Early detection of hemodynamic decompensation using taut-string transformation

• US 9,974,488 B2
• Filed: 06/26/2015
• Issued: 05/22/2018
• Est. Priority Date: 06/27/2014
• Status: Active Grant

• Current Assignee: Board of Regents of the University of Michigan (University of Michigan)
• Original Assignee: Board of Regents of the University of Michigan (University of Michigan)
• Inventors: Najarian, Kayvan, Belle, Ashwin, Ward, Kevin, Derksen, Harm
• Primary Examiner(s): Sereboff, Neal
• Application Number:
  US14/751,260
• Publication Number:
  US 20150374300A1
• Time in Patent Office: 1,061 Days
• CPC Class Codes: A61B 5/7253 characterised by using transforms
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

Techniques develop models for classification for physical conditions of a subject based on monitored physiologic signal data. The models for classification are determined from data transformed and feature extracted using a Taut-string transformation and in some instances using a further Stockwell-transformation, applied in parallel or in series. Physical conditions, specifying the state of hemodynamic stability and reflective of the cardiovascular and nervous systems, are thus modeled using these techniques.

Claims

1. A method of developing classification models for classifying one or more physical conditions of a subject, the method comprises:

receiving, from a sensor equipment configured to monitor the subject, raw signal data indicative of a physical state of the subject;

applying, in a signal processor, a windowing filter to the raw signal data to create a plurality of windows of the raw signal data;

applying, in the signal processor, a pre-processing to each of the plurality of windows of the raw signal data to filter each window by removing spurious noise induced on the raw signal data by the sensor equipment;

performing, in the signal processor, a signal decomposition on each window, where the signal decomposition comprises applying a Taut-string transformation to each window to produce Taut-string transformed signal data for each window;

performing, in the signal processor, a feature extraction on the Taut-string transformed signal data for each window, wherein the feature extraction is configured to identify one or more features for the raw signal data; and

based on the one or more features, developing at least one classification model for classifying the one or more physical conditions of the subject.

2. The method of claim 1, wherein applying the pre-processing comprises:

performing a baseline draft removal on the raw signal data; and

filtering the raw signal data to cancel noise based on motion.

3. The method of claim 1, wherein the one or more extracted features are selected from the group consisting of a mean of frequencies of each window, a sum of frequencies of each window, a product of frequencies of each window, standard deviation of frequencies of each window, range of frequencies of each window, mean of max frequencies of each window, and mean of absolute deviation of frequencies of each window.

4. The method of claim 1, wherein performing the signal decomposition on each window, further comprises:

applying a Stockwell transformation (S-Transformation) to the raw signal data in each window to produce an S-transformed data set.

5. The method of claim 4, wherein applying a Taut-string transformation to each window comprises applying the Taut-string transformation to the S-transformed data set.

6. The method of claim 4, wherein the raw signal data is electrocardiogram signal data, electroencephalogram data, pulse oximetry waveform data, arterial blood pressure waveform data, venous blood pressure waveform data, intracranial waveform data, or a combination thereof.

7. The method of claim 4, further comprising measuring the raw signal data using an electrocardiogram sensor, a respiratory rate sensor, a pulse oximetry sensor, photoplethysmography sensor, inspiration/expiration sensor, arterial blood pressure sensor, and venous blood pressure sensor, impedance sensor, piezoelectric sensor, a Doppler signal, an ultrasound sensor, or intracranial pressure sensor.

8. The method of claim 1, wherein the classification is of hemodynamic decomposition.

9. The method of claim 1, wherein the classification is of trauma, hemorrhage, sepsis, cardiogenic shock, stroke, cardiac arrest, brain injury, pain, arrhythmias, heart failure, infection, level of consciousness, and/or a combination thereof.

10. A therapeutic delivery system for administering a therapeutic treatment to a subject, the delivery system comprising:

a diagnostic apparatus implementing the method of claim 1; and

an administration apparatus comprising a therapeutic delivery mechanism in communication with a processor that controls delivery of the therapeutic treatment in response to a received patient status data from the diagnostic apparatus, the administration apparatus containing the processor and

(i) coupled to receive the raw signal data, in a closed loop manner,

(ii) implemented to store the one or more identified features of the raw signal data, and

(iii) applying the classification model to determine as severity of the one or more physical conditions for the subject

(iv) implemented to administer the therapeutic treatment, in response to the determined severity of the one or more physical conditions for the subject.

11. An apparatus comprising:

a memory; and

a signal processor coupled to store data on the memory, the signal processor is configured to

receive, from a sensor device configured to monitor a subject, raw signal data indicative of a physical state of the subject;

apply a windowing filter to the raw signal data to create a plurality of windows of the raw signal data;

apply a pre-processing to each of the plurality of windows of the raw signal data to filter each window by removing spurious noise induced on the raw signal data by the sensor equipment;

perform a signal decomposition on each window, where the signal decomposition comprises applying a Taut-string transformation to each window to produce Taut-string transformed signal data for each window;

perform a feature extraction on the Taut-string transformed signal data for each window, wherein the feature extraction is configured to identify one or more features for the raw signal data; and

based on the one or more features, develop at least one classification model for classifying the one or more physical conditions of the subject.

12. The apparatus of claim 11, wherein the signal processor is further configured to:

perform a baseline draft removal on the raw signal data; and

filter the raw signal data to cancel noise based on motion.

13. The apparatus of claim 11, wherein the one or more extracted features are selected from the group consisting of a mean of frequencies of each window, a sum of frequencies of each window, a product of frequencies of each window, standard deviation of frequencies of each window, range of frequencies of each window, mean of max frequencies of each window, and mean of absolute deviation of frequencies of each window.

14. The apparatus of claim 11, wherein the signal processor is further configured to:

apply a Stockwell transformation (S-Transformation) to the raw signal data in each window to produce an S-transformed data set.

15. The apparatus of claim 14, wherein the signal processor is further configured to apply the Taut-string transformation to the S-transformed data set.

16. The apparatus of claim 14, wherein the raw signal data is electrocardiogram signal data, electroencephalogram data, pulse oximetry waveform data, arterial blood pressure waveform data, venous blood pressure waveform data, intracranial waveform data, or a combination thereof.

17. The apparatus of claim 14, wherein the signal processor is further configured to measure the raw signal data using an electrocardiogram sensor, a respiratory rate sensor, a pulse oximetry sensor, photoplethysmography sensor, inspiration/expiration sensor, arterial blood pressure sensor, and venous blood pressure sensor, impedance sensor, piezoelectric sensor, a Doppler signal, an ultrasound sensor, or intracranial pressure sensor.

18. The apparatus of claim 17, wherein the classification is of hemodynamic decomposition.

19. The apparatus of claim 11, wherein the classification is of hemorrhage, sepsis, cardiogenic shock, stroke, cardiac arrest, brain injury, arrhythmias, heart failure, infection, level of consciousness, and/or a combination thereof.