Method and system for determining cardiovascular parameters

1. A method for determining a blood pressure of a patient, comprising:

receiving a plurality of images of a finger region of the user, wherein the images are recorded using an image recording device of a user device associated with the patient wherein the image recording device is in contact with the finger region of the user;

generating a photoplethymogram (PPG) dataset from the plurality of images;

reducing noise in the PPG dataset to generate a denoised PPG dataset by:

filtering the PPG dataset using a longpass filter configured to remove signals from the PPG dataset below 0.5 Hertz;

decomposing the PPG dataset into a set of modes using enhanced empirical mode decomposition;

extracting intrinsic modes from the set of modes; and

aggregating the intrinsic modes to generate the denoised PPG dataset;

segmenting the denoised PPG dataset into a segmented PPG dataset using moving average segmentation;

determining a processed PPG dataset by averaging a predetermined number of segments of the segmented PPG dataset;

extracting a set of fiducials from for each segment of the processed PPG dataset comprising:

calculating the first, the second, and the third derivative of the segment;

determining time points corresponding to zeros of the first, second, and third derivative;

determining values of the segment at the time points;

determining values of the first, second, and third derivatives of the segment at the time points;

determining a position on a patient cardiovascular manifold, relating a heart function, nervous system, and vascular system of the patient, based on the set of fiducials;

transforming the position on the patient cardiovascular manifold to a position on a universal cardiovascular manifold; and

determining the blood pressure of the patient based on a mapping from the position on the universal cardiovascular manifold to a blood pressure value;

wherein the set of fiducials comprises:

the values of the segmented PG dataset for time points corresponding to the first four zeros of the first and second derivatives;

the values of the first derivative of the processed PG dataset for the time points corresponding to the first four zeros of the first and second derivatives; and

the values of the second derivative of the processed PG dataset for the time points corresponding to the first four zeros of the first and second derivatives.

2. The method of claim 1 , wherein the predetermined number of segments comprises at least 10 segments of the PPG dataset.

3. The method of claim 1 , wherein at least one of the universal cardiovascular manifold or the transformation are determined by:

for each patient of a group of calibration patients:

measuring a blood pressure using at least one of a sphygmomanometer, a digital blood pressure monitor, radial artery tonometry, arterial catheter, or electrocardiogram;

contemporaneously with measuring the blood pressure, determining a set of fiducials corresponding to the calibration patient; and

determining a relationship between the measured blood pressure and the set of fiducials; and

combining the relationship for each patient to determine at least one of the universal cardiovascular manifold or the transformation.

4. The method of claim 1 , wherein extracting the set of fiducials from the processed PPG dataset comprises fitting the processed PPG dataset to a fitting function, wherein the set of fiducials comprises one or more fit parameters of the fitting function.

5. The method of claim 4 , wherein the fitting function comprises a plurality of gaussians.

6. The method of claim 4 , wherein the fit parameters are constrained to values of a hyperplane.

7. A method for determining a cardiovascular state of a patient, comprising:

receiving a plethymogram (PG) dataset measured at a finger region of the patient;

removing noise from the PG dataset by aggregating a set of intrinsic modes corresponding to the PG dataset;

segmenting the denoised PG dataset into a segmented PG dataset, wherein each segment corresponds to a cardiac cycle of the patient;

extracting a set of fiducials for each segment of the segmented PG dataset, wherein the set of fiducials is determined based on values of at least one of:

the segmented PG dataset,

a first derivative of the segmented PG dataset,

a second derivative of the segmented PG dataset, or

a third derivative of the segmented PG dataset; and

determining the cardiovascular state of the patient by applying a transformation to the set of fiducials;

wherein extracting a set of fiducials for each segment of the segmented PG dataset comprises:

calculating the first, the second, and the third derivative of the segment;

determining time points corresponding to zeros of the first, second, and third derivative;

determining values of the segment at the time points;

determining values of the first, second, and third derivatives of the segment at the time points; and

wherein the set of fiducials comprises:

the values of the segmented PG dataset for time points corresponding to the first four zeros of the first and second derivatives;

the values of the first derivative of the processed PG dataset for the time points corresponding to the first four zeros of the first and second derivatives; and

the values of the second derivative of the processed PG dataset for the time points corresponding to the first four zeros of the first and second derivatives.

8. The method of claim 7 , further comprising: interpolating between data points in the PG dataset to fill gaps in the PG dataset.

9. The method of claim 7 , wherein the cardiovascular state of the patient comprises one or more of: a blood pressure, a cardiac event, cardiovascular drift, and a drug reaction of the patient.

10. The method of claim 7 , wherein reducing the noise in the PG dataset further comprises filtering the PG dataset using a long pass filter.

11. The method of claim 7 , wherein the set of intrinsic modes are determined using enhanced empirical mode decomposition.

12. The method of claim 11 , wherein removing noise from the PG dataset comprises aggregating the first six modes of the set of intrinsic modes.

13. The method of claim 7 , wherein segmenting the denoised PG dataset comprises segmenting the denoised PG dataset by detecting a moving-average crossover.

14. The method of claim 7 , further comprising determining an average PG dataset by averaging a predetermined number of segments of the segmented data, wherein the set of fiducials is determined based on the averaged PG dataset.

15. The method of claim 7 , further comprising:

determining a patient cardiovascular manifold, relating a heart function, nervous system, and vascular system of the patient, based on the set of fiducials; and

determining a manifold transformation that converts the patient cardiovascular manifold to a universal cardiovascular manifold;

wherein the transformation comprises the manifold transformation.

16. The method of claim 7 , wherein the transformation is determined based on a calibration dataset generated based on cardiovascular state measurements derived from a group of calibration patients.

17. The method of claim 7 , wherein the transformation is a linear transformation.