Blood pressure estimation method using cascade forest regression model

1. A blood pressure estimation method using a cascade forest regression model comprising:

extracting a plurality of blood flow characteristics from PPG signals for training,

calculating systolic and diastolic blood pressures from ambulatory blood pressures for training,

labeling the systolic and diastolic blood pressures with the plurality of blood flow characteristics to train a cascade forest regression model, and

inputting the PPG signals of a target user into the trained cascade forest regression model to determine the systolic and diastolic blood pressures of the target user.

2. The blood pressure estimation method using a cascade forest regression model of claim 1 ,

wherein the step of extracting a plurality of blood flow characteristics comprises

passing the PPG signals for training through a high pass filter (HPF) to remove noise from the PPG signals for training.

3. The blood pressure estimation method using a cascade forest regression model of claim 1 ,

wherein the step of extracting a plurality of blood flow characteristics comprises

sampling the PPG signals for training at unit time intervals,

detecting R-R interval and heart rate variability (HRV) from the sampled signals, and

extracting the plurality of blood flow characteristics based on the R-R interval and heart rate variability.

4. The blood pressure estimation method using a cascade forest regression model of claim 3 ,

wherein the plurality of blood flow characteristics comprise BPM, SDNN, RMSSD, IBI and SDSD defined by the following equations:

BPM

=

60000

⁢

ms

⁡

(

1

⁢

minute

)

RR

_

SDNN

=

1

n

-

1

⁢

∑

i

=

1

n

⁢

(

RR

i

-

RR

_

)

2

RMSSD

=

1

n

-

1

⁢

∑

i

=

1

n

⁢

(

RRdiff

i

)

2

IBI

=

1

n

-

1

⁢

∑

i

=

1

n

⁢

RR

i

SDSD

=

1

n

-

1

⁢

∑

i

=

1

n

⁢

(

RRdiff

i

-

RRd

⁢

ι

⁢

ff

_

)

2

in which RR i is the sequentially indexed R-R interval in the sampled signal, n is the number of RR i in the sampled signal, and RRdiff i is the heart rate variability sequentially indexed in the sampled signal.

5. The blood pressure estimation method using a cascade forest regression model of claim 4 ,

wherein the plurality of blood flow characteristics further comprise at least one of SD1, SD2, S, SD, pNN20, pNN50 and HR mad defined by the following equations:

SD

⁢

1

=

1

2

⁢

SDSD

2

SD

⁢

2

=

2

⁢

SDNN

2

-

1

2

⁢

SDSD

2

S

=

π

×

SD

⁢

1

×

SD

⁢

2

SD

=

SD

⁢

1

/

SD

⁢

2

pNNxx

=

RR

x

n

HR

⁢

mad

=

median

(

❘

"\[LeftBracketingBar]"

RR

i

-

median

(

RR

)

❘

"\[RightBracketingBar]"

)

in which RR x is the number of RR i with an interval of x or more.

6. The blood pressure estimation method using a cascade forest regression model of claim 1 ,

wherein the step of calculating the systolic and diastolic blood pressures comprises:

sampling the ambulatory blood pressures for training at unit time intervals,

detecting a plurality of high peaks and a plurality of low peaks in the sampled signals, and

averaging out the plurality of high peaks and the plurality of low peaks, respectively, to calculate the systolic and diastolic blood pressures.

7. The blood pressure estimation method using a cascade forest regression model of claim 6 ,

wherein the step of detecting the plurality of high peaks and the plurality of low peaks comprises

detecting only peaks above a reference value.

8. The blood pressure estimation method using a cascade forest regression model of claim 1 ,

wherein the step of training the cascade forest regression model comprises:

inputting the plurality of blood flow characteristics into the cascade forest regression model, and

supervising training the cascade forest regression model such that the difference between the predicted values output from the cascade forest regression model and the labeled systolic and diastolic blood pressures is minimized.

9. The blood pressure estimation method using a cascade forest regression model of claim 1 ,

wherein the cascade forest regression model comprises a plurality of layers, and each of the plurality of layers comprises a plurality of random forests and a plurality of extra trees.

10. The blood pressure estimation method using a cascade forest regression model of claim 1 ,

wherein the cascade forest regression model comprises a plurality of sequentially connected layers, and the predicted values output from the previous layer is combined with the blood flow characteristics and input into the next layer.

11. The blood pressure estimation method using a cascade forest regression model of claim 1 ,

wherein the cascade forest regression model comprises a plurality of sequentially connected layers, and

the step of training the cascade forest regression model comprises calculating the accuracy of the cascade forest regression model through a validation dataset comprising a portion of the PPG signals for training and a portion of the ambulatory blood pressures for training, and adding a layer to the cascade forest regression model if the accuracy is less than a reference value.

12. The blood pressure estimation method using a cascade forest regression model of claim 1 ,

wherein the step of determining the systolic and diastolic blood pressures of the target user comprises:

collecting the PPG signals of the target user,

sampling the PPG signals at unit time intervals,

inputting the sampled signals into the cascade forest regression model, and

determining the predicted values output from the cascade forest regression model as the systolic and diastolic blood pressures of the target user.