Detection probabilities for time-domain velocity estimation

Estimation of blood velocities by time-domain cross-correlation of successive high frequency sampled ultrasound signals is investigated. It is shown that any velocity can result from the estimator regardless of the true velocity due to the nonlinear technique employed. Using a simple simulation programs, it is demonstrated that the probability of correct estimation depends on the signal-to-noise ratio, transducer bandwidth, number of A-lines and number of samples used in the correlation estimate. The influence of applying a stationary echo-canceler is explained. The echo canceling can be modeled as a filter with a transfer function depending on the actual velocity. This influences the detection probability, which gets lower at certain velocities. An index directly reflecting the probability of detection can easily be calculated from the cross-correlation estimated. This makes it possible to assess the reliability of the velocity estimate in real time