Provided that the operation is carried out with care, the main source of error in the conventional process of compiling data from radiographic headplates is landmark identification. When a digitizer is used, provided that the instrument is in order and is correctly operated, landmark identification is the only source of error. The present study aimed to quantify the error in the identification of 15 commonly used cephalometric landmarks and to determine the variation in direction of the error.
The study was carried out using a digitizer which registered the coordinates of the landmarks twice in a coordinate system with the X-axis parallel to NSL. Generally speaking the method error was small, but different landmarks showed considerable variation. The error was 4–5 times as large for basion as for incision inferius—the two extremes in landmark location reliability. In individual cases there were some pronounced discrepancies between the two registrations. The deviations between the first and second registrations were diagrammatically spread along the axes in a way which appeared characteristic for each landmark.