Method of physiological data analysis and measurement quality check using principal component analysis

• US 7,769,434 B2
• Filed: 11/30/2006
• Issued: 08/03/2010
• Est. Priority Date: 11/30/2006
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Xue, Joel Qiuzhen
• Primary Examiner(s): Layno; Carl H
• Assistant Examiner(s): Bays; Pamela M
• Application Number:
  US11/564,870
• Publication Number:
  US 20080132799A1
• Time in Patent Office: 1,342 Days
• Field of Search: 600/509, 600/513, 600/516, 600/517, 600/515
• US Class Current: 600/509
• CPC Class Codes: A61B 5/316 Modalities, i.e. specific diagnostic methods
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients

Abstract

In a method of analyzing patient physiological data, the data is subjected to principal component analysis and compared to a model physiological data principal component analysis. The comparison is used to identify correlations present in the morphology of the patient physiological data. The present invention further includes determining a confidence interval for the detection of a morphological feature and utilizing this confidence interval for improving the quality of the detection of morphological features of the patient physiological data, including automated morphological feature identification.

Claims

1. A computer-implemented method of analyzing patient physiological data collected as part of a pharmacological study, comprising a non-transitory computer-readable medium programmed with computer readable code, such that when a computer processor executes the computer readable code, the processor performs the steps of:

receiving patient physiological data;

normalizing the patient physiological data;

performing a principal component analysis of the normalized patient physiological data;

comparing the principal component analysis of the normalized patient physiological data to a model patient physiological data principal component analysis, wherein the model physiological data principal component analysis is representative of a specific physiological morphological feature;

using the comparison to identify if the specific physiological morphological feature is present in the patient physiological data.

2. The method of claim 1 wherein the patient physiological data is patient ECG signal data.

3. The method of claim 2 wherein the model patient physiological data principal component analysis is created by a principal component analysis of a plurality of ECG signal data exhibiting the specific morphological feature.

4. The method of claim 3 wherein a model physiological data principal component analysis is created using a plurality of edited normal and abnormal ECG signals.

5. The method of claim 3 wherein the specific physiological morphological a T-wave characteristic.

6. The method of claim 3 wherein the specific physiological morphological feature comprises at least one of a first upslope line, a second upslope line, a T peak line, and a down slope line.

7. The method of claim 2 further comprising the step of sorting the patient ECG signal data into morphological bins based upon the presence of the specific morphological feature identified in the comparison of the patient ECG signal principal component analysis to the model physiological data principal component analysis.

8. The method of claim 7 further comprising the step of performing a statistical analysis on the ECG signals in each morphological feature bin to determine the quality of an individual signal within the bin.

9. The method of claim 7 further comprising the step of determining a confidence interval for each morphological feature bin, the confidence interval reflecting the amount of variance in the specific morphological feature in the patient ECG signal data.

10. The method of claim 9 further comprising the step of rejecting patient ECG signal data based upon its deviation from the other patient ECG signal data within the same bin.

11. The method of claim 7 further comprising the step of learning identifying characteristics of each bin and using the learned identifying characteristics to automatically detect and identify desired predetermined selected signal morphology in the analysis of an ECG signal.

12. The method of claim 9 further comprising creating a morphological identification algorithm for a plurality of the patient ECG signal data in a morphological bin.

13. The method of claim 12 wherein the morphological identification algorithm is an average of the patient ECG signal data in the morphological bin.

14. The method of claim 13 further comprising applying the morphological identification algorithm to newly collected patient ECG signal data to determine if the specific morphological feature is present in the newly collected patient ECG signal data.

15. A method of analyzing an ECG signal collected as part of a pharmacological drug trial, the method comprising the steps of:

collecting ECG signal data;

performing principal component analysis on the collected data;

identifying characteristics of the ECG signal data by comparing the principal component analysis of the ECG signal data to a plurality of model principal component analysis, each of a specific morphological feature;

sorting the ECG signal data into groups based upon the identified characteristics;

performing a statistical analysis on the ECG signal data groups to determine an indication of ECG signal data group quality;

using the ECG signal data group and the indication of ECG signal data group quality to derive a morphological detection algorithm.

16. The method of claim 15 wherein the step of collecting ECG signal data comprises receiving the ECG signal data from a database of collected ECG signal data.

17. The method of claim 15 wherein the indication of signal quality is a confidence interval for detecting a specific morphological feature in collected ECG signal data.

18. The method of claim 17 further comprising:

receiving new ECG signal data;

applying the morphological detection algorithm to the ECG signal data to determine if the specific morphological feature is present in the new ECG signal data.

19. The method of claim 18 wherein the specific morphological feature is Torsade-de-Pointes.

20. A computer program embodied on a non-transitory computer-readable medium having computer logic for enabling at least one processor in a computer system to facilitate analyzing an ECG signal collected as part of a pharmacological drug trial, the non-transitory computer-readable medium comprising:

a first computer readable program code portion for collecting ECG signal data;

a second computer readable program code portion that perform a principal component analysis (PCA) of the collected ECG signal data to obtain a collected ECG signal data PCA, comparing the collected ECG signal data PCA to a plurality of model morphological feature PCA, and sorts the collected ECG signal data into groups based upon to which of the plurality of model morphological feature PCA the collected ECG signal data PCA is the most similar;

a third computer readable program code portion creating a model of the ECG signal data in each of the groups; and

a fourth computer readable program code portion using the model to detect a morphological feature of a future ECG signal.