We have developed a novel, semiautomated carotid intima-media thickness (CIMT) border detection program (AUTO) and evaluated its measurement reproducibility and accuracy. Images from 6 carotid segments were acquired in 50 subjects, for a total of 300 segments. Mean and maximum CIMT values were measured blindly at a reference (REF) lab and in duplicate by experienced (EXP) and novice (NOV) readers using manual (MAN) and AUTO methods. Coefficients of variation for AUTO measurements of mean (3.2%) and maximum (4.1%) CIMT were low, and the AUTO method improved the NOV reader's reproducibility. Compared with the REF lab, mean (0.012 +/- 0.006 mm) and maximum (0.144 +/- 0.006 mm) CIMT biases were small and equivalent to those of the REF lab ( P < .001). The AUTO method shortened reading times by 35% to 46% ( P < .001). We conclude that our novel AUTO CIMT measurement program improved reproducibility and was accurate. Compared with MAN tracing, the AUTO method agreed better with the REF lab and decreased reading time.