Download Verghese

dition was then detected by a two-depth transcranial Doppler (TCD) systemic venous pressure, establishes the downstream pressure for
ultrasound, and the corresponding external pressure was taken as the cerebral perfusion. This is a consequence of the Starling resistor effect,
estimate of ICP.
resulting from the collapse of the cerebral veins owing to ICP being
Some nICP estimation methods feed simultaneous measurements greater than venous pressure (27); it is also the reason that CPP is
of peripheral arterial blood pressure (ABP) along with TCD measure- defined as the difference between mean ABP and ICP, rather than bements of CBFV into multiparameter mappings to generate the ICP tween mean ABP and systemic venous pressure.
estimate. Examples are mappings involving nested regressions (22),
Our model is conveniently specified by its electrical circuit analog
neural networks (23), or support-vector machines (24). The recording (Fig. 1C), where pressures are represented by voltages, and flows by
of ABP and CBFV waveforms in the clinical setting is quite routine; currents. The instantaneous ABP and CBF at time t are represented by
ABP measurement is already necessitated in a wide spectrum of crit- the voltage pa(t) and the current q(t), respectively. The effective
ical care patients, and CBFV is the standard of care in patients with resistance of the cerebral vasculature supplied by the middle cerebral
certain neurovascular pathologies. However, the large number of artery (MCA) is represented by the resistor R, and the effective comparameters and the lack of an underlying mechanistic model mean pliance of this cerebral vasculature and surrounding brain tissue is
that such “black box” mappings can fail to adequately and robustly represented by the capacitor C. Our algorithm for estimation of ICP—
capture the relevant physiology.
with simultaneous estimation of R and C—resulted from requiring the
Almost all the above noninvasive methods require calibration or model constraints to be satisfied as closely as possible by the obtained
tuning of parameters that relate the measured quantities to the ICP measurements, over an estimation window comprising the data assoestimates. Such calibration or tuning typically involves the use of ICP ciated with several consecutive beats, and under the assumption that
measurements obtained invasively on the patient or from some ref- ICP, R, and C are constant over that window.
erence population. Furthermore, training on a reference population
For each estimation window, the algorithm generated one nICP
causes the accuracy of the ICP estimates to depend on how well a par- estimate, which can be considered an estimate of the mean ICP over
ticular patient is represented in the training set. As noted by Popovic the estimation window. The estimation window had to be long
and coauthors (10), after surveying nearly 30 nICP methods patented enough (more than five beats) to allow some averaging of the data
over the last 25 years, none of the methods is sufficiently accurate to over multiple beats, with a corresponding attenuation of the effects
allow for routine clinical use. An additional factor in the way of clin- of measurement noise, respiratory artifacts, and other such perturbaical adoption for some of the proposed approaches is the difficulty or tions. However, the window also needed be short (≤60 beats) comexpense (hardware, computation, human resources) of the involved pared to the time scales of significant transients in the underlying ICP.
measurements. None of the previously proposed approaches to nICP
The ABP in our model was arterial pressure at the MCA, whereas
estimation has transitioned from the research setting to accepted clin- our ABP measurement was made at the radial artery. These two arical practice, although commercial products based on the methods in terial pressure waveforms undoubtedly differ in transit time from the
(17), (21), and (22) are available.
heart and in pulse morphology; their mean values are close, however,
Here, wevulnerable
present a model-based
approach
to obtaining
estimates
provided
measurements
taken with
respect
a common
Brain tissue is highly
to unbalanced
oxygen
demand
and supply.
A few
seconds ofareoxygen
deficit
may to
trigger
neurological symptoms, and sustained oxygen
of ICP on a beat-by-beat time scale from noninvasive waveform reference. Although there is no straightforward way to correct for
deprivation over a few minutes may result in severe and often irreversible brain damage. The rapid dynamics coupled to the potential for severe injury necessitate
measurements of CBFV and ABP. Our approach does not require morphological differences, our algorithm determines and applies an
continuous, patient-specific
and ideally noninvasive,
monitoring
in theappropriate
populations
greatest
risk forradial
developing
oronexacerbating
brain injury. One of the key variables
calibration orcerebrovascular
training on a reference
population.
time at
shift
to the measured
artery ABP
the estiassociatedwith
computational
burden
is negligible, thereby
allowing
mation window
to obtain
waveform on
that brain
can serve
as aand
plausible
to monitor The
in patients
brain injury
is intracranial
pressure
(ICP), which
determines
theapressure
tissue
also affects cerebral perfusion. Current
near–real-time
ICP. the penetration of the skull and
proxythe
for ABP
at the MCA.
measurement
modalitiesestimation
for ICP of
require
placement
of a pressure-sensitive probe in the brain parenchyma or cerebral fluid spaces.
Computational Physiology and Clinical Inference Group
Multi-scale
Bioengineering
and
Biophysics
Downloaded from stm.sciencemag.org on October 9, 2012
Principal Investigators: George C. Verghese, Thomas Heldt
Graduate Students: Shamim Nemati, Becky Asher, Greg Ciccarelli, Ehi Nosakhare
Undergraduate Students: James Noraky, Aditya Kalluri, Yu-chi Kuo, Ben Frank
The Computational Physiology and Clinical Inference Group develops and applies
computational models of human physiology for clinical monitoring and inference. Our
current research focuses on cardiovascular, cerebrovascular, respiratory and
neurological applications.
Projects
Non-invasive ICP Estimation
We are developing and validating non-invasive methods to assess ICP continuously and robustly in a patient-specific and calibration-free manner.
A
B
C
RESULTS
Skull
Skull
ICP
Dynamic model and
CSF
CSF
estimation algorithm
ABP
Detailed dynamic models of the cerebroArteries
Veins
ICP
CBF
vascular space (Fig. 1A) have been develBT
oped in the literature (25–27). We obtained
Brain
ICP
a highly simplified model that focuses
AN
on the major intracranial compartments—
brainabstraction
tissue, cerebral
and CSF
Progressive
of vasculature,
cerebrovascular
physiology:
(A) Relevant
cerebrovascular
anatomy.
(B) Schematic
representation
of the
main cerebrovascular compartments and associated
Fig. 1. Progressive
abstraction
of cerebrovascular
physiology.
(A) Relevant
cerebrovascular
anatomy:
space—and the associated variables (Fig.
brain tissuevariables.
(BT), cerebrospinal
fluid (CSF),
and cerebralrepresentation
arterial network (AN).
(B) Schematic representaphysiological
(C) Lumped
circuit-model
of cerebrovascular
physiology.
1B). The variables involved in the model tion of the main cerebrovascular compartments and associated physiological variables: cerebral blood
Kashif, Faisal M. et al. “Model-Based Noninvasive Estimation of Intracranial Pressure from Cerebral Blood Flow Velocity and Arterial Pressure.” Science Translational Medicine 4, 129 (2012).
are ABP at the level of the cerebral vascu- flow (CBF), arterial blood pressure (ABP), and intracranial pressure (ICP); the collapsed venous segment is
lature, CBF at the inlet of a major cerebral also shown. (C) Lumped circuit-model representation of cerebrovascular physiology: CBF q(t), cerebral
artery, and ICP. Our lumped model rep- arteriovenous flow q1(t), and ABP pa(t). ICP denotes both extraluminal pressure and the effective
resents the relevant physiological mecha- downstream pressure for cerebral perfusion.
Disease Classification with Capnography
Assessment of brain-heart interaction via ECG variability
CHF
50
40
30
20
10
0
0
5
10
15
20
25
30
20
25
30
COPD
50
40
PeCO2 (mmHg)
Capnography refers to the non-invasive measurement of the concentration of carbon dioxide exhaled in the breath. Carbon
dioxide is a byproduct of tissue metabolism, and its concentration, [CO2], can be measured noninvasively as a function of time
www.ScienceTranslationalMedicine.org 11 April 2012 Vol 4 Issue 129 129ra44
2
in exhaled breath. This process is called capnography, and the resultant time series is known as a capnogram. Existing methods
for extracting diagnostic information from the capnogram are qualitative, through visual inspection, and therefore imprecise.
We are working to quantify the capnogram in order to discriminate among various lung disorders, including obstructive lung
disease (e.g. chronic obstructive pulmonary disease) and restrictive lung disease (e.g. congestive heart failure). Our initial results
demonstrate the diagnostic potential of capnography.
Capnogram
Template
Detected Exhalations
Excluded Exhalations
30
20
10
0
0
5
10
15
Normal
50
40
30
20
10
0
0
Heart rate variability is a measure of how the heart rate varies with the autonomic activity. It is a quantitative way of viewing
Waveform strips from normal and abnormal
the response of the cardiovascular system to the two major components of the autonomic nervous system: the sympathetic
capnograms. Detected exhalations (green) are
and parasympathetic activity. Hence it is one of the important non-invasive assessments of the modulation of heart frequency. overlaid with the record’s template exhalation
Its dynamics serve as an indicator of how heart rate, respiration, blood pressure and temperature are controlled by the ANS (black) and outlier exhalations are displayed in red.
through the brain. We are developing methods to use this variability in order to assess brain-heart interaction.
5
10
15
Time (s)
People
Pictured (l-r)
Principal Investigators: George C. Verghese,
Thomas Heldt
Graduate Students: Shamim Nemati, Becky Asher, Greg Ciccarelli, Ehi Nosakhare
20
25
30