Ultrasound
Ultrasound • Ultrasound imaging uses high frequency sound waves beyond the range of human hearing to generate images within objects of interest
Ultrasound • Among the most popular imaging modalities – Safe – Real-time – Relatively more portable than MRI, CT, X-ray • Today, medical ultrasound has become a predominant tool for diagnostic medicine with 159 million exams given annually in the U.S.A., and 250 million worldwide, which is over 72 million exams greater than the next oft used modality (X-ray computed tomography (CT)) in the U.S.A. and over 200% greater than CT exams done worldwide) • Benefits – No ionizing radiation from X-rays or radioactive isotopes – Ability to do real-time 2D and 3D imaging that other modalities • Challenges – Image quality and diagnosis is highly sensitive to the ability and experience of the operator.
Ultrasound imaging • Based on the pulse-echo and doppler concepts • A discrete pulse is sent into the object of interest by a ultrasound transducer that is composed of a piezoelectric material that is excited by a electrical high voltage (HV) pulser. • Once a pulsing pattern has been sent the transducer switches from transmit to receive mode and waits over a preset period for return echoes. • In pulse-echo imaging , the returning echo’s amplitude represents the presence or absence of an object at a given position. • With a longer period more depth is interrogated as the furthest return echoes will have a chance to be received.
Receiver Operation • The echoes are received by the transducer and converted from sound energy into electrical pulses • The return echo signals are amplified depending on the amount of time elapsed since the transducer sent the initial pulse. • A longer delay means the echo signal has traversed more distance and has thus lost an exponential amount of sound energy. • The signal is recovered through a time gain compensation (TGC) implemented with a variable gain amplifier (VGA). These are all part of the receive AFE (Rx AFE) • The analog signals are then converted by an analog-to-digital converter (ADC) into digital samples where further processing is handled by a DSP before the data is displayable as an image or video.
Major components 1) Analog Front-end (AFE), 2) Digital Front-end 3) Digital Mid-end 4) Digital Back-end
Mid-end processing
Detail hardware implementation
Windowing • a window function is a mathematical function that is zero- valued outside of some chosen interval. For instance, a function that is constant inside the interval and zero elsewhere is called a rectangular window, which describes the shape of its graphical representation.
Hamming Window The window is optimized to minimize the maximum (nearest) side lobe, giving it a height of about one-fifth that of the Hann window, a raised cosine with simpler coefficients.
Decimation (downsampling) • Since the typical spectral doppler bandwidth is 20kHz we only need 40kHz to have a valid representation of the signal without aliasing (Nyquist). In reality there are noises that create spectral energy above 40kHz, which need filtering otherwise you will see it after doing the FFT as frequency spikes between 0 and 40kHz. Thus you need an anti-alias filter (AAF) with a sharp cutoff at 40kHz. Since this puts a big burden on the analog filter (high order filter) before the A/D the usual method is to oversample so that the AAF filters out higher frequency spectra and leaves the rest of the noise to be filtered in the digital domain. In this case the AAF filters anything above 75kHz (half-Nyquist of the 150kHz ADC). The bandpass filter takes care of the rest, which is why it is set to 10-11kHz cutoff since we don't expect to see anything above that frequency and downsample to around 2x that (the new Nyquist). • Since we had oversampled the signals we should downsample them in digital domain. The down
Decimation (Downsampling) • Downsample by M means to retain only every M th sample • Ex: Downsample by 4 3 3 2 2 Signal Value Signal Value 4 1 1 0 0 -1 -1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Time (sec) Time (sec)
Downsampling Hardware implementation • Use the basic idea of keeping 1 every Mth sample. – Use a counter that counts to M and keep the Mth sample • Problem: How to throw out the middle samples? – Use a control signal to validate/unvalidate the wanted/unwanted samples • Makes the control logic more difficult – Divide Clk by M such that the wanted samples are only detected by the rising edge of clk/M
Clk divider by M • If M is even: easy, use a counter that counts upto M • If M is odd: more difficult
Clk divider by odd number • It is a shift register that gives a special count. These are called Johnson Counters and provide clock dividers that are even integers, which don't have to only be powers of 2. The starting position provides the phases we need. always @(posedge clk or posedge reset) begin if (reset) count1 <= 3'b001; else count1 <= {~count1[0],count1[2:1]}; end if (reset) count2 <= 3'b100; else count2 <= {~count2[0],count2[2:1]}; end end assign clk_M = count1[0] ^ count2[0]; • This assumes that y[n] from the FIR comes out at the first cycle though. In reality since the FIR design is probably pipelined y[n] will be X's and not valid. Then you may have to adjust the starting reset bits of the counters to correspond to the actual valid x[n] signals.
Scaling • Need log compression and scaling after FFT calculation of samples and computing the magnitude (Im 2 +Re 2 ) • In grayscale there are 8-bits with 0 representing black and 255 representing white. Ideally you want the largest magnitude (after the log compression) to represent 255. So 255/max(mag) * ALL SAMPLES = pixel value. • Of course you can't know what the max is going to be and it takes time to SORT everything once you get a good number of samples. In real life a user chooses this gain so it's simply a user variable into a multiplier that multiplies all the log magnitude data with.
Recommend
More recommend
