HK1208143B - System, method, and computer algorithm for characterization and classification of electrophysiological evoked potentials - Google Patents

Description

System, method and computer algorithm for characterization and classification of electrophysiological evoked potentials

Technical Field

The present invention relates generally to detecting changes in Evoked Potentials (EPs), and more particularly to automatically detecting changes in EPs using computer algorithms.

Background

Somatosensory evoked potentials are accumulated potentials typically recorded from the head or neck region after repeated stimulation of peripheral nerves. Monitoring patients during surgery using somatosensory evoked potentials has been shown to allow early identification of impending positioning effect (impairment) that can then be avoided by repositioning the patient to relieve stress or strain.

For example, as described in the following documents: hickey, C, Gugino, l.d., Aglio, l.s., Mark, j.b., Son, s.l., and Maddi, r. (1993), "inductive electronic based on compressed nuclear input dual wheel heart supply," analytical science 78(1), 29-35; kamel, I.R., Drum, E.T., Koch, S.A., Whitten, J.A., Gaughan, J.P., Barnet, R.E., and Wendling, W.W, (2006), "The use of a passive porous to a derivative of The relative position between The passive porous positioning and embedding approach surface measuring approach surface supply: ecological emission analysis," animal analysis 102 (1545), 1538-2; and Labrom, R.D., Hoskins, M., Reilly, C.W., Tredwell, S.J., and Wong, P.K.H. (2005), and "clinical laboratory of microbiological infected sites for detection of brachialplexus therapy to clinical laboratory in science supply" Spine 30(18), "Spine 30-2093, the entire contents of which are incorporated herein by reference.

Such monitoring is typically performed by a trained technical specialist under the supervision of a physician using sophisticated multi-channel amplifiers and display equipment. Unfortunately, the availability of such personnel and equipment is limited, requires advance reservations, and is expensive. Furthermore, such monitoring has traditionally not been possible in many areas where positioning effects occur outside of the operating room (where unconscious, infirm, or confined patients may have positioning effects).

Typically, technical experts examine waveforms, while neurologists examine EP waveforms simultaneously on-site or remotely over the internet. Technical experts and neurologists are trained and are experts that determine whether changes in EP waveforms are significant and indicate impending nerve damage. The cost of fully engaging experts in interpreting these waveforms results in service rations to almost all high risk procedures.

U.S. patent application publication No. 2008/0167574 describes a semi-automatic device that can be used to automatically measure biological signals during a surgical procedure to avoid nerve damage. However, this device focuses on muscle or motor recording to measure nerves near the surgical instrument and does not address the positioning effect.

The difficulty in determining the localization effect by analyzing and classifying waveforms is the large variation in amplitude, frequency, and shape of the waveforms. These changes are caused by a number of factors, including anesthesia, electrical interference with other devices, and any pre-existing abnormalities of the nerve.

Accordingly, there is a need for systems and methods that overcome the shortcomings of previous systems and methods.

Disclosure of Invention

In exemplary embodiments of the invention, systems, methods, and computer algorithms for characterization and classification of electrophysiological EPs are disclosed. EP can be defined as the voltage versus time signal obtained by the nervous system using appropriate electrodes. For example, when EP is obtained from a somatosensory system, a signal can be obtained by ensemble averaging electrophysiological responses to repeated stimuli of the somatosensory system detected using appropriate electrodes. Examples of EP are somatosensory, auditory or visual EP. The algorithm is applied to a time series of EPs acquired during an ongoing clinical procedure. The algorithm establishes a baseline (baseline)/normal EP signature and then characterizes subsequent EPs relative to the baseline EP and previous EPs to determine whether the function of the underlying sensory nervous system has been significantly affected by the ongoing clinical procedure. The algorithm communicates with ancillary hardware (analog hard) and algorithms developed to acquire sequences of EPs and provide appropriate feedback to ensure an efficient clinical workflow. The algorithm provides a basis for clinically effective application, thereby minimizing false positives and false negatives.

Further features and advantages of the invention, as well as the structure and operation of various embodiments of the invention, are described in detail below with reference to the accompanying drawings.

Drawings

The foregoing and other features and advantages of the invention will be apparent from the following, more particular description of preferred embodiments of the invention, as illustrated in the accompanying drawings in which like reference numbers generally indicate identical, functionally similar, and/or structurally similar elements.

Fig. 1 shows an exemplary depiction of stimulation of a physiological system of interest with context dependent (context dependent) stimulation according to an exemplary embodiment of the present invention.

Fig. 2 shows an exemplary depiction of a sequence of suitable stimuli and a sequence of corresponding responses applied to a physiological system of interest, according to an exemplary embodiment of the present invention.

Fig. 3 illustrates an exemplary depiction of an EP establishing an ensemble average based on multiple responses in accordance with an exemplary embodiment of the present invention.

Fig. 4A illustrates an exemplary flow chart of a process for acquiring and classifying EP responses according to an exemplary embodiment of the present invention.

Fig. 4B illustrates an exemplary flowchart of a process for determining whether a change has occurred in a sequence of EPs according to an exemplary embodiment of the present invention.

FIG. 5 illustrates an exemplary flow chart of a process for calculating a baseline response in accordance with an exemplary embodiment of the present invention.

Fig. 6 shows an exemplary flowchart of a process for determining an analysis range according to an exemplary embodiment of the present invention.

FIG. 7 illustrates an exemplary flow chart of a process for updating a baseline response in accordance with an exemplary embodiment of the present invention.

Fig. 8 illustrates an exemplary embodiment of a relationship diagram in metric calculations for characterizing an EP, according to an exemplary embodiment of the present invention.

Fig. 9 illustrates an exemplary flowchart of a process for a good state according to an exemplary embodiment of the present invention.

Fig. 10 shows an exemplary flowchart of a process for a bad state according to an exemplary embodiment of the present invention.

Detailed Description

Various exemplary embodiments of the present invention, including preferred embodiments, are discussed in detail below. While specific exemplary embodiments are discussed, it should be understood that this is done for illustrative purposes only. A person skilled in the relevant art will recognize that other components and configurations may be used without parting from the spirit and scope of the invention.

Embodiments of the present invention relate to computer signal processing and pattern recognition algorithms for real-time characterization and classification of EPS. The algorithm may replace expert analysis typically provided by technical experts and physicians. Computer algorithms running on software installed on the EP machine can be used in any procedure or situation where the patient is at risk to detect, alert and ameliorate the effects of localization or any abnormalities.

Fig. 1 shows an exemplary stimulation of a physiological system of interest with context dependent stimulation according to an exemplary embodiment of the present invention. For somatosensory systems, for example, the stimulation may be application of a current pulse of appropriate size and shape to the superficial nerves. Assuming that the appropriate stimulus is applied, the electrophysiological response is a coherent firing (volley) of action potentials along the axon that is evoked by the applied stimulus.

Fig. 2 shows an exemplary depiction of a sequence of appropriate stimuli and corresponding response sequences applied to a physiological system of interest. According to an exemplary embodiment of the present invention, the corresponding response sequence may be detected using suitable electrodes suitably configured at the appropriate recording locations. These responses consist of time-sampled and digitized measurements of the body conduction voltage field established by the electrophysiological response of the physiological system of interest when evoked by an applied stimulus.

Fig. 3 illustrates an exemplary depiction of an EP establishing an ensemble average based on multiple responses in accordance with an exemplary embodiment of the present invention. The individual responses may be contaminated by the electrophysiological responses from other physiological systems as well as the voltage contribution of ambient electrical noise. Thus, to obtain an appropriate signal-to-noise ratio, multiple responses may be averaged over the whole to establish a synthetic Evoked Potential (EP). As N (the number of responses averaged) increases, the signal-to-noise ratio of the synthesized EP increases. In embodiments, N may range from 10 to 1000 depending on the physiological system of interest.

The EP can be processed to assess the state of the physiological system of interest. The physiological system in the normal operating mode may be considered to be in a "good" condition. If the physiological system is stressed, fatigued, or injured, the system may be considered to be in an "undesirable" state. Starting from a physiological system in good condition, the detected changes in the characteristics of the EPs of the EP sequence can be used to predict whether the physiological system is in a good or bad condition.

Fig. 4A illustrates an exemplary flow chart of a process for acquiring and classifying EP responses according to an exemplary embodiment of the present invention. Each EP may be initially filtered to remove unwanted instrument noise to better present the electrophysiological response of the system of interest. The EP may be filtered based on likelihood estimates.

If a baseline response is not present, the captured response can be analyzed to estimate the baseline response and establish an analysis range. For example, if there is no N_IGood response to reception (where N_IIs the number of initial EP responses required to establish a baseline response), the baseline response may not exist. The analysis to estimate the baseline response and establish the analysis range is further described below.

If a baseline response exists, the baseline may be updated based on the current response. Updating the baseline is described further below.

Once the current baseline response is determined, then the current response is characterized relative to the current baseline and previous responses. For example, the characterization may be at least one of a Euclidean distance, a pseudo-correlation, a cross-correlation, or an energy ratio between the current response and the current baseline. The energy ratio may be a ratio of the energy between the current response and the current baseline. The energy ratio may indicate a change in the magnitude of the EP response. The current response may then be classified based on its characterization.

EP can be classified into four possible types based on characterization: good, bad, uncertain and unreliable. A good classification may indicate that the EP characterization corresponds to no significant waveform change. For example, when there is no positioning effect. Poor classification may indicate that EP characterization corresponds to significant waveform variation. For example when having a positioning effect. Uncertain classification may indicate that EP characterization may be of uncertain significance. For example, EP characterization may not be sufficient to classify well, but not well. For example, an EP may correspond to a localization effect or no localization effect. Unreliable classification may indicate that an EP contains too much noise to be properly characterized and classified.

Each classification may correspond to a particular threshold. The threshold may indicate how similar the EP response should be to a baseline considered a good response or how different the EP response should be from a baseline considered a poor response. The threshold may be based on characterization of the EP response. For example, the threshold may be based on at least one of euclidean distance, pseudo-correlation, cross-correlation, or energy ratio between EP response and baseline. The threshold may also indicate how much noise may be included in the EP response before the EP response is deemed unreliable.

The threshold for classification may be determined by analyzing the training data. The training data may include a plurality of EP responses known to correspond to a particular classification. Using a plurality of sets of thresholds determined from the analysis of the training data, the current response may be classified as belonging to the category of interest based on the metric values computed for it.

Fig. 4B illustrates an exemplary flowchart of a process for determining whether a change has occurred in a sequence of EPs according to an exemplary embodiment of the present invention. FIG. 4B continues with FIG. 4A. Given a sequence of classified EPs, it can be determined whether the state of the physiological system of interest has changed (from good to bad or vice versa) or whether the state of the physiological system of interest has not changed. If the status has changed, the system may establish an alarm.

FIG. 5 illustrates an exemplary flow chart of a process for calculating a baseline response in accordance with an exemplary embodiment of the present invention. The currently loaded response may be repeatedly represented as a node within a Minimum Spanning Tree (MST) established using euclidean distances between pairs of responses. Each line in the MST of the linked response pair may represent a euclidean distance value. The currently loaded response may be the initially acquired response. The response pair may be a combination of the responses of any two current loads. For example, three responses may result in three response pairs. The euclidean distance may be based on the sum of the squares of the differences between the responses in each response pair or the sum of the absolute values of the differences between the responses in each response pair.

The MST may be divided into clusters (clusters) based on a dividing line (trimming line) that is greater than a threshold. The threshold value may be based on an average of the line lengths and a standard deviation of the line lengths. Clusters may be classified based on the size of the cluster. The size of the cluster may be the number of responses within the cluster. The cluster with the largest size may be selected such that a temporary baseline is calculated based on the responses within the cluster. All responses within the largest cluster may be aligned using the default analysis scope and pseudo-correlation. The response members of the cluster with the most members may be averaged to estimate the baseline response.

Fig. 6 shows an exemplary flowchart of a process for determining an analysis range according to an exemplary embodiment of the present invention. Initial responses were characterized and classified using initial baseline response estimates and default analysis ranges. First, an initial good response is used to locate the default width analysis range by adjusting the position of the range until a minimum consistency value is obtained. Using the initial good response, the width of the analysis range is then adjusted by increasing it to the left or right until a minimum consistency value is obtained. For analysis range position and size, the consistency metric may be:

where NormED is the normalized Euclidean distance and CC is the cross-correlation. Although not shown in fig. 6, the calculated new baseline response may be used to recalculate the analysis range.

FIG. 7 illustrates an exemplary flow chart of a process for updating a baseline response according to an exemplary embodiment of the invention. As shown in fig. 7, if the previous response is classified as good, the current baseline may be recalculated based on the previous response and the previous baseline. For example, the current baseline may be set to 25% of the previous response and 75% of the previous baseline. If the previous response is not classified as good, the current baseline may be set to the previous baseline.

Regardless of how the new current baseline is determined, the new current baseline may be used to realign the current response relative to the new current baseline. Metric calculations may then be performed on the realigned responses.

Fig. 8 illustrates an exemplary embodiment of a relationship diagram in metric calculations for characterizing an EP, according to an exemplary embodiment of the present invention. As shown in fig. 8, the current response may be compared with previous responses to give euclidean distances, pseudo-correlations, and cross-correlations between the responses. The current response may be compared to the current baseline to give euclidean distance, pseudo-correlation, cross-correlation, and energy ratio between the response and the baseline. The current response may be classified based on these different results.

After acquiring the next response, the current response may also be used to give euclidean distance, pseudo-correlation and cross-correlation between the current response and the next response.

Fig. 9 shows an exemplary flowchart of a procedure for a good state according to an exemplary embodiment of the present invention. If in good conditionWhen a bad response is received, the system may check to see if the bad count is greater than or equal to the bad count threshold N_B. The bad counts may indicate many bad responses. Bad count threshold N_BThe number of bad responses or indeterminate responses received before the next bad response changes state to a bad state may be indicated. A bad count threshold N may be set for each state according to the physiological system of interest_B。

If the bad count is greater than the bad count threshold N_BThen the current state may change to a bad state and an alarm may be established. The alert may be delivered to a user of the system in a variety of ways (e.g., by displaying a visualization, generating a sound, generating a vibration, etc.). If the bad count is not greater than the bad count threshold N_BThen the bad count may be incremented and a bad response added to a bad tracker (bad tracker). The bad tracker may track bad responses as well as received uncertain responses.

If the received response is not a bad response, the system may check if the received response is an indeterminate response. If the received response is an indeterminate response, the bad count is also incremented and the indeterminate response is added to the bad tracker.

If the received response is also not an indeterminate response, the system may check if the received response is a good response. If the received response is a good response, then if the bad count is less than or equal to the bad count threshold N_BThen the bad count is reset to zero and the bad tracker is cleared. If the bad count is greater than the bad count threshold N_BThen the good count may be incremented and a good response added to the good tracker.

If the received response is also not a good response, the system may determine that the response is an unreliable response and may ignore the response.

Based on the bad count, the bad tracker, the good count, and the good tracker, the system may provide different indications to the user. The system may change the color of the displayed icon such that the icon appears green when the failure count is zero and gradually becomes more red as the value of the failure tracker increases.

Fig. 10 shows an exemplary flowchart of a procedure for a bad state according to an exemplary embodiment of the present invention. If a good response is received while in the bad state, the system may increment a good count, and if the bad count is less than a bad count threshold N_BThe bad tracker check is cleared.

The system may check to see if the good count is greater than or equal to the good count threshold N_G. A good count may indicate the number of good responses. Good count threshold N_GMay indicate the number of good responses that need to be received to change the status to good. Good count threshold N_GSettings may be made for each state according to the physiological system of interest. If the good count is greater than the good count threshold N_GThen the current state may be changed to a good state and an alarm may be established. If the good count is not greater than the good count threshold N_GThen a good response may be added to the good tracker. A good tracker may track a good response of the reception.

If the received response is not a good response, the system may check if the received response is an indeterminate response. If the received response is an indeterminate response, the bad count is incremented and the indeterminate response is added to the bad tracker.

If the received response is also not an indeterminate response, the system may check whether the received response is a bad response. If the received response is a bad response, if the good count is less than or equal to the good count threshold N_GThen the good count is reset to zero and the good tracker is cleared. If the good count is greater than the good count threshold N_GThen the bad count may be incremented and a bad response added to the bad tracker.

If the received response is also not a bad response, the system may determine that the response is an unreliable response and may ignore the response.

A signal processing procedure may be applied to reduce noise in acquired EPs and to detect when EPs with inadequate signal-to-noise ratios (SNRs) are acquired so that these EPs may be excluded from further analysis and reported poor signal quality. The number of unreliable signals received may be tracked and compared to a threshold to determine when to generate an alert regarding poor signal quality.

The applied filtering technique may use an average-based likelihood estimate to reduce instrumental and background-based noise and improve the SNR of the obtained EP, so that the baseline EP may be defined more clearly, and changes in EP may then be better characterized and compared to the baseline and previous EPs.

Pattern recognition algorithms can be used to characterize EPs to measure changes in the latter obtained EPs from baseline and previous EPs and to detect when changes in EP occur that are indicative of altered function of the underlying sensory nervous system. The EPs may be characterized using their energy, euclidean distance and pseudo-correlation with respect to a defined baseline template response and previous EPs. Using these metrics, classification rules can be applied to determine whether the current response indicates a significant (adverse or improvement) change in the underlying physiological system that generated the EP.

In embodiments, components may be added to allow medical or other participating personnel to reset the baseline response when changes in the acquired EP are not correlated with any underlying physiological changes (e.g., changes related to stimulation or electrode factors).

In an embodiment, the system may be an automated EP analysis device for monitoring, detecting, and identifying changes (poor or good) in the physiological system that produces the analyzed EP, wherein the device is adapted to characterize and classify the EP and establish an alert for changes (poor or good) in the physiological system that produces the EP when the acquired EP waveform changes significantly in delay, amplitude, or morphology. The system may also include a system for integrating such apparatus into other devices in a surgical environment.

The apparatus may also feed information to other devices in the surgical environment that allows these devices to manually or automatically ameliorate or mitigate physiological changes and improve subsequently acquired EP waveforms.

The device may also obtain information from the anesthesia or sphygmomanometer that is used to calculate when changes in the EP waveform are due to anesthesia or blood pressure changes.

The apparatus may perform a method of automatically identifying potential damage to a peripheral nerve structure, comprising: stimulating peripheral nerves with electrical pulses; recording the resultant electrical waveforms generated by the nervous system via electrodes placed on the neck or head; measuring a change or trend in the obtained EP waveform; alerting the user to the change; allowing a user to select to decide whether the data is accurate; communicating the information to an automated surgery room table; and automatically or semi-automatically readjust the patient's position to mitigate or avoid injury through adjustment of the table.

While various embodiments of the present invention have been described above, it should be understood that they have been presented by way of example only, and not limitation. Thus, the breadth and scope of the present invention should not be limited by any of the above-described exemplary embodiments, but should be defined only in accordance with the following claims appended hereto and their equivalents.

Claims (4)

1. An automated Evoked Potential (EP) analysis apparatus for automated baseline acquisition and subsequent monitoring, detection and identification of changes in the nervous system that produce an analyzed EP, wherein the apparatus is adapted to:

providing an electrical pulse to stimulate a peripheral nerve structure;

recording, by electrodes placed on the body, a synthetic EP waveform generated by the nervous system;

measuring a change in the recorded EP waveform relative to a baseline response;

comparing the change to one or more threshold change values to classify the recorded EP waveform; and

determining whether a current state of the peripheral neural structure has changed by determining whether a count of EP waveforms in a classification exceeds a count threshold based on the classification of the recorded EP waveforms.

2. The apparatus of claim 1, wherein the apparatus is integrated with other devices in a surgical environment.

3. The apparatus of claim 1, wherein the apparatus is configured to feed information to other devices in a surgical environment that allows these devices to manually or automatically ameliorate or mitigate physiological changes and improve subsequently acquired EP waveforms.

4. The apparatus of claim 1, wherein the apparatus is configured to obtain information from anesthesia or blood pressure machines for calculating when changes in the EP waveform are due to anesthesia or blood pressure changes.