Methods for consistent forewarning of critical events across multiple data channels

• US 20040087835A1
• Filed: 07/12/2002
• Published: 05/06/2004
• Est. Priority Date: 07/12/2002
• Status: Active Grant

• Current Assignee: UT-Battelle LLC
• Original Assignee: UT-Battelle LLC
• Inventors: Hively, Lee M.
• Granted Patent: US 7,139,677 B2
• Time in Patent Office: Days
• US Class Current: 600/300
• CPC Class Codes: A61B 5/4094 Diagnosing or monitoring seizure diseases, e.g. epilepsy
• CPC Class Codes: A61B 5/7275 Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
• CPC Class Codes: G06F 2218/12 Classification; Matching
• CPC Class Codes: G16H 10/60 for patient-specific data, e.g. for electronic patient records
• CPC Class Codes: G16H 50/50 for simulation or modelling of medical disorders

Abstract

This invention teaches further method improvements to forewarn of critical events via phase-space dissimilarity analysis of data from biomedical equipment, mechanical devices, and other physical processes. One improvement involves conversion of time-serial data into equiprobable symbols. A second improvement is a method to maximize the channel-consistent total-true rate of forewarning from a plurality of data channels over multiple data sets from the same patient or process. This total-true rate requires resolution of the forewarning indications into true positives, true negatives, false positives and false negatives. A third improvement is the use of various objective functions, as derived from the phase-space dissimilarity measures, to give the best forewarning indication. A fourth improvement uses various search strategies over the phase-space analysis parameters to maximize said objective functions. A fifth improvement shows the usefulness of the method for various biomedical and machine applications.

Claims

1. A method for processing data to provide a forewarning of a critical event, comprising:

acquiring a plurality of sets of data with a plurality of channels of data for at least one test subject or process;

computing a renormalized measure of dissimilarity for distribution functions derived from a connected phase space for each respective channel of data;

comparing said renormalized measure of dissimilarity to a threshold (U_C) for a number of occurrences (N_OCC) to indicate a condition change in said renormalized measure of dissimilarity;

detecting a simultaneous condition change in a plurality (N_SIM) of renormalized measures of dissimilarity to determine a forewarning of the critical event;

determining true positive, true negative, false positive and false negative indications of condition change forewarning of the critical event for each channel of data in the plurality of sets of data;

calculating a total true rate for forewarning indications for each channel of data; and

comparing the total true rates for respective channels to determine at least one channel with the greatest channel-consistent total-true rate in said at least one channel.

2. The method of claim 1, wherein the test subject is a human patient.

3. The method of claim 1, wherein the test subject is a mechanical device or physical process.

4. The method of claim 1, further comprising:

testing a plurality of parameters for each channel to determine optimum values for the parameters corresponding to a highest channel-consistent total-true rate for a respective channel; and

setting the plurality of parameters to the optimum values for processing data from other channels of data.

5. The method of claim 1, wherein the connected phase space is constructed by computing equiprobable symbols for the data in the data sets.

6. The method of claim 1, wherein the total true rate is calculated as Σ

_i(TP_i+TN_i)/Σ

_i(TP_i+TN_i+FP_i+FN_i), where TP are true positives, TN are true negatives, FP are false positives and FN are false negative forewarning indications and wherein “

i”

is the data set number.

7. The method of claim 1, further comprising determining a sequence of renormalized phase space dissimilarity measures from data sets collected during increasingly severe fault conditions;

summing said renormalized measures into a composite measure, C_i, for the i-th data set;

performing a least-squares analysis over a window of m points of the said composite measure to obtain a straight line, y_i=ai+b, that best fits said composite data in a least-squares sense;

determining the variance, σ

₁²=Σ

_i, (y_i−

C_i)²/(m−

1), of said composite measure with respect to the straight line fit;

obtaining the variance of a sequel window of n sequential points via the statistic, χ

²=Σ

_i(y_i−

C_i)²/σ

₁²;

comparing said value of χ

² to the maximal value of the same statistic, χ

²(C_i) for a window of n sequential points from said C_i values; and

determining the onset of a critical event, such as a machine failure, when χ

² is significantly more than χ

²(C_i).

8. The method of claim 1, further comprising:

selecting a set of parameter values (N, w, S, and d) for computing the measures of dissimilarity for distribution functions in connected phase space for the data sets to be processed; and

searching over the values of the forewarning threshold (U_C) and a corresponding number of occurrences (N_OCC) for each channel to find the best channel for forewarning indication.

9. The method of claim 8, wherein the connected phase space is constructed by computing equiprobable symbols for the data in the data sets.

10. The method of claim 1, wherein

a plurality of renormalized measures of dissimilarity are computed for distribution functions derived from a connected phase space for each respective channel of data;

and

wherein said plurality of renormalized measures of dissimilarity are compared to respective thresholds to indicate respective condition change forewarning of the critical event.

11. The method of claim 10, wherein

a second plurality of renormalized measures of dissimilarity are also computed for distribution functions derived from an unconnected phase space for each respective channel of data;

and

wherein said second plurality of renormalized measures of dissimilarity are compared to respective thresholds to indicate respective condition change forewarning of the critical event.

12. The method of claim 1, wherein

a plurality of renormalized measures of dissimilarity are computed for distribution functions derived from a connected phase space for each respective channel of data;

wherein said renormalized measures of dissimilarity are summed to provide a composite measure of dissimilarity; and

wherein said composite measure of dissimilarity is compared to a threshold to indicate a respective condition change forewarning of the critical event.

13. The method of claim 12, further comprising:

computing a chi-squared statistic, χ

²=Σ

_i(y₁−

C_i)²/σ

₁², for the composite dissimilarity measure;

testing a plurality of parameters for each channel to determine optimum values for the parameters corresponding to a largest value of χ

² for a respective channel; and

setting the plurality of parameters to the optimum values for processing data from other channels of data.