Hypnometer(1).pdf

(213 KB) Pobierz
Hypnometer
Measuring Hypnosis: Relating
the Subjective Experience to
Systematic Physiological Changes
Solomon Gilbert Diamond and Robert D. Howe
Harvard University
Division of Engineering and Applied Sciences
sdiamond@fas.harvard.edu
1. Introduction
Clinical hypnosis is a mind-body technique that operates at the intersection of
subjective perceptions and objective physiological changes. A fundamental problem
with hypnosis research is that the subjective mental state of patients during hypnosis
cannot be measured directly. Experimental paradigms that neglect to measure
changes in mental state at best yield a correlation between the treatment procedure
and outcomes but cannot demonstrate a causal link. Current practices rely on the
subjective reports of the subject to distinguish whether a negative experimental
outcome arises because the patient never achieved the hypnotic state or because
hypnosis was an ineffective treatment. The purpose of this research is to bridge this
gap between the subjective perception of the hypnotic state and objective
measurement of concomitant physiological changes. In addition, we seek a technique
that can estimate depth of hypnosis during the course of a hypnosis session, on a time
scale of less than a minute.
2. Background
2.1. Previous attempts to identify physiological changes specific to hypnosis
Hypnosis is commonly thought to be associated with EEG alpha frequencies but
reproducibility has been difficult to demonstrate (Perlini and Spanos 1991). PET and
fMRI studies have found significant but inconsistent differences during the hypnotic
state (Maquet and al. 1999), (Rainville 1999), (Ulrich, Meyer et al. 1987). The
hypnotic state can also be monitored by examining variability in the heart rate
2
Measuring Hypnosis: Relating the Subjective Experience to Systematic
Physiological Changes
measured between subsequent beats of the heart – the heart rate variability (HRV)
signal. The heart rate exhibits spontaneous fluctuations even at rest that reflect the
continuous influence of the autonomic nervous system (ANS) on the heart’s
pacemaker cells (Akselrod and al. 1981). The HRV signal typically contains a high-
frequency (HF) component near respiratory rate (@0.25 Hz). The spectral power in
the HF component has been shown to increase during conscious relaxation compared
with rhythmic breathing at 0.25 Hz (Sakakibaba, Takeuchi et al. 1994). Peng et al.
found exaggerated heart rate oscillations associated with slow breathing during
meditation that were significantly different from metronomic breathing and from
spontaneous nocturnal breathing by normal adults or elite athletes (Peng, Mietus et al.
1999). There are accounts in the hypnosis literature that HRV is affected by mental
absorption (Zachariae, Jogensen et al. 2000) and by the hypnotic state (DeBenedittis
and Cigada 1994). These studies show that parameters calculated from HRV change
in specific ways during meditation, mental absorption and hypnosis.
2.2. Experimental hypotheses
1. A single parameter can be calculated from HRV that will change
systematically during the hypnotic state when compared with a control
condition that is commensurate with the hypnotic state.
2. The average values of such a parameter would increase when more hypnotic
phenomena are experienced thereby providing evidence for a hypnosis-
specific measure.
3. Dynamic self-rating of hypnotic depth during hypnosis will also correlate
with a single dynamic HRV parameter. A dynamic HRV parameter that
correlated with dynamic self-rating would enable real-time monitoring of
hypnotic depth.
2.3. Need for dynamic HRV parameterization
The HRV signal is complex due to the many sources of physiological variation with
varying degrees of interdependence. Sources of variation include vagal tone,
baroreflex mechanisms, circadian rhythms, respiration, and stress levels. These
multiple sources of variation are embedded within the single HRV time series. This
matter is further complicated because the variation introduced by each source is
dynamic. Lumped statistics such as mean and variance fail to distinguish between
sources of variance. Power spectrum analysis requires many minutes of HRV data to
achieve reasonable frequency resolution and is therefore incapable of tracking
dynamic HRV frequency dynamics that occur on the order of seconds.
3. Methods
3.1. Proposed dynamic HRV model
Since HRV is sampled only once per heart beat, dynamic parameterization of HRV on
the time scale of seconds requires that parameters are extracted from just a few data
Measuring Hypnosis: Relating the Subjective Experience to Systematic
3
Physiological Changes
points. By treating HRV as a single oscillator, a sinusoid model with an offset and
additive noise can be applied.
w (1)
In this model, the amplitude a , frequency w , phase angle f , offset d and noise n are
allowed to vary on a larger time scale than the sampling rate.
Fitting the parameters within the temporal window to the proposed model is
difficult because of its nonlinearity and potential for aliasing. Conventional
optimization algorithms are sensitive to noise and often converge to higher frequency
alias solutions. A least-squares method was devised to fit the model parameters
thereby increasing the computational speed and noise rejection.
HRV
=
a
(
t
)
sin
[
(
t
)
t
+
f
(
t
)
]
+
d
(
t
)
+
n
(
t
)
ECG
t D t
ECG( D t )
HRV
Interpolated
HRV
Phase-Space
HR
HR
HR
t
t
+
+ t
t
t
t
+
D
t
Heartbeat
t
t
+
D
t
HR
+ t
-
Best-Fit
Ellipse
Sinusoidal
Model
HRV Dynamic Parameters
1 t
)
HR
t
t
+
+ t
D
t
HR
n ( t )
a ( t )
HR
+ t
-
Time
d ( t )
Figure. 1. Least-squares fit of parameters to the proposed dynamic HRV model. A sample of
ECG data of length D t is extracted from an ECG record. The heart rate per heartbeat (HRV) is
calculated by taking the inverse of the beat-to-beat times. HRV is then interpolated to a
regularly sampled time series. The time series is plotted in pseudo-phase space verses itself at
a time lag of t . Sinusoidal oscillations in HRV become an ellipse in this phase-space. The
quadratic coefficients for an ellipse are fit to the data in phase-space using a least squares
method. The quadratic coefficients are mathematically transformed back to the sinusoidal
model. The resulting HRV dynamic parameters ( HRVdp ) are the amplitude a , frequency w
offset d and a goodness of fit parameter that is inversely related to the additive noise n ( t ); the
phase angle f is not used.
The chosen size of the temporal window was D t = 6 sec. so that typically 5 estimates
of the heart rate were contained within the window. To correct the irregular sampling
rate of HRV, the heart rate estimates for each heartbeat were dropped onto the nearest
time point of a 10Hz temporal grid and intermediate data points were interpolated
with a cubic spline. The phase-space transformation used a time lag of t = 1 sec.
yielding a frequency range of 0 to 0.5 Hz. Once the parameters were estimated at
time t the temporal window was advanced by 1 sec. and then the process was
repeated, yielding a 1 Hz moving estimate of the model parameters.
D
t
t
w
t
t
88822009.037.png 88822009.038.png 88822009.039.png 88822009.040.png 88822009.001.png 88822009.002.png 88822009.003.png 88822009.004.png 88822009.005.png 88822009.006.png 88822009.007.png 88822009.008.png 88822009.009.png 88822009.010.png 88822009.011.png 88822009.012.png 88822009.013.png 88822009.014.png 88822009.015.png 88822009.016.png 88822009.017.png 88822009.018.png 88822009.019.png 88822009.020.png 88822009.021.png 88822009.022.png 88822009.023.png 88822009.024.png 88822009.025.png 88822009.026.png 88822009.027.png 88822009.028.png 88822009.029.png 88822009.030.png 88822009.031.png 88822009.032.png 88822009.033.png 88822009.034.png
4
Measuring Hypnosis: Relating the Subjective Experience to Systematic
Physiological Changes
The transformed interpolated HRV from time t to
t + expressed in phase
t
space with the time lag t is given by:
x
=
HRV
(
t
˛
(
t
i
,
t
i
+
D
t
-
t
)
)
=
a
sin
(
w
t
+
f
)
+
d
+
n
(2)
(
)
(
)
y
=
HRV
t
˛
(
t
+
t
,
t
+
D
t
)
=
a
sin
w
(
t
+
t
)
+
f
+
d
+
n
i
i
If the noise component of the HRV signal is small, then the geometric
representation of the HRV data in the xy -plane is an ellipse that can be described by
the generalized quadratic form
Ax
2
+
Bxy
+
Cy
2
+
Dx
+
Ey
+
F
=
0
(3)
where the quadratic coefficients are related to the HRV model parameters as
A
=
1
Ł
tan
2
wt
+
2
+
1
ł
D
=
-
d
(
tan 2
2
wt
+
1
4
a
2
2
tan
2
wt
a
2
2
B
=
1
Ł
tan
2
wt
-
1
ł
E =
D
(4)
2
a
2
2
tan
2
wt
2
D
2
C =
A
F
=
-
1
.
2
A
+
B
From equations (4) it is apparent that if the quadratic coefficients could be
estimated for the HRV data then the amplitude, frequency and offset parameters could
be calculated with the inverse relationship,
a
=
4
A
-
w
=
2
t
atan2
2
A
+
B
d
=
-
D
+
.
(5)
4
A
2
B
2
2
A
-
B
2
A
B
Pilu et al. describe a method for direct least squares fitting of an ellipse that is
ideally suited to this application (Pilu, Fitzgibbon et al. 1996). Pilu’s method is
constrained to yield the quadratic coefficients subject to the elliptical constraint of
0
- AC
<
3.2. Normalizing the HRV dynamic parameters
In order to make inferences based upon HRVdp , the values must be statistically
compared to a control condition. Statistical properties of the control condition
HRVdp can also be used to normalize HRVdp values for both the control and
experimental conditions. Subtracting the control condition means and then dividing
by the control standard deviation scales the distributions of the control HRVdp values
to the standard normal distribution and makes all of the parameters dimensionless.
This also places all of the experimental HRVdp values onto the same normal
distribution axis. Averaging the four scaled HRVdp ’s yields a single normalized
HRV dynamic parameter ( nHRVdp ). When averaging the four scaled HRVdp ’s, sign
changes can be used to normalize the expected direction of parameter change during
the experimental condition. In this study, the sign of all the parameters was flipped so
that a decrease in a , w , d , and n all result in a more positive nHRVdp on the standard
normal distribution scale. The significance of the experimental condition nHRVdp
values can now be statistically tested with the null hypothesis that the mean is zero.
2
B and it is computationally efficient. The HRV dynamic parameters
(HRVdp) a , w , d , and n are calculated using equation (5) from the best-fit quadratic
coefficients.
4
88822009.035.png
Measuring Hypnosis: Relating the Subjective Experience to Systematic
5
Physiological Changes
The 1Hz-sampling of nHRVdp enables reasonable comparisons to be made with 10 to
30 seconds of historical data.
3.3. Experimental Proceedure
Eleven subjects participated in the study (5 male, 6 female, mean age 21). Two
subjects reported having some familiarity with hypnosis. None of the subjects had
any history of psychological disorders, trauma or cardiac health problems. None of
the subjects were currently taking medications.
During the hypnosis condition, subjects were instructed to sit comfortably with
their eyes closed while listening to a hypnotic induction and suggestions spoken by
the experimenter. Subjects were instructed to move a lever periodically to indicate
how hypnotized they felt on a scale of 1 to 5 during the experiment. Subjects were
reminded to move the lever every 1 to 2 minutes. The hypnotic suggestions
encouraged focus on imagined sights, sounds and feelings. During the control
condition subjects were instructed to sit comfortably and relax with their eyes closed
while listening to the experimenter. Subjects were asked a series of true or false
questions designed to ensure that the subjects stayed awake and focused. The content
of the questions had minimal emotional content and required only commonly held
knowledge. Subjects indicated their responses by moving the same lever used to
indicate hypnotic depth.
After giving informed consent, subjects were questioned about their previous
experiences with hypnosis. The Hypnotic Induction Profile (Speigel and Speigel
1978) was used to measure subjects’ hypnotizability. Subjects were then prepared for
data logging sampled at 200Hz with an ECG (HP 78354A) and respirometer
(custom). Data was recorded for 10 minutes of the control condition followed by 10
minutes of the hypnosis condition. After the hypnosis session, subjects were asked
how hypnotized they felt and if they experienced various hypnotic phenomena.
4. Results
4.1. Hypnotizability and hypnotic phenomena
Of the 11 subjects tested with the Hypnotic Induction Profile, 10 were determined to
have intact hypnotic ability and 1 was not responsive to the hypnotizability test. Data
from this subject was subsequently excluded from the analysis.
Table. 1. Overall self-rating of hypnotic depth (n=10)
Subjective Self-Rating
1
2
3
4
5
Number Reporting
0
2
5
2
1
Table. 2. Hypnotic phenomena reported by subjects immediately after hypnosis (n=10)
Hypnotic Phenomenon
Num. Subj. Hypnotic Phenomenon
Num. Subj.
Vivid mental imagery
9
Imagined textures
6
Heaviness/sinking into chair
8
Drifting sensations
5
Time distortion
8
Tingling sensations
4
Floating sensations
7
Imagined smells or tastes
4
Clear mental sounds
6
Unusual temperature changes
3
88822009.036.png

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika:

Zgłoś jeśli naruszono regulamin