medical signal processing
TRANSCRIPT
KAD 2021.12.08
Medical signal processing
2
A signal is any kind of physical quantity that conveys/ transmits/stores information
e.g. (1)electrical voltage, that can be measured on the surface of the skin/head as a result of the heart-/muscle-/brain activities (ECG/EMG/EEG)
e.g. (2)gamma photon detection in radioisotope diagnostics
(1) (2)
1mV/cm
25 mm/s
3
Classification of signals
static – time-dependentperiodic – non-periodicrandom – deterministicpulsed – continuouselectric – non-electricanalog – digital
1mV/cm
25 mm/s
4
in a very special role
electric signals
non-electric signals are transferred to electric ones
advantages of electric signals:they are easy to transform, amplify, transmit
digital signals
analog signals are transferred to digital ones
advantages of digital signals:they are easy to store, the noise can be engineered and influence can be reduced
5
quantity that compares the magnitudes of two signals:
Signal level or Bel-number (or Decibel-number): n
(named after A. Bell)
unit of n : Bel (B) or decibel (dB)
BlgBlgBlg1
2
1
2
1
2
EE
JJ
PPn
decimal logarithm of ratio of two powers (intensities, energies)
6
arcradius
radiusarc
cf. radian
1radmm
dBlg101
2PPn (10d = 1)
instead of Bel number we are using decibel-number
cf. pH (power of Hydrogen)
M 1
HlgpH
7)7(110lgpH
M10H:..e7-
-7
g
7
the relation between power and voltage:
dBlg20dBlg10
dBlg10dBlg10
1
22
1
22
1
21
2
22
1
2
UU
UU
RURU
PPn
):Ohm(2
IRURUIUP
the characteristic unit: power (or intensity/energy),the practical unit: (electric) voltage
signal level with voltages:12 RR
8
3dBdB0,310
dB2lg1021
2
PP
3dB21
1
2 PP
10dBdB110
dB10lg10101
2
PP
20dBdB210
dB100lg101001
2
PP
cf. half life,half value thickness
9
empirical density function
spectrum, as a special density function
H
h
body height, h (cm)
n: number of data
body height, h (cm)
10
hH
h (cm) h (cm)
hn
cm101
160 170 180 190 200 210 160 170 180 190 200 210
Density function Spectrum
area under the curve: n
area under the curve: H
11
all (usual) periodic functions can be expressed as a sum of sine (and cosine) functions from the fundamental frequency and the overtones
where f is the frequency
the sine function, which has the same frequency as theperiodic function: fundamental frequency
2f, 3f, 4f, ... : overtones
(line spectrum)
periodic function: there is a period, T
T
t
Fourier’s theorem for periodic functions (signals)
,1 fT
in music: pitch
in music: timbre/tone color
12
square pulse trainfundamental fr(equency)
fundamental fr.+3rd overtone
fundamental fr.+3rd overtone +5th overtone
fundamental fr.+3rd overtone +5th overtone +7th overtone
spectrumfunction
0
0.5
1
0 5 10 15 20
0
0.5
1
0 5 10 15 20
0
0.5
1
0 5 10 15 20
0
0.5
1
0 5 10 15 20f
f
f
f
13
fundamental fr.+3rd overtone ++ ... +9th overtone
0
0.5
1
0 5 10 15 20
0
0.5
1
0 5 10 15 20
0
0.5
1
0 5 10 15 20
0
0.5
1
0 5 10 15 20f
f
f
f
spectrumfunction
fundamental fr.+3rd overtone ++ ... +11th overtone
fundamental fr.+3rd overtone ++ ... +13th overtone
fundamental fr.+3rd overtone ++ ... +15th overtone 14
0
0.5
1
0 5 10 15 20
0
0.5
1
0 5 10 15 20
.
.
.
.
.
.
f
f
spectrumfunction
fundamental fr.+3rd overtone ++ ... +17th overtone
fundamental fr.+3rd overtone ++ ... +17th overtone ++ ...
15
2...81
41
211...
21
21
21
21
21
32100
kkcf. infinite series
0 1 2 3
16Textbook, Figure VII.3.
Creating an ECG signal from sine functions
17
flute clarinet
Measured spectra
18
all (usual) functions can be expressed as a sum of sine (and cosine) functions
spectrum: continuous
cf. emission spectra of incandescentlight sources
Fourier’theorem for non-periodic functions (signals)
J
19
t
t
t
t
t
U
U
U
U
U
f
f
f
f
f
P
P
d P/df
d P/df
d P/df
T
T
T
f = 1/T0
f = 1/T0
f = 1/T0
f = 1/T0
A A’
B B’
C C’
E E’
D D’
T
line spectrum (1 line)
line spectrum
band spectrum
band spectrum
continuous spectrum
periodicfunction
sine function
non-periodic function
a few periods
more periods
eg. pulsed ultrasound
spectrumfunction
20
fsine
t
Music in time-frequency representation
pitch
timbre
fund
amen
tal
over
tone
s2n
d3r
d4t
h
21
Voiceprint
http://www.nrips.go.jp/org/fourth/info3/index-e.html
oscillogram (sound intensity vs. time)
time –frequency representation
pronounced consonants and vowels22
t (s)
f sine
(Hz)
systole diastole
Heart beats in time-frequency representation(+ oscillogram)
t (s)
U
23
Frequency and amplitude ranges of biological signals
Practical manual, titel page of meas. 1724
Frequency dependent unit: Electronic amplifier
functionssame:and)2()1(
outin
outin
PPPP
1where),()()2()1( outin PP AtPtPA
same: „fundamentalist“ requirementsimilar: realistic requirement
,in
out
PPAP
,in
out
UUAU
power gain (amplification)
voltage gain (amplification)
25
Uin Uout
R1
R2
in21
2out U
RRRU
(frequency independent) voltage-divider
Uin
Uout
R1
R2
frequency dependent voltage-divider: with capacitor26
CR 1
C
in222in2
22
out11
UCR
RCUR
C
RU
Uin Uout
C
R
High-pass/low-cut filter
0),0(if out0 Uat very low frequencies:
at low frequencies: inout0,if URCU
logf
n(dB)
↔ 6 dB/octave
because of the phase difference, the sum should be calculated as vectors
inout,if UU at high frequencies :
at high frequencies the capacitor is a shortcut
supplementarymaterial
stray/parasiticcapacitance
27
Low-pass/high-cut filter
Uin
R
C
in222in
222
out1
11
1
UCR
U
CR
CU
inout0 ),0(if UU at low frequencies:
at very high frequencies : 0),(if out0 U
logf
n(dB)
Uout
at high frequencies: inout01,if U
RCU
↔ -6 dB/octave
the capacitor at low frequencies is a discontinuity
CR 1
C
supplementarymaterial
28
for (1 [on page 24]): AP >1,
n=10 lg AP = 20 lg AU > 0 dB
for (2 [on page 24]): frequency characteristics
f (logarithmic scale)
n(dB)
fl futransfer band (passband)
fl : lower frequency limit fu : upper frequency limit
nmax
nmax-3
ideal amplifier
29
Amplifier with feedback
Without appropriate perspective we may not recognize our real position.
U
UUUUUU
UUUUUUU
U
U
AAAAAAA
AAAAUUA
UAUU
UAU
UUAc
UUAbUUUa
1
)()(
)()(
***
*
0
2
0
20
0
1
0
2*
1
2201
30
:
:,1
**
U
UU
UU
A
AA
AA
positiv feedback:(a) AU=1, amplification: „infinite“
– sine wave oscillatore.g.: ultrasound generator,
heat theraphy(b) AU 1, amplification: very big
– regenerative amplifiere.g.: hearing, outer haircells
negativ feedback: „all“ amplifier
voltage gain with feedback
voltage gain without feedback
> 0, positiv feedback (same phase), AU*> AU (advantage)
< 0, negativ feedback (in opposite phase), AU*< AU (disadv.)
31
3 dBpositiv feedback
without feedback
negativ feedback
positiv feedback: transfer band – narrrower (big disadvantage)higher gain (advantage)
negativ feedback: transfer band – broader (advantage)less gain (small disadvantage)
logf
n(dB)
transfer band
‐50
‐40
‐30
‐20
‐10
0
10
20
30
‐50
‐40
‐30
‐20
‐10
0
10
20
30
‐50
‐40
‐30
‐20
‐10
0
10
20
30
T 1T 2T 3T 4T 5T 6T 7T 8T 9T 10T 11T 12T
‐50
‐40
‐30
‐20
‐10
0
10
20
30
T 1T 2T 3T 4T 5T 6T 7T 8T 9T 10T 11T 12T
32
analog signal: time- and value-contin.
time-discrete value-contin.
time-cont. value-discrete
digital signal: time- and value-discrete
Analog signal – digital signal
33
Nyquist–Shannon sampling theorem:highest frequency component of the signal
fsampl = fmax, reconstructed signal is constant
fsampl = 1.5 fmax, freq. of reconstructed signal is wrong
fsamp = 2 fmax, freq. of reconstructed signal is correct
e.g.: hifi, fmax = 20 kHz
fsampl = 44.1 kHz > 2*20 kHz
time-discrete: the value of the signal is not known for all moments in time
value discrete: the value of the signal can not be arbitrarye.g.: hifi, 16 bit = 216 = 65 536 (CD standard)
24 bit = 224 = 16 777 216 (“best” audio card)
for complete reconstruction the minimum sampling frequency should be twice the frequency of the highest overtone of the signal
34
Pulse processing
integral discrimination differential discrimination
to select only those pulses that are larger than a preset amplitude
to select only those pulses whose amplitudes lie within a preset window
in in
out out
discr
Textbook, Figure VII.32.
35
Distribution functions and ID/DD ”spectra”
0
1
2
3
4
5
6
150 160 170 180 190 200 210 220
02468
101214
150 160 170 180 190 200 210 220
02468
101214
150 160 170 180 190 200 210 220
density function
h (cm)
h (cm)
h (cm)
ID-”spectrum”
DD-”spectrum”
”cumulative-distribution-function”
hN
cm101
M=N0-N
N
how many pulses are smaller than h?
cumulative-distribution-function
how many pulses are larger than h?
36
Concentration of white blood cells
Coulter counter