Accuracy of computerized automatic identification of cephalometric landmarks

Computerized cephalometric analysis can include both landmark identification and determination of linear or angular measurements. Although its use is time saving compared with a manual method, the accuracy of automatic landmark identification remains unclear. The purpose of this study was to evaluate the accuracy of a computerized automatic landmark identification system that used an edge-based technique. The technique divides the scanned cephalogram into 8 rectangular subimage regions. After the resolution of these subimages is reduced, the edges are detected and the landmarks are located automatically. Thirteen landmarks were selected for assessment on a set of 10 test cephalograms. The results showed that the errors between manual and computerized identification for landmarks were not significantly different (P > .05) for 5 of 13 landmarks: sella, nasion, porion, orbitale, and gnathion. These results suggest that the accuracy of computerized automatic identification is acceptable for certain landmarks only. Further studies to improve the accuracy of computerized automated landmark identification are needed.