Introduction
In the platform of next-generation flexible endoscopes such as the Master and Slave
Translumenal
Endoscopic Robot (MASTER) [1]
[2]
[3]
[4] and EndoSAMURAI [5]
[6]
[7]
[8], attention has been mainly focused on achieving remote forceps manipulation. Meanwhile,
manipulation of the flexible endoscope itself depends on traditional manipulation.
This dependence limits developments in this field and accomplishing treatment with
one endoscopist using intuitive manipulation is generally considered important. We
therefore developed the Endoscopic Operation Robot (EOR) to first achieve the necessary
conditions of remote manipulation and a platform for manipulation of the flexible
endoscope itself [9]
[10]
[11]
[12]. On the other hand, hepatic feedback reportedly is useful in robot-assisted medical
procedures [13]
[14]
[15]
[16]. Thus, the current EOR is the third generation, and includes haptic feedback (feelings
of manipulation) felt through the master unit and a bilateral haptic feedback function
that transmits the amount of force applied by the operator to the master unit to the
tip of the scope with equal force. In addition, manipulation of the entire scope can
be done with one hand [12].
Inclusion of haptic feedback makes the system larger and more complex and will be
a source of increased costs. If the utility of haptic feedback were found to be low,
its inclusion in the system, therefore, would not be warranted. In this study, we
developed a new program in which haptic feedback in EOR version 3 does not function
and investigated differences in manipulability with and without haptic feedback to
clarify the system’s utility.
Methods
EOR version 3 system
The EOR ver.3 incorporates haptic feedback to provide complete remote control of flexible
endoscope manipulation ([Fig. 1] and [Fig. 2]) [12]. The endoscope is an Olympus CF-240I (Tokyo, Japan), mounted on slave unit. Maximal
length of an Olympus CF-240I is 130 cm. The colonoscopic insertion length is the same
as a master device with the slave device to 60 cm. The colonoscopy training model
was added to these to conduct this study.
Fig. 1 Components of master unit of EOR ver. 3. a Knob-like rotating part (a) (rotating knob). b Mini joystick. c Rotary motor. d Torque sensor. e Load cell. f Circuit switch.
Fig. 2 The slave unit of Endoscopic Operation Robot (EOR) ver. 3. to which the colonoscope
(Olympus PCF-240; Tokyo, Japan) is attached.
The master unit of EOR ver. 3 ([Fig. 1]) consists of a knob-like rotating part (a) (rotating knob), a mini joystick (b),
a rotary motor (c), torque sensor (d), a load cell (e) and the circuit switch (f).
The master unit is an original device that enables four-axis movement of the flexible
endoscope with one hand to provide intuitive manipulation. In brief, an operating
knob equipped with a knob-like rotating part (a) (rotating knob) is installed on a
linear motor with a long axis of 60 cm. By operating the linear motor in the long-axis
direction with the rotating knob, the scope is inserted or retracted; by rotating
the rotating knob, the scope can be rotated. To enable force feedback in these two
axes, a load cell (e) is attached to the linear motor, and a rotary motor (c) and
torque sensor (d) are installed on the rotation knob. Up-down and left-right angulation
are performed with the thumb or index finger using a mini joystick (b) placed on top
of the torque sensor. Because the master unit is short (60 cm) compared to the scope
length of ≥ 1 meter, if the rotating knob reaches the edge of the master unit, target
insertion and retraction cannot be performed. In this case, the master-slave circuit
is first turned “off” using the circuit switch (f) on the left side of the joystick
for angulation, then the rotating knob is moved to the opposite edge as operation
only of the master unit. Later, the switch is turned “on” to connect the master-slave
circuit, and the target insertion-retraction can be performed.
Because the slave unit ([Fig. 2]) has bidirectional haptic feedback, its construction is similar to that of the master
unit. Instead of an operating knob, there is a rotating actuator that houses the endoscope
handle. The rotating actuator includes a system to operate up-down angulation and
left-right angulation and a system to operate the air supply/water supply buttons
and air suction button (operated by 2-foot switches).
A new program in which haptic feedback does not function was added so that operability
with and without haptic feedback could be compared. By checking the area of “Control”
in a personal computer monitor, the program of haptic feedback was carried out ([Fig. 3]).
Fig. 3 The personal computer monitor. a The setting with haptic feedback checks “Master Free” and “MSON”. b Without haptic feedback, “Force Feed Off,” “Force Rot Off,” and “MSON” are checked
in the area of “Control.”
Study design and protocol
A colonoscopy training model produced by KYOTO KAGAKU Co., LTD. (Kyoto, Japan) was
used. This model has six training patterns (beginner’s grade 1 – 3, intermediate grade
1 – 2, and higher grade). For this study, beginner’s grade 2 was used.
In this study, a scope was inserted up to the cecum in a colonoscopy training model
and the haptic sensations during insertion, time until insertion to the cecum, and
incidence of overstretching of the sigmoid colon during insertion were compared between
groups with and without haptic feedback. The robot operation monitor, including animation
of the colonoscopy training model, was recorded using a video recorder. Overstretching
of the sigmoid colon was confirmed by watching this recording. Overstretching of the
sigmoid colon was considered present if, during push insertion, the sigmoid colon
could be confirmed on the monitor to touch the right or transverse colon or abdominal
wall even once.
Haptic sensations were recorded every 0.2 seconds with master unit insertion-retraction
(push/pull; N), clockwise torque (N.m), and counterclockwise torque (N.m) as parameters.
Maximum and mean values for the different parameters were obtained first with each
insertion, after which the median values for these maximum and mean values were compared.
To achieve familiarity with manipulation of the EOR version 3 master unit, the endoscopists
initially practiced scope insertion to the cecum twice. They then performed the same
procedure five times with haptics, and the next five times without haptics. During
this time, data were recorded for evaluation.
The endoscopists had more than 2 years’ experience performing colonoscopy or hadg
performed more than 200 procedures.
Statistical analysis
Median and quartiles were calculated. P values were also computed by using Student’s t test. A probability less than 0.05 was considered to represent a significant difference
between the samples studied.
Results
Total colonoscopy using the training model was performed a total of 50 times (with
haptics (hap +) 25 times, without haptics (hap –) 25 times) by the five endoscopists.
The cecum was reached in 100 % of cases. The following results are shown as median
values. Insertion time was 70 seconds for hap + and 87 seconds for hap –. Maximum
pull was 5.235 N for hap + and 7.335 N for hap –, and mean pull was 0.939 N for hap +
and 1.158 N for hap –. Mean clockwise was 0.041 N.m for hap + and 0.072 N.m for hap –.
Maximum counterclockwise was 0.064 N.m for hap + and 0.156 N.m for hap – and mean
counterclockwise was 0.012 N.m for hap + and 0.029 N.m for hap –. Incidence of sigmoid
colon overstretching was 8 % in hap + and 32 % in hap –. Significant differences (P < 0.05) were seen in all seven parameters ([Table 1]).
Table 1
Force parameters compared between groups with and without hepatic feedback.
|
Hap (+)
|
Hap (–)
|
|
|
Parameter
|
Median
(Q25, Q75)
|
Median
(Q25, Q75)
|
P value
|
|
Max push (N)
|
17.90
(16.33, 24.08)
|
22.08
(19.49, 25.85)
|
0.12
|
|
Max pull (N)
|
5.27
(3.37, 7.99)
|
7.34
(5.51, 15.49)
|
< 0.05
|
|
Max clockwise (N.m)
|
0.163
(0.122, 0.305)
|
0.259
(0.202, 0.312)
|
0.38
|
|
Max counterclockwise (N.m)
|
0.064
(0.040, 0.134)
|
0.156
(0.051, 0.385)
|
< 0.05
|
|
Mean push (N)
|
4.07
(3.49, 4.52)
|
4.34
(3.79, 5.21)
|
0.21
|
|
Mean pull (N)
|
0.94
(0.68, 1.14)
|
1.16
(0.97, 1.63)
|
< 0.05
|
|
Mean clockwise (N.m)
|
0.041
(0.037, 0.049)
|
0.072
(0.047, 0.099)
|
< 0.05
|
|
Mean counterclockwise (N.m)
|
0.012
(0.009, 0.025)
|
0.029
(0.017, 0.128)
|
< 0.05
|
|
Examination time (min)
|
70
(63.5, 76.5)
|
87
(62.26, 119.5)
|
< 0.05
|
|
All insertion length (cm)
|
64.05
(63.75, 65.00)
|
64.10
(63.00, 65.25)
|
0.94
|
|
Overstretching (%)
|
8
|
32
|
< 0.05
|
Max, maximum; min, minutes.
Median (Q25, Q75), Q25, lower quartile (25 % quantile); Q75, upper quartile (75 %
quantile)
Significant at < 0.05
Discussion
With the da Vinci surgical robot system, there have been calls for inclusion of tactile/haptic
feedback. Even so, this system is currently in clinical use without such feedback
and results have been good. In developing the EOR, we used an existing flexible endoscope
with the aim of capturing an “intuitiveness” that reflects the feeling of scope operation
that endoscopists have cultivated over many years. In EOR version 3, we included haptic
feedback for the first time, but had not objectively evaluated whether this function
is necessary. Inclusion of a haptic feedback function has disadvantages: among other
things, it makes the system more complex with the addition of various sensors, heavier
and larger to achieve sufficient rigidity that vibrations are not picked up. If this
system did not offer clear benefits, we would clearly want to eliminate it from the
system design.
This study showed that insertion time was shorter and incidence of sigmoid colon overstretching
was lower with haptic feedback than without, and that insertion could be done without
using excessive force. This demonstrates for the first time that haptic feedback is
beneficial in remote manipulation of flexible endoscopes. When such feedback was not
used, great force was needed in pulling, in particular, and a higher incidence of
overstretching the sigmoid colon seemed to be a distinguishing characteristic. This
was attributed to the fact that operators cannot feel resistance during insertion
from the master unit with vision only and they exert too much force when pulling to
shorten the sigmoid colon because they are overly conscious of the risk of overstretching
the sigmoid colon. However, the actual frequency of overstretching the sigmoid colon
is high. As a compromise for the purpose of inserting the scope to the cecum, endoscopists
insert the scope with a push and cause overstretching. On the other hands, all insertion
lengths were approximately the same in both groups. This is because the operator achieved
intestinal shortening in the deep part beyond the sigmoid colon. Moreover, both maximum
push and mean push did not show significant differences between hap (+) group and
hap (–) groups. In our previous study, the colon was divided into two zones: Zone
A was from the rectum to the sigmoid/descending colon transition and zone B was from
the sigmoid/descending colon transition to the cecum [12]. Both maximum push and mean push were significantly higher for zone B than for zone
A, demonstrating that a stronger force was required for deeper insertion. We opined
that the strong force required for insertion of the colonoscope into the sigmoid colon
did not significantly influence either the maximum or mean force required for insertion
of the scope into the cecum. In achieving intestinal shortening, manipulation is needed
without too much insertion while feeling the resistance as the scope is being inserted,
and bilateral haptic feedback that can ascertain the haptic force in the master unit
and transmit suitable force to the slave unit is considered necessary.
In the current, the overall time interval of the control cycle was 1 minute with a
hardware interval trigger, while the data recording cycle was 0.2 second, as shown
in the paper. The force and position data in the actual system were captured every
1 minute in the control interval cycle for the bilateral control method. Even with
utilization of a preemptive multitasking thread program, the excessive usage of the
data bus for accessing Solid state drive (SSD) might disturb the hardware interval
period because today's computer systems tend to prefer access for data storage. Hence,
the recording cycle was reduced to every 0.2 seconds. If we require more frequent
data recording, Random access memory (RAM)-based data acquisition would be useful,
given the advances in today's personal computer systems with the gigabyte (GB) order
of dynamic random access memory (DRAM). Of course, we realize that the stability problem
derives not only from the interval time length, but also from other software and mechanical
elements.
While retaining its possibilities as a robot to assist in colonoscope insertion, the
ultimate aim for the EOR is to provide a treatment robot. The current investigation
was conducted as a preliminary step to demonstrate whether haptic feedback is useful
in scope manipulation.
Creating a robot system for endoscopic treatment requires: 1) a flexible endoscope
robot for remote forceps manipulation in the master; and 2) a robot that can perform
scope manipulation to access the treatment target. This means, in addition to access
in the broad sense of arriving in the vicinity of the treatment target, there is access
that demands fine movements at treatment targets that cannot be handled with forceps
manipulation only in the restricted workspace of the intestinal tract. For example,
access manipulations include activities such as proximate operations that maintain
a proper sense of distance with the lesion and approach angle manipulations. This
second element is what EOR acts on. Endoscopic treatment was originally developed
as a low-invasive method that can be performed in a relatively short time by a single
endoscopist without the need for general anesthesia. By simply combining a flexible
endoscope robot for remote manipulation of the master and other forceps together with
EOR, the manipulation system may become more complex than useful, and doubts have
been raised as to whether use of such robots offers any advantages. We think there
is a need to take apart both operation systems and rebuild a platform that enables
more intuitive manipulation, and this represents our next task [17].
