Artificial intelligence and hypertension: new

approach in understanding of possible

mechanisms

Milovanovic Branislav,Drasko Furundzic,GligorijevićTatjana

University Clinical Center Bezanijska Kosa

Neurocardiological Laboratory,Medical Faculty , Belgrade

Mihajlo Pupin Institute, Volgina 15, 11000 Belgrade, Serbia

Artificial intelligence will control of

cockpits

Not for nervous flyers! Boeing to test pilotless

planes next year as artificial intelligence takes

control of cockpits

Autonomous Cars and Artificial Intelligence

A prototype Audi A7 with self-driving technology is

seen during testing on the A9 autobahn in

Germany in May 2016.

Applications of Neural Networks

They can perform tasks that are easy for a human but difficult for a machine −

Aerospace − Autopilot aircrafts, aircraft fault detection.

Automotive − Automobile guidance systems.

Military − Weapon orientation and steering, target tracking, object discrimination, facial recognition, signal/image

identification.

Electronics − Code sequence prediction, IC chip layout, chip failure analysis, machine vision, voice synthesis.

Financial − Real estate appraisal, loan advisor, mortgage screening, corporate bond rating, portfolio trading

program, corporate financial analysis, currency value prediction, document readers, credit application evaluators.

Industrial − Manufacturing process control, product design and analysis, quality inspection systems, welding

quality analysis, paper quality prediction, chemical product design analysis, dynamic modeling of chemical process

systems, machine maintenance analysis, project bidding, planning, and management.

Medical −Cancer cell analysis, EEG and ECG analysis,

prosthetic design, transplant time optimizer. Speech − Speech recognition, speech classification, text to speech conversion.

Telecommunications − Image and data compression, automated information services, real-time spoken language

translation.

Transportation − Truck Brake system diagnosis, vehicle scheduling, routing systems.

Software − Pattern Recognition in facial recognition, optical character recognition, etc.

Time Series Prediction − ANNs are used to make predictions on stocks and natural calamities.

Signal Processing − Neural networks can be trained to process an audio signal and filter it appropriately in the

hearing aids.

Control − ANNs are often used to make steering decisions of physical vehicles.

Anomaly Detection − As ANNs are expert at recognizing patterns, they can also be trained to generate an output

when something unusual occurs that misfits the pattern.

Artificial Neural Networks

Artificial Neural Networks (ANNs)?

The inventor of the first neurocomputer, Dr. Robert

Hecht-Nielsen, defines a neural network as −

"...a computing system made up of a number of

simple, highly interconnected processing elements,

which process information by their dynamic state

response to external inputs.”

Introduction

Artificial neural networks (ANN) are data driven

learning structures based on the principles of

morphological and functional organization of

biological neurons. Basic quality of trained neural

structures, generalization, association and self-

organization, enable them reliable nonlinear multi-

variate regression, classification and clustering. The

models trained on the representative sample

generalized knowledge on the unknown test sample

with high reliability even at low level of

representativeness of the training set.

Heart rate variability

Artificial Neural Networks

Heart rate variability and ANN

The heart rate variability is used as the base

variable from which certain parameters are

extracted and presented to the ANN for

classification

NEUROCARDIOLOGICAL

LABORATORY

Center for noninvasive electrocardiology

Center for autonomic nervous system

testing in clinical medicine

Syncope center

METHODOLOGY

Data were obtained using short ECG analysis (Shiller AT-10), non-invasive beat-to-beat heart rate variability and baroreflex sensitivity (Task Force monitor) and 24 hour ambulatory ECG monitoring with long term HRV analysis

ECG parameters were obtained from the signals of all 12 ECG channels over the past 5 minutes using commercial software (Schiller AT-10, Austria)

The Task Force Monitor (CNSystems, Graz, Austria), was used to monitor beat-to-beat heart rate (HR) by ECG and beat-to-beat blood pressure by the vascular unloading technique [12], which was corrected automatically to the oscillometric blood pressure measured on the contralateral arm. The Task Force Monitor automatically provides beat to beat spectral analysis of heart rate, systolic and diastolic blood pressure variability, applying an autoregressive methodology

Baroreceptor reflex sensitivity (BRS) was automatically assessed using the sequence technique according to Parati

Twenty-four-hour ambulatory ECG recordings were acquired by a 12 leads electrocardiogram, sampling rate 1000 Hz per each lead (Cardioscan, D.M.S.USA) and analyzed by an experienced analyst

TASK FORCE monitor

TASK FORCE monitor

Finger blood pressure monitoring

PORTAPRES

The Portapres® is the ambulatory Finapres

technology solution. The Portapres® offers

on top of standard ambulatory blood

pressure monitoring (ABPM) insight into

hemodynamic parameters such as stroke

volume and cardiac output. For almost 20

years the technology has proven itself in

clinical settings, high altitude research on

mountain heights and in space by top

scientific institutes like NASA

Overview parameters

The following parameters are available using the Portapres® in combination with BeatScope® software:

Parameter Abreviation

1 Blood Pressure SYS SYS

2 Blood Pressure DIA DIA

3 Blood Pressure MEAN MEAN

4 Heart rate HR

5 Inter beat interval IBI

6 Cardiac output CO

7 Stroke volume SV

8 Pulse rate variability* PRV

9 Baroreflex sensitivity* BRS

10 Total peripheral resistance TPR

11 Total arterial compliance CwK

12 Max. steepness of current upstroke dp/dt

13 Ascending aortic impedance at DIA Zao

14 Left ventricular ejection time LVET

15 Rate pressure product PS*HR

1. Autonomic Nervous System Activity

Method

ANSA SCAN METHOD

2. Autonomic Nervous System Activity

Scanning

ANSA SCAN SOFTWARE

ANSA SCAN SOFTWARE

o 33 parameters of short and long time

HRV analysis

o 21981 in statistical analyse

o Scanning of type of autonomic balance

o Individualized approach related to drugs

> Ansa Scan Plus

ANSA SCAN PLUS

Short time parameters

EKGSpectral

parameters BRS and BPQTc

QT

PR

QRS

P

Paxis

QRSaxis

Taxis

Mean RR

SDRR

PNN50

RMSSD

LF nu

LF ms

HF nu

HF ms

VLF ms

VLF nu

LF ms

HF ms

TP ms

LF/HF ms

LF/HF nu

BRS

BI

HR

sBP

dBP

mBP

ANSA SCAN PLUS

Long time parameters

Time domain Spectral Blood pressure

Mean RR

Avg HR

SDNN/24

SDANN INDEX

SDNN INDEX

rMSSD

PNN50

TP ms

VLF ms

VLF nu

LF ms

LF nu

HF ms

HF nu

LF/HF ms

ULF ms

ULF nu

sBP

dBP

PULS PRESSURE

ANSA SCAN PLUS

Groups

Groups II Groups III Groups IV

1.Parasympathetic predominance

2.Sympathetic predominance

1.

Parasympathetic

predominance

2.

Balance state

3.

Sympathetic

1.

High parasympathetic

predominance

2.

Mild parasympathetic

predominance

3.

Mild sympathetic

predominance

4.

High sympathetic

predominance

ANSA SCAN PLUS

Combination of parameters with TP,VLF,BRS

4 groups

1.

Low parameter with low total power

2.

Low parameter with high total power

3.

High parameter with low total power

4.

High parameter with high total power

ANSA SCAN PLUS

Combination of parameters with TP,VLF,BRS

9 groups

1.

Low parameter with low total power

2.

Low parameter with normal total power

3.

Low parameter with high total power

4.

Normal parameter with low total power

5.

Normal parameter with normal total power

6.

Normal parameter with high total power

7.

High parameter with low total power

8.

High parameter with normal total power

9.

High parameter with high total power

ANSA SCAN PLUS

Combination of parameters with TP,VLF,BRS

8 groups

Very low parameter with low total power

2.

Very low parameter with high total power

3.

Mild low parameter with low total power

4.

Mild low parameter with high total power

5.

Mild high parameter with low total power

6.

Mild high parameter with high total power

7.

Very high parameter with low total power

8.

Very high parameter with high total power

Heart rate variability intervals,HRVI

NEUROCARD 2017

Hypertension-Healthy

Clasterisation of groups

0 50 100 150 200 250 300 3500

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

FP

FNTN

TP

Healthy-Hypertension

Healthy-Hypertension

Groups

-1 -0.5 0 0.5 1 1.5 2 2.5-1

-0.5

0

0.5

382 116

Hits

83 274274 83116 382

ECG (5)(identical 6 parameters !! )

Impact of ECG parameters on clusters

Parameters Hipertension Controls

1.0000 'QTc' 1.0084 0.96122.0000 'QT‘ 0.8163 0.79103.0000 'PR‘ -0.0829 -0.12454.0000 'QRS' -0.3432 -0.36965.0000 'P' -0.2843 -0.31216.0000 'mRR' 2.5784 -0.63827.0000 'Paxis' -0.5547 -0.68878.0000 'QRSaxis' -0.6086 -0.58589.0000 'Taxis' -0.5693 -0.551010.0000 'SDRR' -0.6489 -0.533311.0000 'PNN50' -0.6939 -0.574212.0000 'RMSSD' -0.6173 2.6253

Healthy-Hypertension

Impact of ECG parameters

(12)

1 2 3 4 5 6 7 8 9 10 11 12-1

-0.5

0

0.5

1

1.5

2

2.5

3

Hiperten. 1

Kontrol. Gr.2

Impact of ECG parameter

Mean RR

400 500 600 700 800 900 1000 1100 1200 1300 14000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

mRR

Impact of ECG parameter

PNN50%

Hypertension Syncope

0 5 10 15 20 25 30 35 40 450

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

PNN50 0 5 10 15 20 25 30 35 40 450

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

PNN50

Syncope C

lass (

1)

and C

ontr

ol C

lass (

0)

Hypertension-Healthy

ECG and Holter ECG parameters

(53)

Hypertension-Healthy

ECG and Holter ECG parameters

(53)

-1 -0.5 0 0.5 1 1.5 2 2.5-1

-0.5

0

0.5

382 116

Hits

83 274274 83116 382274 83382 11682 275274 83

Hypertension-Healthy

Impact of ECG and Holter ECG parameters

(53)

0 10 20 30 40 50 60-1

0

1

2

3

4

5

6

Serial number of examples

Dejstvo parametara na Klase Hipertenz I Kontrolnu Klasu

Hipert. 1

Kontrol 2

Holter ECG parameters

Impact on hypertension and controls

Parameters Hypertension Controls

19.0000 TP 5.1540 3.6695

20.0000 VLF 3.3694 2.3626

23.0000 Mean RR 1.2626 0.6988

Impact of Holter ECG parameter

Total Power (TP)

0 2000 4000 6000 8000 10000 12000 140000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

TP

Impact of Holter ECG parameter

Very Low Frequency (VLF)

0 2000 4000 6000 8000 10000 120000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

VLF

Hypertension Syncope

Impact of Holter ECG parameter

Mean RR interval

550 600 650 700 750 800 850 900 950 1000 10500

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

mRR

Artificial network,ANN

Hypertension

ANN

Low Frequency,(LF) and hypertension

Sympathetic activity

0 1000 2000 3000 4000 5000 6000 70000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

LFms

ANN

Duration of P wave and hypertension

50 100 150 2000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

P

Hypertension

ANN

Duration of QRS axis and

hypertension

-200 -150 -100 -50 0 50 100 150 2000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

QRSaxis

ANN

Systolic blood pressure and

hypertension

80 100 120 140 160 180 200 2200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

sBP

Lyme disease

Blood pressure changes

Very low baroreflex sensitivity

Blood pressure variability

Syncope and Epilepsia

Dysfunction of baroreflex activity

Syncope

Patient M.V. 6 years old

Syncope

Epilepsia

Hypertension!!

Low value of vitamin D

Acute infection with

COXSACKIE VIRUS

Lyme disease

Syncope

Baroreflex sensitivity Blood pressure

Panic atack with Syncope before the

head up tilt testing

Patient K.V.20 years old

Hypertension

Acute infection with Borelia burgdorferi,Adeno

virus,Influenza A,Influenza B,Echo virus

Syncope (children) and Control Group

(III)

0 10 20 30 40 50 60 70 80 90 100-0.2

0

0.2

0.4

0.6

0.8

1Klasifikacija test uzorka Sinkope Deca

redni broj primera

0 -

Kontr

oln

i pacije

nti

1-

Pacije

nti d

eca s

a S

inkopam

a

Artificialis neural network-ANN

LYME DISEASE AND SYNCOPE

Lyme disease and syncope

ANN

Lyme disease and control group

Partial impact of variables

0 5 10 15 20 25 30 35 40 450

1

2

3

4

5

6

7

Valu

es in p

erc

enta

ges(%

)

Serial number of variables

Partial impact of variables in the outcome

Conclusion

ANN models can be used in modeling different

types of pathology and diagnostics. In this

particular case the ANN structure enabled us a

highly reliable discrimination of patients with

hyperetnsion,syncope and patients without risk,

based on standard cardiologic examination

procedure