Diagnostic classifications of pulse signal waveform data

• US 9,220,437 B2
• Filed: 07/23/2010
• Issued: 12/29/2015
• Est. Priority Date: 05/14/2010
• Status: Expired due to Fees

• Current Assignee: Centre For Development of Advanced Computing
• Original Assignee: Centre For Development of Advanced Computing
• Inventors: Dhurandhar, Medha Sanjeev, Tillu, Girish Shrikrishna
• Primary Examiner(s): Mallari, Patricia
• Assistant Examiner(s): Ramachandran, Vasuda
• Application Number:
  US13/256,568
• Publication Number:
  US 20120184861A1
• Time in Patent Office: 1,985 Days
• Field of Search: 600500-501
• US Class Current: 1/1
• CPC Class Codes: A61B 5/02007 Evaluating blood vessel condition, e.g. elasticity, compliance
• CPC Class Codes: A61B 5/0205 Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition
• CPC Class Codes: A61B 5/0245 by using sensing means generating electric signals, i.e. ECG signals
• CPC Class Codes: A61B 5/0535 Impedance plethysmography for measuring blood flow A61B5/0295
• CPC Class Codes: A61B 5/681 Wristwatch-type devices
• CPC Class Codes: A61B 5/6824 Arm or wrist
• CPC Class Codes: A61B 5/7257 using Fourier transforms
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

Diagnostic classifications of pulse signal waveform data is provided. In one example, diagnosis or prediction of a disease may be performed by analyzing pulse signal waveform data, and comparing aspects of the pulse signal waveform data with a morphology pattern that has been found to indicate a subject is suffering from a specific disease. A pulse signal can be captured via wrist electrodes using bio-electric sensors based on an impedance plethysmographic principle, for example, and processed using feature extraction for diagnosis and assessment of the subject'"'"'s proneness to a disease. Analysis of the pulse signal and pulse morphology patterns provides an in-vitro, non-invasive, and low-cost method for diagnosing diseases.

Claims

1. A method of making diagnostic classifications of pulse signal waveform data, the method comprising:

receiving, at a processor, pulse signal waveform data;

identifying, by the processor, systolic peak data points in the pulse signal waveform data corresponding to a systolic peak indicating a heartbeat;

based on the systolic peak data points, identifying, by the processor, complexes in the pulse signal waveform data, wherein a complex comprises a sub-waveform in the pulse signal waveform data approximately between an on-set data point and an off-set data point, wherein the on-set data point is a data point preceding a given systolic peak data point that has a local minimum amplitude and the off-set data point is a data point subsequent to the given systolic peak that has a local minimum amplitude, and wherein each complex includes deflection data points between the on-set point and the off-set point, wherein one of the deflection data points is a systolic peak corresponding to a heartbeat;

determining, by the processor, a pattern distribution of the complexes based on relative magnitudes of at least three deflection data points of the complexes, wherein the pattern distribution indicates a frequency of pulse patterns appearing in the complexes of the pulse signal waveform data; and

based on the pattern distribution, making, by the processor, a diagnostic classification of the pulse signal waveform data;

wherein making the diagnostic classification of the pulse signal waveform data comprises determining a relationship between a first previous data point preceding the systolic peak, a first subsequent data point following the systolic peak, and a third subsequent data point following the systolic peak and a second subsequent data point;

wherein making the diagnostic classification of the pulse signal waveform data comprises determining that the pulse signal waveform data is indicative of Diabetes Mellitus when the following relationship is substantially satisfied;

A≧

X>

Y′

,

where A is the first previous data point preceding the systolic peak, X is the first subsequent data point following the systolic peak, and Y′

is the third subsequent data point following the systolic peak and the second subsequent data point.

2. The method of claim 1, wherein making the diagnostic classification of the pulse signal waveform data comprises determining that the pulse signal waveform data is indicative of Diabetes Mellitus or Hyperglycemia induced vascular changes.

3. The method of claim 1, further comprising for each complex determining the deflection data points by:

identifying the first previous data point preceding the systolic peak data point that has a local minimum amplitude, the first preceding data point being the on-set data point;

identifying the first subsequent data point following the systolic peak data point that has a local minimum amplitude;

identifying the second subsequent data point following the first subsequent data point that has a local maximum amplitude;

identifying the third subsequent data point following the second subsequent data point that has a local minimum amplitude;

identifying a fourth subsequent data point following the third subsequent data point that has a local maximum amplitude; and

identifying a fifth subsequent data point following the fourth subsequent data point that has a local minimum amplitude, the fifth subsequent data point being the off- set data point.

4. The method of claim 1, further comprising receiving personal health and medical data of a patient from which the pulse signal waveform data was recorded, and based on the personal health and medical data of the patient, making the diagnostic classification of the pulse signal waveform data.

5. The method of claim 1, further comprising converting the pulse signal waveform data into time series data that is stored in ASCII format.

6. The method of claim 5, further comprising filtering the pulse signal waveform data by:

performing a discrete Fourier transform on the time series data to decompose data points into components of different frequencies;

determining a power spectral density of the decomposed data points;

identifying frequencies at which a signal power is concentrated;

based on the identified frequencies, determining a threshold value reflecting low signal power;

setting decomposed data points at frequencies that have less signal power than the threshold value to zero to provide modified data points; and

performing an inverse fast Fourier transform (iFFT) to the modified data points to output de-noised pulse signal waveform data.

7. The method of claim 1, further comprising identifying systolic peaks corresponding to heart beats in each complex by identifying a data point in each complex having a maximum amplitude.

8. The method of claim 1, further comprising based on relationships between the deflection data points, characterizing each complex in the pulse signal waveform data.

9. The method of claim 1, further comprising identifying within the pattern distribution of complexes a pattern appearing most frequently in the pulse signal waveform data.

10. The method of claim 9, further comprising comparing the pattern appearing most frequently in the pulse signal waveform data with a pattern classified as being indicative of a disease.

11. The method of claim 1, further comprising comparing deflection data points of patterns of complexes within the pattern distribution of complexes with deflection data points in known patterns of pulse signals to identify matches.

12. The method of claim 11, further comprising when deflection data points of a given pattern of a complex do not sufficiently match with deflection data points in any of the known patterns of pulse signals, populating a database of distinct pulse signal patterns observed in pulse signal waveform data with the given pattern.

13. The method of claim 1, wherein receiving pulse signal waveform data comprises receiving Impedance phlebography/impedance plethysmography based pulse signal data as a function of time.

14. The method of claim 1, wherein receiving pulse signal waveform data comprises receiving pulse signal waveform data collected over a time period, and wherein a time interval of the sub-waveform in the pulse signal waveform data is selected based on a patient'"'"'s resting heart rate so that approximately one systolic peak indicating the heart beat occurs during the time interval of the sub-waveform.

15. A non-transitory computer readable medium having stored therein instructions executable by a computing device to cause the computing device to perform functions of:

receiving pulse signal waveform data;

identifying systolic peak data points in the pulse signal waveform data corresponding to a systolic peak indicating a heartbeat;

based on the systolic peak data points, identifying complexes in the pulse signal waveform data, wherein a complex comprises a sub-waveform in the pulse signal waveform data approximately between an on-set data point and an off-set data point, wherein the on-set data point is a data point preceding a given systolic peak data point that has a local minimum amplitude and the off-set data point is a data point subsequent to the given systolic peak that has a local minimum amplitude, and wherein each complex includes deflection data points between the on-set point and the off-set point, wherein one of the deflection data points is a systolic peak corresponding to a heartbeat;

determining a pattern distribution of the complexes based on relative magnitudes of at least three deflection data points of the complexes, wherein the pattern distribution indicates a frequency of pulse patterns appearing in the complexes of the pulse signal waveform data; and

based on the pattern distribution, making a diagnostic classification of the pulse signal waveform data;

wherein making the diagnostic classification of the pulse signal waveform data comprises determining a relationship between a first previous data point preceding the systolic peak, a first subsequent data point following the systolic peak, and a third subsequent data point following the systolic peak and a second subsequent data point;

wherein making the diagnostic classification of the pulse signal waveform data comprises determining that the pulse signal waveform data is indicative of Diabetes Mellitus when the following relationship is substantially satisfied;

A≧

X>

Y′

,

where A is the first previous data point, X is the first subsequent data point, and Y′

is the third subsequent data point.

16. The non-transitory computer readable medium of claim 15, wherein the functions further include determining that the pulse signal waveform data is indicative of Diabetes Mellitus.

18. The non-transitory computer readable medium of claim 15, wherein the functions further include:

identifying within the pattern distribution of complexes a pattern appearing most frequently in the pulse signal waveform data; and

comparing the pattern appearing most frequently in the pulse signal waveform data with a pattern classified as being indicative of a disease.

17. A non-transitory computer readable medium having stored therein instructions executable by a computing device to cause the computing device to perform functions of:

receiving pulse signal waveform data;

identifying systolic peak data points in the pulse signal waveform data corresponding to a systolic peak indicating a heartbeat;

based on the systolic peak data points, identifying complexes in the pulse signal waveform data, wherein a complex comprises a sub-waveform in the pulse signal waveform data approximately between an on-set data point and an off-set data point, wherein the on-set data point is a data point preceding a given systolic peak data point that has a local minimum amplitude and the off-set data point is a data point subsequent to the given systolic peak that has a local minimum amplitude, and wherein each complex includes deflection data points between the on-set point and the off-set point, wherein one of the deflection data points is a systolic peak corresponding to a heartbeat;

identifying a first previous data point preceding the systolic peak data point that has a local minimum amplitude, the first preceding data point being the on-set data point;

identifying a first subsequent data point following the systolic peak data point that has a local minimum amplitude;

identifying a second subsequent data point following the first subsequent data point that has a local maximum amplitude;

identifying a third subsequent data point following the second subsequent data point that has a local minimum amplitude;

identifying a fourth subsequent data point following the third subsequent data point that has a local maximum amplitude; and

identifying a fifth subsequent data point following the fourth subsequent data point that has a local minimum amplitude, the fifth subsequent data point being the off- set data point,

determining a pattern distribution of the complexes based on relative magnitudes of at least three deflection data points of the complexes, wherein the pattern distribution indicates a frequency of pulse patterns appearing in the complexes of the pulse signal waveform data; and

based on the pattern distribution, making a diagnostic classification of the pulse signal waveform data;

wherein making the diagnostic classification of the pulse signal waveform data comprises determining that the pulse signal waveform data is indicative of Diabetes Mellitus when the following relationship is substantially satisfied;

A>

X>

Y′

,

where A is the first previous data point, X is the first subsequent data point, and Y′

is the third subsequent data point.

19. A system comprising:

a database storing known patterns of pulse signals;

an input interface for receiving pulse signal waveform data; and

a processor configured to identify systolic peak data points in the pulse signal waveform data corresponding to a systolic peak indicating a heartbeat, the processor further based on the systolic peak data points configured to identify complexes in the pulse signal waveform data, wherein a complex comprises a sub-waveform in the pulse signal waveform data approximately between an on-set data point and an off-set data point, wherein the on-set data point is a data point preceding a given systolic peak data point that has a local minimum amplitude and the off-set data point is a data point subsequent to the given systolic peak that has a local minimum amplitude, and wherein each complex includes deflection data points between the on-set point and the off-set point, wherein one of the deflection data points is a systolic peak corresponding to a heartbeat,

wherein based on relative magnitudes of at least three deflection data points of the complexes, the processor is configured to compare a pattern of each complex with the known patterns of pulse signals in the database, and to make a diagnostic classification of the pulse signal waveform data;

wherein making the diagnostic classification of the pulse signal waveform data comprises determining a relationship between a first previous data point preceding the systolic peak, a first subsequent data point following the systolic peak, and a third subsequent data point following the systolic peak and a second subsequent data point;

wherein making the diagnostic classification of the pulse signal waveform data comprises determining that the pulse signal waveform data is indicative of Diabetes Mellitus when the following relationship is substantially satisfied;

A≧

X>

Y′

,

where A is the first previous data point, X is the first subsequent data point, and Y′

is the third subsequent data point.