Medical diagnostic systems 

(Orvosbiológiai képalkotó rendszerek) 

B-mode imaging components 

( B-mód képalkotás összetevői) 
Miklós Gyöngy 


The origins: pulse-echo ranging [Szabo 2004, pp. 1-12] Sonar: SOund NAvigation and Ranging 
– 
Titanic disaster (1912) 

– 
Anti-submarine warfare (1916-) Radar: RAdio Detection and Ranging 

– 
Tesla (1917) 


• 
Early experiments in medical ultrasound came from equipment and experience in above two fields 

• 
Ranging (distance measurement based on time of arrival information) relies on relatively constant speed of sound 




The origins: using an oscilloscope 
• 
Echo returning from transmission observed on oscilloscope 


• 
Amplitude-mode (A-mode): traditional oscilloscope display 


• 
Brightness-mode (B-mode): display envelope of each A-line on top of each other 


transverse received distance voltage 
time - longitudal distance 



Multiple reflections from a boundary. Left: A-line. Right: B-mode image 


The role of technology [Szabo 2004, pp. 16-20] Advances in transducers 
– 
piezoelectricity (Curie brothers, 1880) 

– 
mass, reproducible manufacture 



– miniaturization (e.g. MEMS) Advances in electronics 
– 
application-specific integrated circuit (ASIC) 


– 
digital signal processors (DSP) 

– 
very large scale integration (VLSI) 


– 
move towards digitization (beamforming, TGC) 

– 
reduced cost of digital storage 




Pulse-echo pathway (A-line) 




ADC (later if analogue beamformer) 




(compression, 
downsampling, 
projection) 






User control/access Transmission 
• 
Typical commercial system: 

• 
choose imaging depth (determines focus) 


• 
choose frequency (determines waveform) 




• 
Research system: arbitrary transmission Reception 

• 
Typical commercial system: 

• 
access to bitmap screen grab 


• 
access to post-beamformed RF data (maybe!) 



• 
Research system: pre-beamformed channel RF 



panel of imaging parameters on the z.one ultrasound system (ZONARE Medical Systems) 


Needs for user control/access Clinician: 
• 
basic parameters (resolution, depth) Researcher of registration/segmentation 

• 
ideally post-beamformed (BF) data Researcher of new imaging modalities: 

• 	
some research possible with BF data (e.g. estimation of acoustic parameters) 

• 	
ideally, total control over imaging parameters 

• 	
calibration of transmitted and received signal for quantitative studies 





panel of imaging parameters on a z.one ultrasound system (ZONARE Medical Systems) 


Ultrasound systems for research use 

Commerical (C)/  Channel data (C)/  
Name  Purpose-built (P)  Post-beamformed (P)  Other options  
Antares (Siemens)  C  P  
DiPhAS  
(IBMT,Fraunhofer)  
LeCoeur (OPEN)  C  C  arbitrary transmission  
RASMUS (DTU)  P  C  arbitrary transmission  
SonixTouch (Ultrasonix) C  C  imaging parameters  
SONOS 500 URP  
(Agilent + U. Virginia)  C/P  C  
SITAU FP (Dasel)  C  C  programmable width transmission  
t3000 (Terason)  C  P  arbitrary apodization, focal depth  
ULA-OP (U. Florence)  P  C  arbitrary transmission  
z.one  ZONARE  C  C  (on request) arbitrary transmission  
[Tortoli et al. 2009; Wilson et al. 2006]  


Transmit/Receive switch 
• 	
Implementations: 

– 	
diode 


– 	
transmission line (frequency selective) 



• 	
Transmission: ~10 V; Reception: ~mV 

• 	Some leakage will always occur 


• 	
Receive circuitry needs to be resistant to saturation blinding (especially from matching layer) 



Multiplexing 
• 	
Reduction of complexity 

• 	
Maintain fixed subaperture during linear scan 


element i 
channel i 
element i+64 


MUX i+128 (if 192 elements) 

• 	Shifting of subaperture during linear scan: (1,2,...64), (65,2,...,64), (65,66,3,...,64), etc. 

Time-gain compensation (TGC) [Brunner 2002] 
• 	
Diffraction loss relatively unimportant. Consider, in the worst case, spherically diverging Tx/Rx beams. Identical scatterer at 5 cm, 10 cm, causes -12 dB signal difference. 

• 	
Tissue attenuation ~1dB/MHz/cm. 5 MHz signal, 10 cm penetration depth, causes -100 dB loss. 

• 	
Linear-in-decibel variable-gain amplifiers (VGA) needed to for time-gain compensation (TGC) 



Frequency-shift compensation [Szabo 2004, pp. 86-88] 
• 	
Tissue causes frequency-dependent attenuation 

• 	
Frequency peak of a Gaussian-modulated pulse shifts with distance (~1 MHz for 5 cm imaging depth, 50% fractional bandwidth) 

• 	
Depth-dependent compensation needed (but where in the signal processing pathway is it most appropriate?) 




Medical diagnostic systems – B-mode imaging 

Focal Point 


Digital beamforming 
adapted from [Brunner2002] 

Analogue beamforming 
– 
Difficult to match channels across delay lines 

– 
Many delay taps needed or phase shifting 


+ Only one ADC needed – can make it high-spec Digital beamforming 
– 
High cost of in-sync, fast (vs) high-resolution ADCs 


– 
Large bit depth and sampling rate incur large storage and computational costs 


+ 
Easier to program/configure 


+ 
Novel implementations (e.g. several receive beams) 


[Brunner 2002] 

ADC considerations 
• 	
Fast MHz applications, flash ADC is used (comparator for every signal level) 

• 	
Oversampling: sample at a higher rate, take average of values. 


E.g. 10 bits at 100 MHz can generate 12-bit data at 25 MHz 


• 	
Sigma-delta processing: “pulse density modulation” – local density of 1s represents value (used both for ADC and DAC) 

• 	
IQ (in-phase/quadrature) modulation/demodulation 



IQ demodulation 
Bandwidth of 
1. 	
Mix bandwidth of interest down to baseband 

2. 	
Apply LFP -fff/2 


c cs(mixing (Nyquist frequency) frequency) 
3. Sample at reduced sample 
rate (less storage cost) 

-2fc 	fs/2 
Low-pass filter 
RF signal recovery 

(LPF) 
1. Upsample to original sample rate (interpolation) B/2 IQ demodulation 
2. 	Remodulate by mixing 
adapted from [Kirkhorn 1999] 
frequency fc 




IQ (in-phase/quadrature) data: interpretation 
xIQ = LPF{exp(-.t) xRF}= 
c

LPF{cos(.t) xRF -jsin(.t) xRF }= xI + jxQ
c	c

• 
Express RF signal as sum of slowly varying signal i(t) modulating in-phase 

cosine oscillation and slowly varying q(t) modulating quadrature sinusoid xRF = i(t)cos(.ct)+ q(t)sin(.ct) where i(t), q(t) are slowly varying 

• 	
IQ signal is then xIQ = 0.5 LPF{i(t)(1+cos(2.t)-jsin(2.t)) + q(t)(sin(2.t)-j-jcos(2.t))} 


cc	cc
= 0.5 i(t) -0.5 jq(t) 

• 	
Low-pass filter removes ±2fc 

• 	
Re{xIQ} contains in-phase signal 

• 	
Im{xIQ} contains quadrature signal 

• 	
|xIQ| gives envelope 





Medical diagnostic systems – B-mode imaging 

IQ example: cosinusoid (in-phase) around t=0 µs, sinusoid 
(quadrature) around t=2 µs (both 3 cycles at 5 MHz) 


signal 
1 


0.1 
0.05


RF signal 0 
-1 
0 


power spectrum 


-1 0 1 2 3 -50 -40 -30-20 -10 0 10 20 30 40 50 


-1 0 1 2 3 -50 -40 -30-20 -10 0 10 20 30 40 50 

IQ signal (after LPF) Note how IQ signal can  0.5 1  
be  sampled  at  much  0  
lower rate!  -1 -0.5  0  1  2  3  -10  0  10  
time (µ
s)  frequency (MHz)  



Envelope detection 
• 	Take magnitude of xIQ 
1 
OR 
0.8 

• 	
Hilbert transform H{.} of 

0.6 reconstructed xRF : 90o phase shift 0.4 


• 	
Analytic function of r(t): 0.2 xRF(t) + j×H{xRF(t)} 0 


-0.2 

• 	In Matlab: abs(hilbert(r(t))) -0.4 
(hilbert(.) actually generates -0.6 analytic function!) -0.8 

• 	In your own time: consider -1 similarities between IQ and Hilbert transforms 


Scan conversion 
• Log compression for perception of large (~60 dB) dynamic range 
• Threshold to reject noise 
0 
0 
0 
10 
10 
10 
20 
20 
20 
30 
30 
30 
40 
40 
40 
50 
50 
50 
60 
60 
60 
70 
70 
70 
80 
80 
80 

im log(im) log(max(im,value)) 






[Brunner 2002] Ultrasound system considerations and their impact on front-end components 
[Kirkhorn 1999] Introduction to IQ-demodulation of RF data. http://folk.ntnu.no/htorp/Undervisning/TTK10/IQdemodulation.pdf 
[Szabo 2004] Diagnostic ultrasound imaging: Inside out 
[Tortoli et al. 2009] ULA-OP: an advanced open platform for ultrasound research 
[Wilson et al. 2006] The Ultrasonix 500RP: a commercial ultrasound research interface