THE COMMON PATTERNS OF BLOOD PERFUSION IN THE FINGERNAIL BED SUBJECT TO FINGERTIP TOUCH FORCE AND FINGER POSTURE

Haptics-e, Vol. 4, No. 3, July 21, 2006
http://www.haptics-e.org
THE COMMON PATTERNS OF BLOOD PERFUSION IN THE FINGERNAIL BED
SUBJECT TO FINGERTIP TOUCH FORCE AND FINGER POSTURE

Stephen A. Mascaro
H. Harry Asada
Department of Mechanical Engineering
Department of Mechanical Engineering
University of Utah
Massachusetts Institute of Technology
smascaro@mech.utah.edu
asada@mit.edu


ABSTRACT
in the capillaries beneath the fingernail. In previous work,
photoplethysmograph fingernail sensors have been designed
When the human fingertip is pressed against a surface or
which optically measure the two-dimensional pattern of blood
bent, the hemodynamic state of the fingertip is altered in a way
perfusion beneath the fingernail [10]. These patterns can then
that is common to all people. Normal force, shear force, and
be used to estimate the fingerpad forces and finger posture.
finger extension/flexion all result in visibly distinct patterns of
In order to better design such fingernail force sensors, it is
blood volume or perfusion beneath the fingernail. These
important to understand the sensing mechanism, including the
patterns of blood perfusion can be used not only to monitor the
mechanics of the fingernail-bone-tissue interaction and its
state of the finger, but also to understand how the fingernail
effect on blood perfusion. In previous research, the mechanism
interacts with the bone and surrounding tissues when various
behind the hemodynamic response to normal force has been
forces or postures are applied.
quantitatively modeled, but the response to shear force and
In this paper, photographic techniques are used to catalog the
finger bending were not understood [11]. It was suspected that
average patterns of fingernail coloration corresponding to
normal forces, shear forces, and changes in finger posture all
various states of applied forces and postures across human
result in measurably different blood perfusion patterns that are
subjects of a variety of size, gender, and skin color. Results
(to some degree) common to all people. Up until now, this has
indicate that there are at least seven different states of force and
not been substantiated.
posture that yield distinct coloration patterns that are
In this paper, the observable fingernail color patterns that are
statistically significant and common to people in general.
representative of blood perfusion are cataloged for a variety of
human subjects in response to a variety of force and posture
1. INTRODUCTION
states or poses. Photographic results are combined to create a
There is an increasing interest in understanding fingertip
set of “average” patterns that apply to all subjects. The patterns
forces in the growing fields of haptics and virtual reality, in
from all the subjects are correlated to these average patterns to
addition to more established fields such as robotics and
determine whether the average patterns can be used to classify
medicine [1]. These forces act as bi-directional feedback
the responses of all people in a statistically significant manner.
between human and environment, either mechanical or virtual.
Finally, important outcomes are discussed, including HCI
Forces applied by a machine or virtual tool are fed back and
applications as well as unified modeling of fingertip behavior
presented to the human, while forces applied by the human are
during touching and grasping.
measured and fed back to the machine or virtual environment.
Both application and measurement of fingerpad forces are
2. FINGERNAIL COLOR PATTERNS
required, and understanding the mechanics and dynamics of the
As the human fingertip is pressed down on a surface or bent,
human fingerpad is important for both.
the blood perfusion in the fingertip is affected. In fact, the
Several researchers have investigated the mechanics and
change in blood perfusion is characteristically non-uniform
dynamics of the human fingerpad [2-5]. Resulting analyses lead
across the nail, resulting in distinct patterns of red and white
to a better understanding of human grasping and manipulation,
fingernail coloration for different types of forces and posture.
characterizations of the human haptic sense, ergonomic design
Figure 1 shows the primary variables of interest that affect the
criteria [2,3], and performance criteria for haptic feedback
coloration of the fingernail. These include the normal force, F
devices [6]. However only a few studies have taken into
z,
the lateral shear force, F
account the role of the fingernail in fingerpad behavior [7]. It is
x, and the longitudinal shear force, Fy,
which occur when the finger is pressed against a surface. Since
well documented in medical literature that the fingernail plays
the finger surface is curved, the direction of force and its
an important role in grasping and fine manipulation [8,9].
influence on coloration vary depending on the location, contact
In addition to applying forces, the fingernail has recently
angle, and contact surface shape. This paper focuses on the case
been discovered to be useful for measurement of forces. When
where the fingerpad is pressed against a large, uniformly flat
forces are applied to the fingerpad, interaction between the
surface parallel to the bone of the distal phalanx, such that the
fingernail, bone, and tissue alters the hemodynamic state of the
contact area is maximized and the contact is most stable. Thus
finger, creating various patterns of blood volume or perfusion

1

---

Haptics-e, Vol. 4, No. 3, July 21, 2006
http://www.haptics-e.org
the contact location is standardized and the three touch force
Second, when sample images are broken down into their RGB
directions are defined with respect to the surface.
(red-green-blue) components, results show that the visibly
Also of interest are the three posture angles, J1, J2, and J3.
“red” and “white” areas of the fingernail differ principally in
However, in practice J1 does not affect the fingernail color; also
their green component, which can be isolated (by software or
J3 is coupled to J2 as long as the finger is not in contact with the
hardware filters) to generate a high-contrast grayscale image.
surface when bending occurs. Therefore, in this paper the only

posture angle of interest is J
1.0 MegaPixel CCD Digital Camera,
2, hereafter referred to as ?.
There are many other variables that may affect blood
10x Optical Zoom, Monochrome Mode
perfusion in the fingertip such as body or finger temperature,
light source:
hand elevation, or any other factors affecting blood pressure or
• 25 Watts for soft illumination
cardiovascular activity. In this paper, experiments are
• positioned behind nail to
prevent glare from nail surface
conducted with the body at rest in a constant temperature
environment with the hand at constant desk level in order to
Green filter
minimize the effects of any such variables.
and close-up lens
no glare

J
y
3
J2
glare
z
x
rubber pad
3-axis force-
z
sensing platform
x
y

J
• Norma
m l
a fo
f rce
c : F
Figure 2. Experimental apparatus. A CCD camera
1
z
images the fingernail coloration while the finger is
• Shea
e r fo
f rces
e : F , F
x
y
pressed against the force platform with specified force.
• Postu
t re an
a gl
g es:
s J , J , J

1
2
3

All 15 subjects were photographed in eight different poses.
Figure 1. Finger variables of interest. The forces are
The first image is the nominal coloration of the fingernail when
defined to be positive when the finger is pressed against
no force is applied and the finger is held straight out. For the
a surface in the positive x, y, and z directions as shown.
second image, a normal force of -3 N is applied. Then lateral
The finger posture is represented by the angle of the
and longitudinal shear forces of negative and positive 2 N are
knuckle (MP) joint, J1, the middle (PIP) joint, J2, and the
added. Finally, images are taken of the finger fully extended
distal (DIP) joint, J3. The angles are defined to be positive
and fully flexed with no force applied. These poses are near the
for flexion and negative for extension.
limits of steady-state force/posture that can be comfortably

maintained (on the order of a minute), and are also near the
In this study, digital images were collected for 15 human
saturation limits of the fingernail coloration effect [10] where
subjects: 10 male and 5 female. For purposes of categorizing
the coloration is insensitive to small fluctuations in
skin color, 7 subjects were white, 7 were Asian, and 1 was
force/posture due to unsteady control by the subjects.
black. In order to uniformly photograph the subjects, an
In order to compile the images for all 15 subjects into
apparatus was designed as shown in Figure 2. A 3-axis force-
meaningful results for any particular pose, the intensity values
sensing platform with rubber surfaces is placed within an
are averaged across all the subjects, resulting in an “average”
enclosed chamber to block ambient lighting. A 1.0 MegaPixel
fingernail image depicting the average pattern of coloration.
CCD digital camera with 10x optical zoom is mounted above
First, the sizes of the images are normalized according to the
the force-sensing platform in order to image the fingernail as
length and width of the fingernail. The length is defined as the
the finger is pressed against the platform with various forces or
maximum distance in the y-direction from the eponychium
held with various postures. The human subjects use visual
(where the nail emerges from the proximal nail fold) to the
feedback from the force sensor to maintain constant desired
hyponychium (where the nail separates from the distal bed),
force. The two key features of the apparatus are the lighting and
and the width is defined as the maximum distance in the x-
the filtering. First, in order to prevent glare from ruining the
direction between the lateral nail folds. Then the images are
image, the light must be placed directly behind the fingernail,
uniformly cropped about the center of the nail to 125% of the
over the hand, as shown in the figure, so that the curvature of
normalized nail size, leaving a small area visible around the
the nail will reflect the light out away from the lens. In this
nail. The resulting images are all 400x400 pixels, with a
situation, the only light that reaches the lens is the light that
320x320 pixel fingernail in the center. Finally, the intensity
penetrates the nail bed and is diffusively reflected back out.
value for each pixel is averaged across the 15 subjects.

2

---

Haptics-e, Vol. 4, No. 3, July 21, 2006
http://www.haptics-e.org
corresponding to the red and white areas, respectively, of the
fingernail. The lunula (visible portion of the nail matrix)
appears as a grayish-white zone that varies in size and thus is
hyponychium
not useful for establishing commonality between subjects.
normalized
lateral nail folds
dimensions
3. ANALYSIS
? 0 N ?
eponychium
In order to evaluate whether these average patterns are
F = ? 0 N ?
?
?
distinct in a statistically significant sense, all of the images for
F = 0
?-3 N?
?
?
each pose should be correlated to a standard set of images that
nominal
-F
? = 0°
z
? = 0°
is representative of the common pattern for each pose. The
logical choice for such a standard set of images is the set of
average images in Figure 3. For example, if the images of all
the subjects for +Fx correlate best in a statistically significant
sense to the average image for +Fx compared to the average
images for the other poses, then it can be concluded that the
response to +Fx is common for all subjects and distinct from all
?-2 N?
?+2 N?
other poses. Moreover, if this is true for all poses, then it will
have been demonstrated that there is a set of common distinct
F = ? 0 N ?
F = ? 0 N ?
?
?
?
?
patterns (namely the set of images in Figure 3) that is
?-3 N?
? -3 N ?
?
?
?
?
representative of all subjects.
-Fx
+F

x
? = 0°
? = 0°
50
100
150
n 200
250
? 0 N ?
? 0 N ?
300
?
?
F = ?-2 N?
?
?
F = +2 N
?
?
350
?-3 N?
?
?
? -3 N ?
?
?
400
100
200
300
400
-F
+F
m
y
? = 0°
y
? = 0°

Figure 4. Correlation zone. The dashed line indicates
the normalized area of the fingernail. The white area
represents the subset that will be used for correlation.

Towards this end, a zone of correlation, Z, is first defined,
which includes the normalized area of the fingernail with the
lunula
lunular portion removed, as in Figure 4.
2
2
2
?
if [(m ? 200) + (n ? 200) < 150 ]
?1
F = 0
F = 0
2
2
2
Z
=
(1)
mn
?
and [(m ? 450) + (n ? 200) > 160 ]
extension
? = -10°
flexion
? = 90°
?

?0
otherwise
Figure 3. Average fingernail coloration for 15 human
subjects for various poses. Images from the 15
The number of pixels in this area is given by:
subjects were normalized to the same size and averaged
pixel by pixel. 5x contrast is applied. Subjects were
400 400
N
= ??Z = 52593
instructed to apply the same constant forces using visual

(2)
pix
mn
feedback from the force sensor.
m 1
= n 1
=

Let I be the set of all images from all Ns=15 subjects for all
The results are shown in Figure 3. As evidenced in the figure,
Np=8 poses. The image intensity for each ith subject and jth pose
the various poses result in visibly different average patterns of
is normalized according to the average and standard deviation
coloration with well-defined dark and white zones,
of the intensities in the correlation zone:

3

---

Haptics-e, Vol. 4, No. 3, July 21, 2006
http://www.haptics-e.org
I
? I
C ? C
ijmn
ij
I? =
Z

jj
jk
t =
(8)
ijmn
mn
?
jk
2
2
ij
S
S
jj
jk
+
i = 1,2...N
N
N
400 400
1
s
s
s

I =
?? I Z
(3)
ij
( ijmn mn)
N
In order to find the confidence levels for statistical
pix m 1
= n 1
=
j = 1,2,...N
significance, the degrees of freedom are computed [12]:
p
2
400 400
2
2
??(
? S
S ?
I
? I
jj
jk
?
+
ijmn
ij )
?
N
N
(N ?1) S + S
? s
s ?
s
( jj jk)2
2
2
m 1
= n 1
?
=
=
ij
? =
=
N
?1

(9)
jk
2
2
4
4
pix
2
2
?
?
?
?
S + S
S
S
jj
jk
jj
jk
?
?
?
?
Let A be the set of average images in Figure 3:
N
N
? s ?
? s ?
+
1 Ns
N ?1
N ?1
s
s

A
=
?I
j = 1,2...N (4)
jmn
ijmn
p
Ns i 1=
The minimum degree of freedom for all jk pairs is ? = 15,
which gives a conservative 95% confidence level of t < 1.75
A is also normalized by the averages and standard deviations
and a conservative 99% confidence level of t < 2.60. For those
of the intensities in the correlation zone:
pairs with greater degrees of freedom, the confidence levels
A
? A
shift to slightly more generous values of 1.65 and 2.33.
jmn
j
A? =
Z
20
jmn
mn
? Aj
99% confidence level
18
95% confidence level
400 400
1
avg nom

A =
?? A Z
(5)
j
( jmn mn)
16
avg -F
N
z
pix m 1
= n 1
=
avg -F
14
k
x
400 400
avg +F
??(
x
A
? A
12
Pose,
avg -F
jmn
j )
y
m 1
= n 1
?
=
=
avg +Fy
Aj
N
?1
t jk 10
avg ext
pix
8
avg flex
The correlations for each ith subject between each jth pose and
each kth average image is now computed:
6
i = 1,2,...N
4
400 400
1
s

C =
??I? A?
j = 1,2,...N (6)
2
ijk
ijmn
kmn
p
N pix m 1= n 1=
k = 1,2,...N
0
p
nom
-F
-F
+F
-F
+F
ext
flex
z
x
x
y
y
Pose, j

Then the average and standard deviations of the correlations
Figure 5. Statistical significance of coloration
across all the subjects is computed:
patterns. The t-value was computed for the correlations
of each pose with true average image vs. the correlation
1 Ns
C =
?C
with each of the other average images.
jk
ijk
N

s i 1
=
Figure 5 plots the t-values for the correlations and the

Ns
?(
(7)
minimum (most conservative) confidence levels. With the
C ? C
ijk
jk )
exception of -Fz and -Fy, all of the force poses correlate best to
i 1
S
=
=
the average image of the true pose with better than 99%
jk
N ?1
s
confidence. The two posture poses correlate best to themselves
with better than 95% confidence. The patterns for -Fz and -Fy
To evaluate statistical significance, the t-distribution is used,
are on average too similar to each other, as one might suspect
where the t-values are given by [12]:
from a visible inspection of the images in Figure 3. Evidently,

4

---

Haptics-e, Vol. 4, No. 3, July 21, 2006
http://www.haptics-e.org
the addition of shear force in the negative y-direction does not
The t-values for the significances can then be computed
change the coloration beyond what already occurs when a
using equations of the same form as (8) and (9) and are plotted
normal force is exerted downward in the z-direction. Also, the
in Figure 7 for the poses of interest. The nominal pose is now
pattern for the nominal coloration of the nail is not significantly
distinguishable from extension in terms of average intensity in
distinguishable from that of extension. However in this case,
the proximal zone with greater than 99% confidence.
returning to the images in Figure 3, one may note that while the
patterns for these two poses are similar, the brightness or
3
intensity of the white zone appears to be much greater for
ies
(proximal zone)
2.5
extension. As a secondary means of distinguishing poses with
t
e
nsit
similar patterns, the statistical significance of the average
in
2
intensities in the white zones may be investigated. Two new
zones corresponding to the distal and proximal regions of the
e
r
age 1.5
(distal zone)
fingernail are defined, illustrated in Figure 6.
r
av
2
2
2
if [(m ? 250) + (n ? 200) > 160 ]
1
s
fo
99% confidence level
ue
95% confidence level

Zd
= Z
and Zp = 0 (10)
mn
mn
mn
0.5
90% confidence level
else Zd
= 0 and Zp = Z
t
-
v
al
mn
mn
mn
0
nom vs. ext -F vs. -F
z
y

distal zone
proximal zone
Figure 7. Statistical significance of average
intensities. The t-values for the average intensity
50
50
differences were computed for poses which did not have
100
100
significantly different patterns. ? = 27 for both.
150
150

n 200
n 200
For comparison, -Fz and -Fy are barely distinguishable with
90% confidence based on average intensities in the distal zone.
250
250
If the -Fy pose is removed from consideration, then it can be
300
300
concluded that there are at least seven different poses that result
350
350
in observable coloration patterns that are common to all
400
400
subjects and distinct from each other with at least 95%
50
100
150
200
250
300
350
400
50
100
150
200
250
300
350
400
m
m

confidence (>99% in most cases).
Figure 6. Distal and proximal zones of fingernail
images. The white areas represent the subset areas of
4. DISCUSSION
the normalized fingernail used to average intensities.

The preceding results serve as strong evidence that there is a
The average intensity for each ith subject for each jth pose
set of characteristic color patterns that are representative of the
within these two zones is computed:
population in general and are distinct for at least seven different
poses of force and posture. In conjunction with continuing
400 400
??(
work [13][14], which show that individual variability and time
I
Zd
?? I Zp
ijmn
mn )
400 400 ( ijmn mn)
dependence are not obstacles to distinguishing individual
m 1
= n 1
=
m 1
= n 1
D =
, P
=
=
(11)
responses, this validates the potential of the fingernail sensor as
ij
400 400
ij
??(Zd
?? Zp
a useful means for facilitating human-computer interaction
mn )
400 400 ( mn)
(HCI) for people in general. At the very least, any user should
m 1
= n 1
=
m 1
= n 1
=
be able to give six discrete commands (nominal pose would be
and the average and standard deviation across the subjects:
a 7th null command) to the computer using the fingernail
sensor, which for example, could be used to move a pointer,
1 Ns
1 Ns
activate scroll bars, or navigate through menus. Results from
D =
?D
P =
?P
j
ij
j
ij
N
previous work show that by calibrating the fingernail sensor to
=
N
s i 1
s i 1
=
the individual user, we can go beyond discrete classification
N
and estimate a continuous range of values of x, y, and z forces,
s
?(D ? D
?
(12)
P ? P
as well as posture angle [10]. Because normal force and
ij
j )
Ns ( ij j)
i 1
=
i 1
SD =
SP
=
=
negative longitudinal shear do not result in a statistically
j
N ?1
j
N ?1
significant difference in color, it may not be possible in general
s
s
to independently recognize or estimate values of -Fy. This

5

---

Haptics-e, Vol. 4, No. 3, July 21, 2006
http://www.haptics-e.org
would place some constraints on the applications of the sensors
REFERENCES
to monitoring arbitrary grasping states of the fingers. However
it would still allow for a variety of HCI applications where the
[1]
Webster, J., ed., Tactile Sensors for Robotics and
mapping from sensor output to computer action could be
Medicine. New York: Wiley, 1988.
tailored to available range of outputs. For example, horizontal
[2]
Pawluk, D.T.V. and Howe, R.D., “Dynamic Lumped
motion of a computer pointer could be controlled by the entire
Element Response of the Human Fingerpad,” J.
range of x-forces, while the vertical motion is controlled by the
Biomech. Eng., vol. 121, no. 2, pp. 178-183, 1999.
positive range of y-forces or the negative range of z-forces, or
[3]
Pawluk, D.T.V. and Howe, R.D., “Dynamic Contact of
some combination thereof.
the Human Fingerpad Against a Flat Surface,” J.
Furthermore, the results of this paper in particular justify an
Biomech. Eng., vol. 121, no. 6, pp. 605-611, 1999.
effort to create a unified model of the color change mechanism
[4]
Serina, E.R., Mote, C.D. Jr., and Rempel, D., “Force
that is applicable to people of various skin colors and fingernail
Response of the Fingertip Pulp to Repeated
sizes. This could be as simple as creating a set of average
Compression—Effects of Loading Rate, Loading Angle,
patterns that would allow the fingernail sensor to work on any
and Anthropometry,” J. Biomechanics, vol. 30, no. 10,
person without calibration. Or it could be as complex as
pp. 1035-1040, 1997.
creating a physically-based model using the common anatomy
[5]
Gulati, R.J., and Srinivasan, M.A., “Human Fingerpad
and physiology of the fingertip that would explain how the
under Indentation I: Static and Dynamic Force
various forces and postures create stress/strain fields within the
Response,” ASME Bioengineering Conference, BED-
fingertip. Preliminary research has begun towards creating such
vol. 29, pp. 261-262, 1995.
a unified physically-based model [15]. We anticipate that this
[6]
Singh, U. and Fearing, R.S., “Tactile After-Images from
model will help to understand what is mechanically transpiring
Static Contact,” Proc. ASME Dynamic Systems and
inside the finger when various forces and posture are applied
Control Division, vol. 64, pp. 163-170, 1998.
during various states of grasping and manipulation.
[7]
Han, H.-Y., Shimada, A., and Kawamura, S., “Analysis
of Friction on Human Fingers and Design of Artificial
Fingers,” Proc. IEEE Int. Conf. on Robotics and
5. CONCLUSIONS
Automation, pp. 3061-3066, 1996.
In conclusion, this paper presented an analysis of the
[8]
Scher, R.K. and Daniel, C.R. III, eds., Nails: Therapy-
photographically measured coloration patterns of the human
Diagnosis-Surgery. Philadelphia: W.B. Saunders, 2nd
fingernail for 15 subjects when various forces and postures are
ed., 1997.
applied to the finger. The analysis indicates there are at least
[9]
Tubiana, R.T., ed., The Hand. Philadelphia: W.B.
seven different states of fingertip force and finger posture that
Saunders, vol. 1, 1981.
result in distinct fingernail coloration patterns that are common
[10] Mascaro, S. and Asada, H., “Measurement of Finger
to all people according to a statistically significant measure.
Posture and Three-Axis Fingertip Touch Force Using
In future work, we propose to develop a unified, physically-
Fingernail Sensors,” IEEE Transactions on Robotics and
based model of the finger that explains the resulting blood
Automation, vol. 20, no. 1, pp. 26-35, 2004.
perfusion patterns that occur in the fingernail bed when various
[11] Mascaro, S. and Asada, H., “Photoplethysmograph
forces and changes in posture are applied to the fingertip. Such
Fingernail Sensors for Measuring Finger Forces Without
a model will help to understand how various combinations of
Haptic Obstruction,” IEEE Transactions on Robotics
forces and postures create regions of compression and tension
and Automation, vol. 17, no. 5, pp. 698-708, 2001,
within the tissue under the fingernail that result in the visible
[12] Weiss, N.A., and Hassett, M.J. Introductory Statistics.
patterns of reddening and whitening zones that are common to
2nd ed. Reading, Mass.: Addison-Wesley, 1987.
all people. This will be useful for designing future generations
[13] Sun, Y., Hollerbach, J. and Mascaro, S., “Measuring
of fingernail sensors as well as for understanding human
Fingertip Forces by Imaging the Fingernail,” Proc. 14th
grasping and manipulation. Eventually, the effect of additional
Int. Symp. Haptic Interfaces for Virtual Environment and
variables should be investigated, such as roll and pitch angles
Teleoperator Systems, pp. 125-131, 2006.
of the finger with respect to the surface, as well as variables
[14] Sun, Y., Hollerbach, J. and Mascaro, S., “Dynamic
that influence the cardiovascular state (e.g. temperature, hand
Features and Prediction Model for Imaging Fingernail to
elevation, heart rate, etc).
Measure Finger Forces,” Proc. IEEE Int. Conf. Robotics
and Automation, pp. 2813-2818, 2006.
ACKNOWLEDGEMENT
[15] Mascaro, S. and Asada, H., “Understanding of
Fingernail-Bone Interaction and Fingertip
This work was funded in part by the National Science
Hemodynamics for Fingernail Sensor Design,” Proc.
Foundation, Grant: NSF IRI-0097700. Experiments in this
10th Int. Symp. Haptic Interfaces for Virtual Environment
work received IRB approval from the MIT Committee on the
and Teleoperator Systems, pp. 106-113, 2002.
Use of Humans as Experimental Subjects (COUHES #2738).

6

---