Accuracy of two systems--conventional (DRF 400, Diasonics) and color-coded (Angiodynograph, Quantum/Phillips) image-directed Doppler ultrasonography--was investigated using an in vitro model that generated both monophasic and triphasic pulsatile flow patterns. Estimated and actual blood volume flow rates showed good correlations, but the sampling with a hand-held transducer led to wide variations in measurement error for the conventional (-69.2% to 50%) and the color-coded (-79.3% to 265.7%) systems. By performing multiple measurements, one could improve accuracy considering only the maximal values of a series instead of the mean values. Accuracy was impaired by interposed muscular or fatty tissue due to false low time-average velocity measurements caused by a loss of Doppler signal. Comparison of both systems revealed significant differences between pulsatility index values (p less than 0.001), blood flow velocities (p less than 0.001), and blood volume flow rates (p less than 0.05 for program flow, p less than 0.001 for manual and automatic flow program of the color-coded system).