IEEE TRANSACTIONS ON SONICS AND ULTRASONICS. zyxwvutsrqpon VOL. SU-32. NO. zyxwvutsr 4, JULY 1985 zyxwvutsr 555 Velocity Profile Reconstruction Using Ultrafast Spectral Analysis of Doppler Ultrasound PIER0 TORTOLI, GIANFRANCO MANES, AND CARLO ATZENI Abstmct-An ultrasound pulse Doppler system is discussed, capable of performing real time spectral analysis of data originating from a number of sample volumes. The use of a surface-acoustic-wave-based ultrafast spectrum analyzer allows for a large amount of Doppler data to be processed in a time equal to a few tens of microseconds for each equivalent channel. Sonograms comparable to those obtained in single gate-fast Fourier transform-based-Doppler systems can thus be simultaneously presented, showing the velocity distribution in different sample volumes as a function of time. Alternately,the instantaneous velocity profile can be mapped as a function of depth,thus allowing for the possibility of bidimensional imaging. A prototype flowmeter capable of processing Doppler data from 32 range cells in real time is described and preliminary results obtained zyxwvutsrqpon in virro under simulated flow condi- tions are presented. D I. INTRODUCTION OPPLER ultrasound techniques areused in detecting blood flow velocity in real time for application to early diagnosis of cerebral and peripheral vascular diseases. In the last two decades [l] efforts to expand the use of these techniques have been directed mainly towards two objec- tives. The first objective is the introduction of true spec- tral analysis techniques, capableof accurately and reliably detecting Doppler frequencies representing the velocity distribution of the insonated vessel; the second is the de- velopment of multigate detectors, capable of instanta- neously mapping the flow velocity profile as a function of range, thus allowing for the possibility of a bidimensional flow imaging. Digital fast Fourier transform (FFT) is undoubtedly the most flexible means employed for spectral analysis of Dopp- ler waveforms [2], [3]. Digital FFT unitsused in com- mercially available flowmeters are typically capable of providing up to 200 spectra per second with 641128 trans- form points allowing the spectral content of a single sam- ple volume to be detected with adequate speed. The use of an analog spectrum analyzer, based on chirp Z-transform (CZT) processing and employing charge-cou- pled devices (CCD’s) as transversal filters, has also been demonstrated [4]. Using a commercially available 512- point CCD-CZT unit operated for zyxwvuts 100 Hz resolution, a bandwidth of less than 16 KHz was analyzed; there were ManuscriptreceivedNovember 6, 1984; revised March 13. 1985. This paper was supported by the National Research Council, Special Project on Biomedical Engineering, by the Italian Ministry of Education, and by An- saldo Elettronica Biomedicale, Genova, Italy. The authors are with the Dipartimento di Ingegneria Elettronica, Via zyxwvutsr S. Marta, 3, 50139 Firenze, Italy. no practical differences between the results obtained with the CCD-CZT unit and those obtained using a bandpass filter array in a continuous wave (CW) flowmeter. Multigate pulse [5] Doppler systems have beendem- onstrated that sequentially process up to 32 range chan- nels using a zero-crossing counter [6], [7], or that parallel process up to 20 range channels using a bank of fre- quency/voltage converters [8]. Brandestini [9] describes a digital implementation of a multigate flowmeter, in which a large number (up to 128) of range cells covers the full observation range, providing the blood flow profiles along the transducer axis at the rate of 50/s. Doppler analysis was performed by a zero-crossing detector or by a new type of frequency estimator taking advantage of the digital format of Doppler data [ 101. Arteriographic cross-sec- tionalimageswere constructed by rapidlyscanningthe vessel area. A multigate system using a digital FFT for Doppler pro- cessing is described by Abdel-Azim and Hottinger [ 111. A distributed processing implementation allows the power spectrum as well as significant parameters like the mean frequency and spectral variance to be computed in real time.Spectral analysis is performed by a special hard- wired processor capable of providing 32 significant trans- form points in about 400 PS. Sequential FFT’sof 32 range cells are thus performed in a time interval on the order of 13 ms, and this is compatible with the requirement for a processing time that is less than 2.5 percent of the cardiac cycle. As discussed in more detail in the following sections, whenthenumber of rangechannelstobeprocessed is higher than some tens (up to 128), the required analysis speed can reach the order of severalhundredkilohertz. The use of digital FFT units operating in parallel or in sophisticated pipelined architectures could be discouraged by considerations of cost, size, and power consumption. As an alternative approach, a multigate flowmeter, whichwas first developed for Doppler processing in co- herentradarsystems, is described.Themultigate flow- meter uses the analog ultrafast spectral analysis tech- nique, based on the chirp Z-transform principle and implemented with surface acoustic wave(SAW) devices. Chirp transform [ 121, [ 131 is a processing technique which displays Fourier transform by premultiplying the signal by a linear frequency modulation (chirp) waveform and han- dling the product through a chirp filter. The CCD-CZT 0018-9537/85/0700-0555$01.00 zyxwv 0 1985 IEEE