METHODS OF EVALUATING THE LEVEL OF CONSCIOUSNESS USING AEP, EEG AND ANFIS
The invention relates to a method of evaluating the level of consciousness of a patient through a numerical index (AAI) which is derived from information from Auditory Evoked Potentials (AEP), a High-Low Electroencephalogram Frequency Ratio (HLR) and the Burst Suppression (BS) measured on the Electroencephalogram (EEG).
The invention also relates to a method to determine a Depth of Anaesthesia
Index (DAI) of a patient which is derived from information from Auditory Evoked Potentials (AEP), the Electromyographic activity (EMG), the Electroencephalogram (EEG) and the Burst Surpression rate (BS).
The AEP signal is an evoked electrical activity embedded in EEG activity that is elicited in a neural pathway by acoustic sensory stimulus provided by a train of acoustic pulses.
A method to monitor AEP signals is disclosed in the published International patent application no. WO 01/74248.
According to this published application it is possible within 6 seconds to measure a reliable AEP signal which is calculated from an autoregressive model with exogenous input.
Even though the method according to WO 01/74248 has shown good and reliable results it is possible that unwanted electrical interferences arising from the EMG are present on the AEP signal.
A system to determine and evaluate the quality of the extracted AEP signals through estimation of the SNR (Signal-to-Noise Ratio) is described
in the Danish patent application no. PA 2001 00381 , which was not published at the day of the filing of the present application.
In some subjects there is the possibility that a part of the EMG interferences may be synchronised with the click stimulus.
These interferences are triggered by what is called the 'startle effect' or involuntary contraction of the facial muscles in response to the sound of the click. Usually the startle effect is a defensive response against an excessive acoustic stimulation.
When this effect is present, the signal extracted from the synchronised average process contains the AEP activity plus the synchronised EMG part, and the extracted signal shape gets distorted by the presence of a high amplitude peak in the range of 15-40 ms.
The presence of the synchronised EMG produces extremely high values of the Signal-to-Noise Ratio (SNR).
The object of the invention is to eliminate these drawbacks and obtain an improved AAI.
The object is solved according to the invention by the method according to claim 1 , i.e. by calculating the index AAI from the formula:
AAI = f(AEP) + f(HLR) +f(BS)
HLR = kHLR.Iog(Eb1 ) / log(Eb2) in which
kHLR is a constant value,
Eb1 is the energy of all frequencies ranging from 0-10Hz until 20-
Eb2 is the energy of all frequencies ranging from 10-20Hz until
20-δOHz.
If, as stated in claim 2, the synchronised EMG activity is eliminated from the
AEP signal by manually or automatically controlling the volume of the acoustic stimulus given to the patient, an elimination of the synchronised interference from the AEP signal is achieved.
If, as stated in claim 3, a SNR value is detected in a monitor and if the SNR value is higher than a prefixed threshold, a message instructs a user to lower the volume of the click stimulus.
In this way a user is instructed to set the volume to a lower level using a control with prefixed volume intensities.
It is further expedient if, as stated in claim 4, for the automatic volume control a closed control system with negative feedback is used, said system comprising a Proportional Integral Derivative (PID) controller fed by a 0- order Sugeno type Fuzzy Inference System (FIS).
In order to further improve the AAI value it is expedient if, as stated in claim 5, the AAI value, along with the corresponding values for effect site concentrations of Propofol (PROP) and Remifentanil (REMI), is used as input to an ANFIS type forecasting system in order to provide a composite
ANFIS-AAI Index (CADI) to evaluate the level of consciousness of a patient.
An expedient embodiment for carrying out the invention in claim 5 is found in claim 6 in that the values of PROP, REMI and the Bispectral Index (BIS) are used as inputs to the ANFIS type forecasting system in order to provide
a composite ANFIS-BIS Index (CABI) to evaluate level of consciousness, or, as stated in claim 7, that the values of PROP, REMI, AAI and BIS are used as inputs to the ANFIS type forecasting system, the ANFIS type forecasting system providing a composite ANFIS-AAI-BIS Index (CADBI).
As mentioned the invention also relates to a method of determining a depth of anaesthesia index.
This method is characterised in that the index DAI is derived from the formula:
DAI = k1* S(AEP) + k2Η(AEP) + k3* S(EMG) + k4*H(EMG) + k5* S(BS) + k6*H(BS) + k7* S(EEG) + k8*H(EEG)
where
S is a symbol from the corresponding symbol sequence, H is Hurst's exponent defined by H(T):=log(R/S)/log(T), where T is the duration of the sample k(1....7) are either constants or delimiter functions.
The principles according to the formula defined in claim 8 are explained in the following:
The basic principle of symbolic dynamics is to transform a time series into a symbol sequence. These provide a model for the orbits of the dynamical system via a space of sequences.
Given an AEP, the symbol sequence is achieved by quantising the AEP into boxes labelled with a symbol. Calculating attributes of the symbol sequence
can reveal non-linear characteristics of the original time series and the examined dynamical system. This variable is termed S.
The Hurst exponent, H, is defined as:
H(T) = log(R/S)/log(T),
where T is the duration of the sample of data, and R/S the corresponding value of rescaled range.
If H=0.5, the behaviour of the time-series is similar to a random walk; if H<0.5, the time-series covers less "distance" than a random walk (if the time-series increases, it is more probable that then it will decrease, and vice-versa); if H>0.5, the time-series covers more "distance" than a random walk (if the time-series increases, it is more probable that it will continue to increase).
The Hurst exponent is a measure related to the existence of long term correlations.
The Depth of Anaesthesia Index (DAI) uses information from the AEP, the EMG activity, the EEG and the Burst Suppression (BS). The DAI is then defined as:
DAI = k1* S(AEP) + k2*H(AEP) + k3* S(EMG) + k4Η(EMG) + k5* S(BS) + k6*H(BS) + k7* S(EEG) + k8*H(EEG) where k(1....7) are either constants or delimiter functions.
An embodiment of the invention will now be explained in connection with the drawing which shows a diagram of an automatic volume comprising a
Proportional Integral Derivative (PID) controller fed by a 0-order Sugeno type Fuzzy Interference System (FIS).
A Sugeno type FIS is defined as a system that uses fuzzy set theory to map a number of inputs to one output. Firstly, a set of membership functions has to be established for each input parameter. These functions define a set of fuzzy rules which combine all possible parameter function values in order to produce a single output.
In this case, the input parameters are:
The Signal-to-Noise Ratio measured in the middle latency auditory evoked potential (SNR).
The Signal-to-Noise Ratio measured in the brainstem auditory evoked potential (SNRb).
The energy of the electromyogram (EMG). The AAI value.
The following fuzzy membership functions are established for each parameter:
SNR values are classified in three categories (LOW, OK, HIGH). EMG values are classified in two categories (LOW, MEDIUM, HIGH). AAI values are classified in three categories (LOW, MEDIUM, HIGH).
The combinations among these membership functions determine a set of n fuzzy rules.
The fuzzy system output (OUTFUZZY) can be considered then as a measure of the global quality of the AEP. OUTFUZZY is a continuous value between 0 and 1.
An OUTFUZZY value close to zero indicates an over-stimulation and signals the convenience of lowering the volume of the click stimulus.
If OUTFUZZY is close to one, the quality of the AEP is considered to be low and an order to increase the volume of the stimulus is issued.
An OUTFUZZY value close to 0.5 is considered to be adequate and the click volume remains unchanged.
The OUTFUZZY value is then fed into a PID controller. The function of the PID controller is to protect the system from sudden changes or instabilities by automatically adjusting a variable (uc) to hold a measurement at a given set point.
The input to the PID is an error variable (e) which is the difference between the desired set-point value and the OUTFUZZY value that comes from the fuzzy system.
The desired value for adequate AEP quality is set to 0.5 and uc is then adjusted according to the equation:
uc = P. e + l. e + D. (d/dt) e
where
uc is the PID output, P is a proportional gain constant, I is an integral gain constant, D is a derivative gain constant.
The value of uc will determine an order from a decisor (d) to increase or lower the volume of the click stimulus by one volume level.
The volume scale lies in the 35 to 100 dB SPL range and is discretised in N integer levels.
Once the AEP signal quality has been improved, its information, along with other quantities pertaining to the patient's behaviour, is analysed according to the SNR and BS conditions.
A High-Low EEG Frequency Ratio (HLR) is defined as:
HLR = kHLR.Iog(Ebl ) / log(Eb2)
where
kHLR is a constant value,
Eb1 is the energy of all frequencies ranging from 0-10Hz until 20-30Hz, Eb2 is the energy of all frequencies ranging from 10-20Hz until 20-50Hz.
Depending on SNR conditions, the proportionate value of HLR can be adjusted according to the equation:
KLIM = f(SNR)
Where KLIM is a constant value in the range (0.9-3)
As a result:
f(HLR) = KLIM(HLR)
The term Burst Suppression defines a phenomenon seen on the EEG that indicates loss of cortical brain activity and is characteristic of very deep anaesthesia.
Burst Suppression is characterised by very low amplitude activity (usually less than 3-5 microvolts). The presence of these low-amplitude periods is quantified as a percentage by unit time over the EEG signal and is termed BS.
The effect of the BS value is also taken into account as a function:
f(BS) = kBS(BS)
where kBS is a constant value in the range (-0.1 , - 1 ).
The value of AAI is then adjusted according to the HLR and BS values as:
AAI = f(AEP) + f(HLR) + f(BS).