.001) of the identifications of landmarks located at points of maximum curvature had errors of position less than 1.41 mm (average distances in positions for each observer and interobservers between 0 and 1.32mm, S.D. 0.7mm). At least 80% (p<0.001) of the identifications of landmarks that were points of extremity of local curvature defined with respect to an external co-ordinate system had errors less than 2 mm (average distances between 0.49mm and 1.76mm, S.D. 1.09mm). The landmark location errors within each observer showed that the most trained observer had smallest errors, despite the observers' calibration prior to the landmark identification. The disagreement across the three observers was significant with Anova p-value <0.01, except for the Lower Border of the Mandible landmarks. Conclusion: We have shown that three-dimensional landmark data collected from MR images are internally consistent and precise, being affected by landmark type and observer training. This method will significantly enhance the potential of MR data collection and analysis of bony structures in three-dimensions for quantitative assessment of patient parameters in craniofacial biology.

Seq #286 - Post-natal Craniofacial Growth
9:00 AM-11:00 AM, Saturday, 9 March 2002 San Diego Convention Center Room 17A (Mezzanine Level)

Back to the Craniofacial Biology Program
Back to the IADR/AADR/CADR 80th General Session (March 6-9, 2002)

Top Level Search