The Haptic Scissors: Cutting in Virtual Environments

A. M. Okamura, R. J. Webster III, J. T. Nolin, K. W. Johnson, and H. Jafry

Department of Mechanical Engineering
The Johns Hopkins University
Baltimore, MD 21218
{aokamura, robert.webster, jnolin, kate}@jhu.edu, hrjafry@mtholyoke.edu

Abstract limited our previous display to a single closing of the

scissor blades.
The “Haptic Scissors” is a device that creates the Cutting has also been explored in other previous work.
sensation of cutting in virtual environments. The scissors The data used in this paper was originally gathered by
have two degrees of freedom of motion and force Greenish and Hayward. However, a Fourier analysis did
feedback, one for cutting (single blade rotation) and one not reveal a clear relationship between force and position,
for translation. An algorithm was developed to velocity, or other dynamic interaction properties from the
simultaneously display translational and cutting forces data acquired [3,4]. In [7], a haptic simulation was created
for a realistic cutting simulation. In previous work, we from data acquired during cutting with a single blade.
used filtered data from cutting biological tissues to create Resolved-force haptic devices, such as the Phantom from
“haptic recordings” of the cutting experience. Here, we SensAble Technologies (Woburn, MA), have been used to
consider two cutting models: one based on real tissue display external cutting forces of a single blade in surgical
data and one that is analytical. The model based on real procedures (e.g., [1,5]). In addition, cutting with scissors
tissue is a segmented linear empirical model of the has been modeled, without the display of internal cutting
original data. Experimental results show that users forces, in [6,8].
cannot differentiate between these models and the haptic
recordings created earlier. The analytical model uses a 2. The Haptic Scissors
combination of friction, assumed material properties, and
user motion (position and velocity) to determine the 2.1 Design
displayed cutting forces.
The current design of the haptic scissors is shown in
1. Introduction Figure 1. The scissors have two degrees-of-freedom, one
for cutting (rotational) forces, and one for translational
Current simulations of biological tissues in surgical forces. The rotational degree-of-freedom has an angular
procedures are mostly, if not completely, visual with resolution of 0.056° and a maximum torque output of 280
regard to cutting. This work is motivated by the need for mNm, which corresponds to a force output of 7.58 N at
haptic display not only in “poking” and “pulling” surgical
tasks, but also in grasping and cutting. Simulations of
tissue cutting will likely be more realistic than cadaver
tissues, which have significantly different mechanical
properties from living tissues. These simulations will also
be more practical than physical phantoms, which cannot
be reused. Also, haptic rendering of tissue cutting has the
potential to reduce the need to sacrifice animals for
surgical training and dissection.
The work in this paper is an extension of our previous
work in the development of a single degree-of-freedom
haptic scissor [2]. This original haptic scissor used “haptic
recordings,” look-up tables of position vs. force values
from real biological tissue cutting data, to display the
forces. However, this system was limited in realism for
two major reasons: (1) the scissors did not translate during
cutting, and (2) the haptic recordings did not take into
account the context of user motion. These two problems Figure 1. The two-degree-of-freedom haptic scissors.
 the cutting process.
The first step is recognition of the direction of motion
of the blades: opening or closing. This is determined by
the sign of the angular velocity of the scissor blades.
Because of noise that would have been seen in
differentiating the position obtained from optical
encoders, position data was passed through a first-order
filter with a corner frequency of 10 rad/sec before
differentiation. Even with filtering, however, the system
still exhibited ringing or vibration effects at low velocity.
This was because the force output was sufficient to move
the scissors in the opposite direction, changing the sign of
the velocity and consequently the direction of the output
force in a cyclical manner. The issue was overcome and
the system stabilized by creating a small linear force
rampdown band around zero velocity. Within this band,
the cutting force was multiplied by a scale factor
proportional to the velocity. This scale factor decreased
Figure 2. The haptic scissors and graphic display. as zero velocity was approached from either positive or
negative directions, and became identically zero when
the handle. The translational degree-of-freedom has a velocity was zero.
resolution of 0.0152 mm and maximum force output of The second step is collision detection, performed to
159 N. (However, the cables are likely to slip before the determine whether the cutting point on the scissors is
maximum force is reached.) intersecting the material. Figure 3 is a diagram of the
A Maxon RE 25mm diameter/20-Watt motor scissors showing the cutting point, xc, with respect to the
(rotational) and a Maxon RE 35mm diameter/90-Watt pivot point of the scissors.
motor (translational) were used for actuation, both with
HP HEDS 5540 encoders for sensing. A PWM amplifier t (θ ) ,
xc = (1)
(25A-series, Model 12A8) from Advanced Motion sin(θ 2 )
Control was used for each motor. A Servo To Go card
(Model 2) was used to read the encoders, and a
where t(θ) is half of the width of the scissor blade (which
Measurement Computing card (CIO-DAS 1600/12) was
used for voltage/force output. The control computer was a changes along the length of the blade) and θ is the angle
Pentium III 800 MHz running Windows 98, and the between the scissor blades. As an approximation, we
development environment was Microsoft Visual C++. The assume that the value of t changes linearly from tmax at the
graphic and haptic update rates were 50 Hz and 1000 Hz, pivot to tmin at the blade tips. Thus, t is calculated by:
respectively. The graphic display used an OpenGL
NURBS scissor model, as shown in Figure 2.  tan (θ 2 ) 
, (2)
t (θ ) = t max  t max − tmin 
 tan (θ 2 ) + lt 
2.2 Haptic Rendering of Scissor Translation
where lt, the length from the pivot to the tip, is shown in
Simultaneous rendering of cutting (rotational) and Figure 3.
translational forces is necessary for realistic simulation of If the edge of the uncut material is at a position xm with

Figure 3. Diagram of scissors and material to be cut, with constants and variables used in haptic rendering.
respect to ground, a collision is detected if (xc + xs) > xm. 30
Empty Scissors
30
Liver

When a collision is detected, the cutting force (generated

20 original data 20
by torque at the pivot), fc, applied to the user is

force (N)

force (N)
simulation data
determined by one of the three cutting algorithms 10 10

presented in Section 3. Although data acquired from 0 0

instrumented scissors handles show the forces displayed to
-10 -10
the user, they do not provide information about the state 0 10 20 30 0 10 20 30
angle (deg) angle (deg)
of the material. Thus, we define a cutting threshold, fmax,
above which the material begins to cut away. When the Skin Tendon
cutting force is below the threshold, the uncut material 30 30

edge position, xm, remains unchanged. If the threshold is 20 20

force (N)

force (N)
exceeded, the material edge position moves, maintaining 10 10
xm = xc + xs. Translational forces are displayed as
0 0

f t = kt ( xm − xc − xs ) , (3) -10
0 10 20 30
-10
0 10 20 30
angle (deg) angle (deg)

where kt is the predefined stiffness of the material in the

translational direction. Figure 4. Data used for haptic recordings. The original
data was obtained from instrumented Metzenbaum
scissors while cutting rat tissues. The simulation data uses
3. Cutting Algorithms filtered data from one closing of the scissors.

To analyze the sensation of cutting on the haptic

scissors, we used tissue-cutting data acquired as described switching from opening to closing, (3) closing scissors,
in [3,4]. We used data from cutting nothing (empty and (4) switching from closing to opening. The lower
scissors) and three rat tissues (liver, skin, and tendon) with section of the plots, where the force appears to remain
a pair of Metzenbaum scissors. We have developed three relatively constant as the angle increases, corresponds to
different methods for displaying the cutting data: haptic the opening of the scissors. The upper section, where the
recordings, segmented empirical models, and analytical force fluctuates more, corresponds to the closing of the
models. scissors. At the left and right of the plots, the forces
change dramatically at a single angle; these regions result
3.1 Haptic Recordings from friction when the scissors are in the stiction state
while changing between opening and closing.
Haptic recordings are implemented by replaying an Using these haptic recordings alone, it is difficult to
angle to force mapping of the acquired data on the haptic create a realistic simulation; we could not guarantee that
scissors. The primary advantage of haptic recordings is the users would maintain the scissor angles used in data
their computational efficiency (a simple lookup table can acquisition. To provide forces for complete closure of the
be used). A disadvantage is that contextual information scissors, forces were smoothed and the force trend at the
(e.g., velocity and grip force) that may affect the force minimum angle was maintained to zero degrees. In
output is not considered. In contrast, analytical and addition, the tendon forces were scaled down to match the
empirical models are difficult to create, due to the real rat tissues available for comparison in [2]. Details of
complex material interactions that give rise to forces the haptic recording method for single degree-of-freedom
during the cutting of biological tissues. However, they scissors (cutting/rotation only) are provided in [2]. That
provide a much better fundamental understanding of the method has been improved as follows.
scissor cutting activity. Current analytical models of tissue A simulation using pure haptic recordings creates
deformation and fracture are nascent, and have not been sharp discontinuities when the scissors move between
verified by experimental data. Similarly, accurate opening and closing. To create a smooth display of cutting
empirically based models would require a significant data forces from the initial point of contact of the scissors with
acquisition effort in order to characterize the many the material until the scissors lose contact with the
parameters affecting cutting forces. material, a force scaling method was used. Consider the
To create the haptic recordings, the data was first force displayed to the user at the handle, fh. In a pure
segmented and filtered. In each set of the data shown in haptic recording, this force is calculated from a lookup
Figure 4, there are several cuts, resulting in overlapping table fopen(θ) or fclose(θ), for opening and closing of the
lines. When surgeons cut, a repetitive motion is used, and scissors, respectively, and linear interpolation is used to
the scissors are never fully closed. The data loops can be find the forces corresponding to intermediate values of θ :
segmented into four phases: (1) opening scissors, (2)
 4 To create segmented empirical models for each kind of
original data
scaled data
tissue as well as the empty scissors, the following
algorithm was used. First, the data from each complete
3
cycle of opening and closing of the scissors was averaged
with the other cycles to create an average data run. A user-
2 defined piecewise linear fit was applied to this average
Force (N)

using a software program developed for the purpose. This

program plots the cutting data for the user and allows him
1 or her to design a piecewise linear model by clicking on
θswitch the plot at various points. The resulting model is then
0
superimposed on the original data plot, allowing the user
to see how closely the model matches the data. When a
4 6 8 10 12 14 16 qualitative visual inspection indicates a good match, the
Scissor Angle (degrees)
resulting function is displayed on the haptic scissors.
Figure 5. Illustration of the force scaling technique for To assess the success of this modeling method, an
smooth display of haptic recordings. experiment was conducted using subjects who had
experience cutting the same group of real rat tissues with
Metzenbaum scissors in [2]. Each of the three subjects
 f open (θ ) for θ& > 0 was asked to compare the feel of a haptic recording to an
 (4) empirical model for each type of tissue. The order of
f h =  f close (θ ) for θ& < 0
 0 haptic display presentation (model or recording) was
 for θ& = 0
random, as well as the tissue order. Subjects were
blindfolded so that the computer graphics and appearance
Consider the quantity ∆θ = |θ − θswitch|, where θ is the of the haptic scissors did not affect the results.
current angle of the handle and θswitch is the angle at which For each type of tissue, the subjects were told what
the user changes from positive to negative velocity or vice kind of tissue they were cutting and asked to determine if
versa. A constant scale factor ks is chosen so that when the “sample A” or “sample B” felt more like cutting the real
scissors have just changed direction (at angle θswitch), ∆θ ks tissue. The subject was allowed to freely manipulate the
= 0, and when the normal look-up table force is desired, scissors. That is, they could open and close the scissors as
∆θ ks = 1. Then the force output to the user at the handle many times as they liked, at whatever speed they desired,
is changing directions at whatever angle they preferred, etc.
 f (θ )∆θ k s for 0 ≤ ∆θ k s ≤ 1 , They were also allowed to switch back and forth between
fh =  (5) the two samples as many times as they felt necessary. The
 f (θ ) for ∆ θ k s > 1 results are summarized in Table 1.
As can be seen, aside from the Blank case, there was
where f(θ) is the lookup table value of force at the current no tissue on which every subject agreed. Additionally,
angle θ. Figure 5 shows how this scaling method affects each subject said that they could feel extremely little if
the haptic recordings. any difference between the haptic recordings and the
models, and had a very difficult time making up their
3.2 Segmented Empirical Models minds. In the Blank case, it is probable that users expect
to feel linear forces (friction) when they open and close
The segmented empirical model is an attempt to empty scissors, and that the haptic recordings were
mathematically model the data collected by cutting real influenced by noise and/or other factors in the data
tissues using a piecewise linear function. The goal of the
model and associated experiment is to ascertain exactly Table 1. Subject responses comparing the realism of the
how complex a haptic representation of the scissor cutting haptic recordings and the segmented linear model. The
forces must be. The advantage of this method is its display identified as most realistic is listed for each
simplicity and adaptability. subject and tissue type.
As can be seen from Figure 6, raw data of scissor
forces is not easily characterized by a mathematical
Tissue Subject 1 Subject 2 Subject 3
function. It is also not clear exactly how accurately a
Type
“good” model must fit the data. The criterion we selected
Tendon Recording Recording Model
for this was qualitative human perception. That is, if
human users could not distinguish the difference through Skin Recording Model Recording
the feel of the haptic device, the model was considered to Liver Recording Model Recording
be acceptable. Empty Model Model Model
 30 30
Empty Scissors Liver

20 20
original data
force (N)

force (N)
simulation data
10 10

0 0

-10 -10
0 10 20 30 0 10 20 30
angle (deg) angle (deg)

30 30
Skin Tendon

20 20

force (N)
force (N)

10 10

0 0

-10 -10
0 10 20 30 0 10 20 30
angle (deg) angle (deg)

Figure 6. Data used for the piecewise linear empirical models.

collection process (e.g., forces not directly in the plane of are assumed to be a summation of forces from friction at
the scissors’ motion generated inadvertently by the hand the scissor pivot, friction at the scissor blades, and cutting
of the experimenter) that the empirical model does not force at the cutting point xc.
include. The piecewise linear models used for this The effect of torsional friction at the pivot point is
experiment are those shown in Figure 6, demonstrating computed from a simple moment arm:
that a small number of line segments (2 to 7) is generally τ pf
sufficient. f h − pf = , (6)
Figure 6 shows the data taken from cutting real tissues lh
overlaid with the scissors closing empirical model. Note where fh-pf is the force felt at the handle due to pivot
the visual similarity between these plots and the haptic friction, τpf is the torsional friction at the pivot, and lh is
recordings shown in Figure 4. For the scissors opening the length of the scissor handle as shown in Figure 3.
segment of each loop a frictional simulation model of Blade friction arises from a small contact area between
scissors opening was displayed to the user. This model the two blades. The scissor blades are shaped so that this
was equal to -½ the model for Empty Scissors (pictured area, approximated as a point, is the only source of blade
above). All users agreed that this felt realistic. friction. The friction at the handle is
f bf xc
3.3 Analytical Models f h −bf = , (7)
lh
Preliminary work on analytical modeling has been where fh-bf is the force felt at the handle due to pivot
performed to characterize the effect of the changing friction, fbf is the friction at the blade contact point, and xc
cutting point on forces sensed at the handle. The model is the cutting point, as shown in Figure 3.
presented here is based on the assumption that the force The force necessary to cut a homogeneous material is
required to cut the material (at the exact location of the assumed to be a constant, fcf. Thus, the force felt at the
cut) is constant. Thus, it is a simplified model that does handle is
not take into account the material variations present in
biological tissues. The forces felt by the user at the handle
 2.5 analytical model. This would certainly yield insights for
basic research in tissue modeling, and robot-controlled
2 cutting. In future work, more cutting data must be
obtained. This data should be acquired by a system with
controlled position, velocity, and grip force. Further
force (N)

1.5 improvements on the scissors interface itself would

include adding tilting (pitch) and steering (yaw) degrees
1 of freedom to the scissors as a whole so that they could be
rocked as a unit up and down and turned as cutting
progresses.
0.5

Acknowledgements
0
0 20 40 60 80
theta (deg) The authors acknowledge Paul Stemniski and Vanessa
Chial for their work in the design of the haptic scissors.
Figure 7. Handle forces versus scissor angle for a This material is based on work supported by the National
constant velocity scissor closing (blank). Science Foundation (grants IIS0136344 and
EEC9731478), a Howard Hughes Fellowship, and a
f cf xc Department of Defense Fellowship.
f h −cf = . (8)
lh References

A summation of these three forces provides the force [1] D. Biesler and M. H. Gross, “Interactive Simulation of
at the handle: Surgical Cuts,” Proceedings of Pacific Graphics 2000,
fh = fh-pf + fh-bf + fh-cf (9) IEEE Computer Society Press, pp. 116-125, 2000.
[2] V. Chial, S. Greenish and A. M. Okamura, “On the
A plot of the handle forces calculated in this way is Display of Haptic Recordings for Cutting Biological
shown in Figure 7. The discontinuities in the plot result Tissues,” 10th International Symposium on Haptic
from the fact that the scissor blades do not actually cut at Interfaces for Virtual Environment and Teleoperator
small angles, and at large angles the blades are no longer Systems, 2002, pp. 80-87.
in contact. The analytical model does not closely match [3] S. Greenish, “Acquisition and Analysis of Cutting
the recorded cutting data because several aspects, Forces of Surgical Instruments for Haptic Simulation,”
including the effects of user grip force, inhomogeneous Master’s Thesis, Department of Electrical and
tissue properties, and elastic forces in the tissue, are not Computer Engineering, McGill University, 1998.
modeled. However, it does capture the rise in forces at [4] S. Greenish, V. Hayward, T. Steffen, V. Chial, and A.
small angles, which can be seen in the blank data from M. Okamura, “Measurement, Analysis and Display of
Figure 6. Haptic Signals During Surgical Cutting,” Presence. (in
press for December 2002)
5. Conclusions and Future Work [5] K. Hirota, A. Tanaka, and T. Kaneko, “Representation
of force in cutting operation,” Proceedings of IEEE
Adding translational motion/forces and context-based Virtual Reality, pp. 77, 1999
force feedback to the haptic scissors has greatly improved [6] C. B. Ho, M. A. Srinivasan, S. D. Small, and S. L.
the system by making the haptic interface feel much closer Dawson, “Force Interaction in Laparoscopic
to the experience of using real scissors. In addition, Simulation: Haptic Rendering of Soft Tissues,”
experiments comparing the segmented empirical models Medicine Meets Virtual Reality, 1998.
and haptic recordings verified that users of the haptic [7] M. Mahvash and V. Hayward, “Haptic Rendering of
scissors could not differentiate between a complex data Cutting: A Fracture Mechanics Approach,” Haptics-e,
recording and a simple piecewise linear model. This The Electronic Journal of Haptics Research
demonstrates that models of cutting forces need not be (www.haptics-e.org), Vol. 2, No. 3, November 20,
very precise to feel real. Thus, for practical engineering 2001
purposes, it is probably not necessary to formulate a [8] G. Picinbono, H. Delingette, and N. Ayache,
precise model for cutting forces at all. A general idea of “Nonlinear and anisotropic elastic soft tissue models
the proper shape of the curve for each type of tissue is all for medical simulation,” Proceedings of the IEEE
that is necessary for an adequate haptic rendering. International Conference on Robotics and
However, a complete understanding of cutting with Automation, vol. 2, pp. 1370-1375, 2001.
scissors requires the development of a more complex