
INTRODUCTION



Volume
blood flow information is important to the diagnosis of cardiovascular
diseases. In
Doppler ultrasound, the mean velocity of the sampled
flow is directly related to the mean frequency of the
echo signal. Perhaps
the major difficulty in velocity estimate is the accurate
determination of the Doppler shift in a noisy
environment. 
Because
Doppler shift is a linear function of the flow velocity, signal
processing
in the frequency domain is natural. However, basing on FFT
techniques, these methods are noisesensitive and
suffer from spectral leakage effects, due
to
windowing that are inherent in finitelength data records. 
An
autocorrelation velocity estimator operates in the time domain. It is
based on using a PW Doppler system that transmits a
gated
tone burst while the perchannel echo signals are
independently range gated. The gated signals are phase aligned, summed
and converted into baseband.
Finally, the
obtained I/Q components are integrated over the combined time interval
of the all range gates and applied to the velocity estimator. 
The
estimation of velocity can also be done by finding the timeshift
between two consecutive RF signals directly.

Typically,
the timedomain velocity estimators use the Kasai
algorithm wherein the mean frequency is directly derived from a
pair
of successive I and Q signals as follows: 
(1) 




ADVANCED
DOPPLER DATA COMPUTATION



It is widely
accepted that the Kasai technique is
both accurate and computationally efficient. As seen
below, its hardware implementation is quite
simple too. However, the signals from the blood is much
weaker than the signal from the tissue. Furthermore, in the
Doppler operating mode, the time period of the signal phase variations
depends upon the blood flow velocity. In slower flows, the period of
the Doppler signal becomes longer. It means that the phase increments
between two adjacent samples N and N  1 are reduced, i.e.,


Consequently,
accurate estimation of the numerator in Eq. 1 requires an
excessive bit resolution or/and spatial averaging.
Unlike to conventional techniques, the proposed
solution is based on
a hybrid analog/digital architecture. This allows to
implement lowerresolution and
hence more costeffective ADCs. 






