Method and apparatus for beat-space frequency domain prediction of cardiovascular death after acute coronary event

• US 9,295,397 B2
• Filed: 06/16/2014
• Issued: 03/29/2016
• Est. Priority Date: 06/14/2013
• Status: Active Grant

• Current Assignee: Massachusetts Institute of Technology
• Original Assignee: Massachusetts Institute of Technology
• Inventors: Liu, Yun, Guttag, John V., Stultz, Collin M.
• Primary Examiner(s): Wieland, Robert N
• Application Number:
  US14/305,260
• Publication Number:
  US 20140371610A1
• Time in Patent Office: 652 Days
• US Class Current: 1/1
• CPC Class Codes: A61B 5/0816 Measuring devices for examining respiratory frequency measuring frequency of electric signals G01R23/00
• CPC Class Codes: A61B 5/316 Modalities, i.e. specific diagnostic methods
• CPC Class Codes: A61B 5/347 Detecting the frequency distribution of signals
• CPC Class Codes: A61B 5/7267 involving training the classification device
• 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: G16H 50/30 for calculating health indices; for individual health risk assessment

Abstract

A method for the machine generation of a model for predicting patient outcome following the occurrence of an event. In one embodiment the method includes the steps of obtaining a physiological signal of interest, the physiological signal having a characteristic; obtaining a time series of a signal characteristic; dividing the time series into a plurality of window segments; converting the time series from time-space to beat-space; computing the power in various frequency bands of each window segment; computing the 90th percentile of the spectral energies across all window segments for each frequency band; and inputting the data into a machine learning program to generate a weighted risk vector.

Claims

1. A method for the machine generation of a model for predicting death within a predetermined period following the occurrence of a cardiac event, the method comprising the steps of:

obtaining an ECG of interest from a plurality of patients having a plurality of outcomes, the ECG having a plurality of ECG characteristics;

obtaining a time series of one of the ECG characteristics of the plurality of ECG characteristics;

dividing the time series into a plurality of window segments;

converting the time series from time-space to beat-space;

computing the power in various frequency bands of each window segment;

computing the 90th percentile of the spectral energies across all window segments for each frequency band to generate spectral energy data; and

inputting the spectral energy data into a machine learning program to obtain a weighted risk vector.

2. The method of claim 1 further comprising the step of removing noise from the time series.

3. The method of claim 1 wherein the machine learning program is an L1 regularized logistic regression machine learning program.

4. A method for the machine generation of a model for predicting patient outcome following the occurrence of a medical event, the method comprising the steps of:

obtaining a physiological signal of interest from a plurality of patients having a plurality of outcomes, the physiological signal having a characteristic;

obtaining a time series of a signal characteristic;

dividing the time series into a plurality of window segments;

converting the time series from time-space to an inverse physiological event space;

computing the power spectrum for a characteristic frequency band for each window segment;

computing the 90th percentile of the spectral energies across all window segments for each frequency band to generate spectral energy data; and

inputting the spectral energy data into the L1-regularized logistic regression machine learning program to generate a weighted risk vector.

5. A method of predicting death within a predetermined period, of a patient following the occurrence of a cardiac event, the method comprising the steps of:

obtaining, using an ECG module, an ECG of the patient of interest, the ECG having a plurality of ECG characteristics;

obtaining, using a processor in communication with the ECG module, a time series of one of the ECG characteristics of the plurality of ECG characteristics;

dividing, using the processor, the time series into a plurality of window segments;

converting, using the processor, the time series from time-space to beat-space;

computing, using the processor, the power in various frequency bands of each window segment;

computing, using the processor, the 90th percentile of the spectral energies across all window segments for each frequency band;

computing, using the processor, a risk by multiplying the 90th percentile spectral energies by a weighted risk vector and comparing the result to a predetermined value, and

displaying, on a display, a prediction of the death of the patient in response to the comparison.