A laser Doppler scanner for imaging blood flow

in skin

T.J.H. Essex and P.O. Byrne

Northern Regional Medical Physics Department, Newcastle General Hospital, Newcastle
upon Tyne, NE4 6BE, UK

ABSTRACT
This paper describes a novel medical instrument that produces an image of bloodfl ow in the capillaries under the skin
surjaa. A laser beam is used to detect blood cell motion from the Doppler broadening of the laser light scatteredfiom the
skin. The image is generated by scanning the t&r beam in a raster. The design of a practical clinical instrument is
outlined and some preliminary results are presented.

Keywords: Blood flow, imaging, laser Doppler scanner, skin

INTRODUCTION Scanner unit

Knowledge of blood flow in the skin is important in

many areas of medicine, such as plastic surgery, bum
management, dermatology and vascular medicine. Ia
Experience from skin blood flow measurement sug- \
gests there is a clinical need for an instrument that /
provides a measurement of skin blood flow in the \,
Operator position
form of an image of the area of skin being studied. :-3 /
An instrument has been developed to do this using
a scanning laser Doppler techni ue. Unlike existing
instruments, which use optical fiB re probes at fixed
sites, this instrument scans a laser beam in a raster
pattern over the skin to build up a laser Doppler
image. It is able to produce an image of an area of
skin in terms of the movement of blood cells under
the skin surface. An area of up to 500 x 700 mm can
Figure 1 Typical clinical arrangement of the laser Doppler scanner
be scanned in 6 min. The spatial resolution at the skin
positioned about 1.6 m above the couch
surface is about 3 mm, which is sufficient for clinical
diagnosis. Trials of the scanner have indicated that it
detects blood cell movement and artefact caused by medicine, including dermatology, vascular disorders
the scanning motion is minimal. and bum assessment, with pro’ected work in breast
Fijpre 7 shows a typical clinical arrangement with tumour detection, and study o 1 pressure sore forma-
the scanner mounted over a couch. The patient lies tion.
on the couch and simply remains still while the scan The scanner has been patented by British Tech-
is made. Normal small movements caused by nology Group (London, UK)‘.
breathing, for example, do not affect the scan. The
fact that the scanner itself does not come near the
patient has some incidental advantages: there is no LASER DOPPLER AND SKIN BLOOD FLOW
risk of contact with open wounds, and the scanner Conventional laser Doppler blood flowmeters”‘” use
cannot influence the blood flow, which can happen optical fibres to carry light from a laser (usually
with a probe-based instrument. helium-neon) to the skin. The light is scattered in the
The scanner is controlled by an IBM-AT-type skin tissue and that which is scattered from moving
computer which is also used for image display and blood cells experiences a Doppler frequency shift
storage. The images are displayed with colours (Figure 2). Optical fibres collect some of this light at
representing a scale of blood flow. Image processing the skin surface and return it to the photodetectors.
and analysis software is now being developed in Typical blood cell velocities are of the order of
conjunction with the clinical evaluation of the instru- 1 mm s-’ in the capillary loops, giving rise to Doppler
ment. frequency shifts of u to 3 kHz, though scattering by
Clinical trials are in progress in various areas of faster moving bloo B in deeper vessels extends the
spectrum to above 10 kHz. Optical frequency shifts of
Correspondence to: T.J.H. Essex this magnitude are too small to detect directly, but
0 1991 Butterworth-Heinemann for BES
0 14 I -.532.5/9 I /O:iO1X9-06
J. Biomed. Eng. 1991, Vol. 13, May 189
Laser Doppler blood flow imaging: T.J.H. Fsex and P.O. Byrne

Photodiodes Lenses k+ >

Transmitting optical fibre Collecting optical fibre

Figure 4 General layout of the scanner

Figure 2 Simplified cross-section of skin. Laser light is scattered by

the skin tissue and by blood cells in the capillary loops
system, and the output is in the form of an image on a
computer display rather than a continuous reading.
because of mixing between the various Doppler
shifted and unshifted components of the light, the
‘IWE SCANNER
Doppler spectrum that appears at the photodetector
output is centred around zero frequency. Optical and mechanical considerations
The light is scattered randomly in the skin tissue, so
the Doppler shifts do not relate directly to the blood F&we 4 is a representation of the layout of the
cell velocities. It is therefore not possible to calibrate scanner. The 2mW helium-neon laser is mounted
laser Doppler flowmeters absolutely. Relationships behind the array of four photodiodes and lenses, and
have been found relating the Doppler spectrum and the beam passes coaxially through this assembly onto
parameters of blood flowing in skin@‘, and the centre of the 250mm diameter plane-scanning
most instruments use analogue circuitry to derive a mirror, which directs the beam onto the target. Light
flow-related variable from the spectrum, which is scattered from the target is collected by the same
expressed in arbitrary units. The term ‘flux’ is often mirror and focused by the lenses (50mm diameter,
used for this output variable to distinguish between a f = 100 mm) onto the four photodiodes (BPW34). The
measurement of true flow, as in a tube, and the function of the lenses is simply to increase the light
general movement of blood cells within a volume of collecting area, the spatial resolution being deter-
tissue, which is what this technique detects. mined by the laser beam diameter, as in the early
Figure 3 shows, on the top, the form of conven- flying-spot television systems. This has the convenient
tional blood flowmeters, and below the main compo- effect of giving the scanner a great depth of field.
nents of the laser Doppler scanner. The fibre optics The scanning mirror is driven horizontally by a DC
are replaced by a moving mirror and light collection motor, gearbox and crank arrangement. This pro-
vides smooth continuous motion and the crank is
arranged to maintain a constant angular velocity of
Laser
1-1 < optics the mirror during each line of the scan. Between lines
the mirror is incremented vertically by a stepper
motor. Mirror position information is derived from an

L Processor
optical encoder on the gearbox shaft and by counting
pulses sent to the stepper motor. A dedicated micro-
processor controls the motors and sends the mirror
-11 position signals to the corn uter to enable it to
assemble the processed Dopp Per information into an
image.
Scanning mirror The use of a single plane mirror for scanning the
Laser L laser beam and collecting the scattered light has the
advantage that the image of the laser spot on the
photodiodes remains stationary as the mirror scans.
Ideally the laser beam and collecting lens should be
coaxial, but the only disadvantage of using off-axis
lenses is that the image of the spot moves over the
photodiode depending on the target distance. Using a
5mm square photodiode keeps the image on the
active surface for a sufficient range of target distance
(0.6-2 m).
Studies with different lens configurations showed
that the total collecting area is the most important
factor for improving the signal-to-noise ratio with no
Figure 3 Comparison of fibre-optic and scanning instruments significant disadvantage of using multiple lenses.

190 J. Biomed. Eng. 1991, Vol. 13, May

 her Z)oppkr blood flow imaging: T2.H. Fsex and P.O. Byrne

The arrangement of four lenses that was finally displayed image 16 colours are assigned to this scale.
used makes the best use of the circular mirror area Some image processing is possible, such as spatial
and by summing the photodiodes in two pairs it is averaging, contrast manipulation and measurement
possible to use a differential input amplifier. The of areas of raised flow. Further processing develop-
completed optical assembly is rugged and easy to ments are being made as clinical work progresses.
align.

Scan size and acquisition speeds SCANNING ARTEFACTS

A continuous scanning motion was used to avoid the Patient movement
need for a mirror settling time, which would have
The Doppler spectrum from the photodiodes results
been necessary with a stepping scan method before
from the mixing of the various Doppler-shifted and
making each pixel measurement. Continuous motion
-unshifted components from the skin, and therefore
decreases the total scan time but there is then the risk
represents movement of blood cells relative to the
of artefacts due to the movement of the beam (see the
static skin tissue. Small movements of the patient,
Beam movement section).
such as those caused by breathing, will therefore not
The lowest frequency of interest in the Doppler
cause additional spectral components affecting the
spectrum is 250Hz, this being a result of filtering to
measurement. It is necessary, however, to avoid gross
limit artefacts at lower fre uencies. This requires a
movements of the patient which would distort the
sample time per pixel of at 8 east 4ms to estimate the
image obtained.
power in the spectrum. Using a continuous smooth
scanning motion at this rate, the line time is 1 s, and
allowing for turnround at the end of lines the total Beam movement
scan time is 6 min, which is quite convenient for
clinical use. The outputs of the analogue processor An important source of artefact arises from the laser
are low-pass filtered at 160 Hz to optimize perform- scanning regions of different reflectivities. This results
in a large AC component in the received light
ance at this pixel rate.
intensity. The differential input to the Doppler
processor virtually eliminates this. The intensity
Doppler signal processing variations appear as a common mode signal to the
The analogue processing circuit used in the scanner photodiodes, i.e. they are highly correlated. The
is based on one developed previously for a conven- Doppler spectra at each photodiode are not signifi-
tional type of laser Doppler flowmeter and validated cantly correlated and the differential input allows
against radioisotope-labelled microspheres in cere- these signals to pass through to the Doppler processor
bral tissue7. (The detailed design of a Doppler signal without the intensity variations. The differential
processor is not the subject of this work.) input also rejects other common mode signals, such
The analogue processor sends two output signals to as laser noise and mains frequency harmonics from
the computer’s A/D converter. One is a light intensity room lighting.
signal, formed by summing the outputs of the four Spectra were obtained from skin at the normal
photodiodes, the other is the flow-related signal scanning speed (1 s per line) and a low scanning
derived from the Doppler spectrum. speed (9s per line), the lower bandwidth limit being
The Do pler spectrum is the AC component of the 5OHz (Figure 5). Additional spectral components at
photodio dp es’ outputs. The top two photodiodes are the higher speed are visible, mainly below 300 Hz,
summed into one side of an AC-coupled differential with minimal difference above this frequency. This
amplifier and the bottom two into the other side. This
provides rejection of any common mode signals. Range.-51 dBV Status Paused
B: Stored RMS: 250
II I I
100
Computer processing
mVrs
Ii ll I
The computer receives the light intensity and blood
flux signals from the Doppler processor via an A/D
converter, and the mirror position information signals
from the motor control microprocessor. Before this 10
information is used to assemble the image, the blood mVrs
/DIV
flux signal is compensated for noise effects.
Circuit noise and shot noise from the photodiodes
appear at the flux output as an offset whose value
depends on the received light intensity. This offset
increases the effective flux reading. To compensate
0
for this a variable offset is subtracted from the flux
value and is obtained from a look-up table of the vrmsA-stop
Start. OHz
noise-intensity relationship of the scanner. The inten- Frequency (Hz)
sity output from the Doppler processor is used to
Figure 5 spectra from skin, showing addItIona spectral
Doppler
index the look-up table. components at low due to scannmg. The lower trace was
frequencies
The compensated flux values are represented on a recorded at a very low scanning speed. the upper a! normal scanning
scale of O-255 arbitrary blood flux units. In the speed

J. Biomed. Eng. 1991, Vol. 13, May 191

Lam Doppler bloodjow imaging: TJ,H. Essex and P.O. Byrne

relationship between the spectra obtained is highly At high flow rates, 14-20 ~1 s-’ and above, the
repeatable, suggesting that the changes are not due to relationship in Figure 6 is a straight line that would
blood flow variations. intercept the origin, as expected. The dip in the line
The range of frequencies of the additional spectral at lower flows is a result of the filtering at low
components suggests that they are the residual effects frequencies to limit scanning artefact. This reduces
of the reflectance variations mentioned above not the sensitivity of the scanner to low blood velocities.
being completely rejected by the differential ampli-
fier. It is therefore necessary to filter the low
Scan speed artefacts
frequency end of the Doppler spectrum to attenuate
these artefact components. A second-order high-pass Sixty pairs of blood flux values were obtained, each
filter with a cut-off frequency of 250 Hz is used for this pair being of an identical area of skin scanned at the
purpose. Some residual corn onents remain after normal and low scan speeds used to obtain Figure 5.
filtering, but the nature of the g oppler signal proces- The 250Hz filter was in place for these measure-
sing is such that lower frequencies carry little weight ments. Each flux value was the mean of 96 pixel
in the computed flux value. measurements. The difference within each pair was
These scanning artefacts and the filtering necessary calculated, as well as the mean and standard devia-
have an effect on the scanner’s measurements. tion of all the differences. The same calculations were
made for 60 pairs of flux values recorded both at the
lower scan speed, representing the variability due to
EVALUATION
blood flow changes and the measurement technique.
Flow tests The mean change in the flux between low and
normal scanning speeds was -0.1 (k8.6) flux units at
The scanner’s response to changes in flow was
a mean flux for all points of 82.5 units. In the group of
evaluated using a suspension of 15pm nylon par-
airs at the low speed, the mean change was -2.5
ticles in water, pumped through a test cell. There was
P56.6) units and the mean flux for all pints 8 1.8 units.
no attempt to mimic skin tissue with the set-up,
These results seem to indicate that any error in the
although the reflectance and Doppler spectra were
flux measurement due to the beam scanning is not
similar to those of skin.
significant compared with the variability in the
The cell consisted of two microscope slides, their
measurement itself. It would be necessary to take a
faces separated by 0.12 mm, sealed down their long
much larger set of readings to extract any systematic
sides, forming a channel of 20 X 0.12 mm. The
error due to the scanning motion.
assembly was sandwiched between two 1 mm thick
pieces of silicone rubber acting as diffusers. The
stationary laser beam from the scanner was aimed
perpendicularly at one of the cell faces. The nylon Clinical results
particle suspensions were pumped through the chan-
Examples of images obtained from the laser Doppler
nel between the slides.
scanner are shown in Figures 7-77.
Figure 6 shows the scanner outputs obtained for
In F&we 7, the dorsal surface of a hand is seen,
various flow rates. These measurements were made at
where the blood flow to one finger was interrupted
a fixed particle concentration of 0.2% v/v, and hence
using an elastic band. There is clearly no flux visible
represent a velocity relationship. The particle concen-
on the scan in this finger.
tration and flow rates chosen were such that the
Figure 8 is the chest of a normal subject after he had
resulting Doppler signals were similar in spectral
made a scratch on the skin with his fingernails. No
shape and amplitude to those obtained from normal
visible changes were apparent on the skin, but the
skin, giving similar output values from the processor.
area affected by the scratch is visible as an area of
The actual velocities in the cell do not relate directly
raised blood flux. The increase in area covered by the
to those of blood cells in skin, however, because of the
scan compared with Figure 7 is achieved by placing
transverse cell flow and the different scattering pro-
the subject further from the scanner (1.6 m compared
perties of the cell compared with skin.
with 600 mm).
The subject in Figure 9 had a skin graft performed
as a child, about 20 years ago. This area shows a
reduced blood flux in the skin.
Figure 70 is the result of a series of tests to obtain a
dose-response curve for a skin irritant. The patches
of raised blood flux are due to various doses of the
irritant and the black square is a known reference
area. Software is being developed to anal se the dose
response in terms of increase in blood Kux and the
size of the area affected. The method is quick and
convenient compared with methods currentIy in use.
Finally, Figure 11 is a scan of a pair of hands, again
0 2 4 6 8 10 12 14 16 18 20 the dorsal surface. The point of interest on this scan is
Flow through test rig (p\ s-‘1 on the right hand, where there was some inflamma-
Figure 6 Relationship between the scanner processor output and tion of the tendon of the first finger. This is reflected
flow through a test rig. Bars show 1 standard deviation from the mean. as an increase in the blood flux in the skin overlying
The processor output is in arbitrary flux units the tendon.

192 J. Biomed. Eng. 1991, Vol. 13, May

 Laser Doppler bloodflow imaging: i?JH. Fsex and P.O. Byrne

Relative perfusion sCOle

Figure 7 her Doppler scan of the dorsal surface of a hand with 2, 2: 5: 7; *

100 125
h 150
?I 175
n 200
n 225
A 250
A Units
blood flow to one finger interrupted Relative perfusion scale

Figure 10 Laser Doppler scans of responses to doses of skin irritant

ii A A n n A
k 2; 5; 7; 100 125 l:O 175 200 225 250 Units
Relative perfusion scale

Figure 11 Laser Doppler scan of the dorsal surface of a pair of

hands; the tendon of the first finger of the right hand is inflamed.

CONCLUSION
A scanning laser Doppler instrument has been
developed that is able to generate an image of blood
flux in the capllaries under the surface of the skin. It
has been demonstrated to detect changes in blood
flux resulting from injury, disease and temperature.
Although it is not possible to calibrate the instru-
ment directly, in most cases a comparison can be
; 2: 5; 7; to”0 125
A 150
A 175
n 200
A 225
A 250
h Units made between the area being studied and an adjacent
Relative perfusion scale or symmetrically opposite part of the body.
Figure 8 Laser Doppler scan of the chest of a normal subject who Clinical studies are in progress to evaluate the
had scratched his skin scanner in various areas of medicine.

J. Biomed. Eng. 1991, Vol. 13, May 193

Laser Doppler blood flow imaging: TJH. Essex and P.O. Byrne

ACKNOWLEDGEMENTS 3. Nilsson GE, Tenland T, oberg PA. A new instrument for

continuous measurement of tissue blood flow by light
The authors are indebted to the numerous clinicians beating spectroscopy. IEEE Trans Biomed Eng 1980;
in the Newcastle hospitals who have contributed BME-27: 12-9.
advice and effort to this work, and from the Medical 4. Bonner R, Nossal R. Model for laser Doppler measure-
Physics Department, Mr E. Horsefield for the mecha- ments in tissue. A# Opt 1981; 20: 2097-107.
nical fabrication of the instrument and Professor 5. Jentink HW, de Mu1 FFM, Hermsen RGAM, Graaff R,
Boddy for supporting the project. Greve J. Monte Carlo simulations of laser Doppler blood
flow measurements in tissue. Appl Opt 1990; 29: 2371-81.
6. Nilsson GE. Signal processor for laser Doppler flowmeters.
REFERENCES Med Biol Eng Comput 1984; 22: 343-8.
7. Eyre JA, Essex TJH, Flecknell PA, Bartholomew PH,
1. British Technology Group UK Patent Nos GB 2 23 1742 A SinclairJI. A comparison of measurement of cerebral blood
and WO 90/ 11044. flow in the rabbit using laser Doppler spectroscopy and
2. Watkins D, Holloway GA. An instrument to measure radionuclide labelled microspheres. Clin Phys Physiol Meas
cutaneous blood flow using the Doppler shift of laser light. 1988; 9: 65-74.
IlL!!X Trans Biomed Eng 1978; BME-25: 28-33.

194 J. Biomed. Eng. 1991, Vol. 13, May