Validity and reliability of a new edge-based computerized method for identification of cephalometric landmarks

To evaluate the validity and inter- and intraexaminer reliability when on-screen landmarks are digitized manually or when these are computer-assisted by means of a new cephalometric software feature.

Twenty radiographs were digitized four times by two experienced orthodontists using a manual method and an edge-based algorithm that helps landmark identification by detecting the edges of anatomical structures.

The computer-assisted method did not agree with manual digitization in 7 of 13 landmarks and 5 of 10 variables. With a tolerance of 0.5 mm or degrees, the two methods did not agree in cephalometric variables. Intraoperator reliability was improved for B point (x-axis), and Menton (x- and y-axis). It got worse for point A (y-axis). Interoperator reliability was improved for B point (x- and y-axis), Soft Labrale Inferior (x- and y-axis), Soft Pogonion (x-axis), and Menton (y-axis). It decreased for point A (y-axis). Intra- and interoperator reliability got better for only one cephalometric variable under study (SNB).

The edge-locking feature seems to be a promising tool for increasing the reliability of on-screen cephalometric analysis. There seem to be difficulties in locating the appropriate edges when artifacts or soft tissue edges are located near the targeted landmark. The existence of very small, but systematic differences between the two digitization methods manifests the need for further improvement.