OBJECTIVE The aim of this paper is to compare the accuracy of cephalometric landmark identification using three-dimensional CT (3D-CT) surface rendering with "high-dose" (200 mAs) and "low-dose" (35 mAs) CT protocols. The absorbed dose levels for radiosensitive organs in the maxillofacial region during the exposure for both 3D-CT protocols were also measured. METHODS The study population consisted of 15 human dry skulls examined with spiral 3D-CT. 12 cephalometric anatomical landmarks at 7 sites were identified on the 3D-CT surface renderings by 2 observers independently, twice each, using high-dose and low-dose protocols. In total, 1440 imaging measurements were made. Thermoluminescent dosemeters (TLDs) were placed at ten sites around the thyroid and submandibular glands and the eyes in an Alderson phantom for measuring the absorbed dose levels. RESULTS The intraobserver mean distances between 3D landmarks were smaller for all sites with the high-dose protocol (P = 0.37). There was a significant difference among the observers (P = 0.000004). Interobserver mean distances between 3D landmarks were smaller for four of the seven sites with the low-dose protocol. However, the global interobserver mean distances between 3D landmarks for all sites were smaller with the high-dose protocol (P = 0.028). The low-dose protocol reduced the radiation dose to the thyroid by 6.12, to the submandibular salivary glands by 5.91 and to the eye by 5.44, resulting in a global reduction factor of 5.71. CONCLUSIONS The accuracy in the landmark's identification was maintained when the milliampere-second values were reduced from 200 mAs to 35 mAs. We recommend use of the low-dose protocol for clinical 3D-CT cephalometric applications.