Electrocardiograph

Prediction of atrial fibrillation, AV block 1 and normal ECG with more than 90% reliability

Yrish Technology

Prediction

Sudden heart attack and increased deaths: growing concern.

In the last decade, deaths from sudden heart attacks have skyrocketed,
particularly in developing countries, apart from genetic and lifestyles, the lack of medical resources in rural areas causes most of the deaths in heart attacks.

The reasons for death from heart attack in different countries are: rural areas without sanitary facilities, lack of awareness about the symptoms of cardiac arrest leads to death and increased cases of heart disease.

ECG Analyzer Machines on the Market and Their Features:

  • Today's IoT medical device sends massive ECG data to mobile or server and analysis is performed on mobile application or high-performance servers.
  • Today's IoT medical device sends massive ECG data to mobile or server and analysis is performed on mobile application or high-performance servers.
  • Today's IoT medical device sends massive ECG data to mobile or server and analysis is performed on mobile application or high-performance servers.

We can summarize it in the following points:

  • Reliance on the Internet for massive transmission of ECG data.
  • Requires system or mobile application.
  • A high computational system is needed. 

Our solution

  • The technology used can analyze the ECG data without relying on the Internet.
  • Latency is lower compared to IoT devices.
  • Our device will analyze the ECG patterns and classify them into Normal, Atrial Fibrillation, and First Degree Heart Block.

ECG chart basics

The ECG graph is divided into 5 waves: P, Q, R, S and T waves.

YRISH Technology

Atrial fibrillation

YRISH Technology

Irregular heart rate: The difference between the current R-R interval and the previous R-R interval is 200 ms

First degree heart block

YRISH Technology

If the P-R interval exceeds 200 ms, it may be indicated as first-degree heart block

ECG Electrode Placement

The ECG electrodes are placed in RA, LA and LL as mentioned in the following diagram and connect to our ECG analyzer.

YRISH Technology

For best results: place the electrodes on the chest wall equidistant from the heart (instead of specific extremities)

Novel approach to producing quality data sets

In machine learning, the accuracy and performance of a model are determined by the quality and divergence of a data set.

If you look at the ECG data, it is really difficult to distinguish the different heart conditions with the normal ECG data in a shorter window time (example: 3 seconds)

YRISH Technology

Normal ECG data

When training a model with filtered ECG data only for atrial fibrillation, normal heart block, and first degree, the accuracy is less than 23%. The reason is that in the shorter window the model cannot differentiate the difference. If a longer window time is sought, the processing time and maximum RAM usage increase quite a bit, but without precision.

Background of the novel approach:

When a doctor or a trained person analyzes the ECG graph, they will be counting the small squares between the R wave to R, the interval P to R and they will write the counts on the graph or save it for calculation.

YRISH Technology

ECG graph reading

The solution is to convert human observations into data sets. This is how the accuracy of the ECG analyzer model is increased.

Separate waveforms are created from filtered ECG data.

New waveforms:

RR interval
PR interval

Condition

Normal


Atrial fibrillation


First Degree Heart Block

Human Observation

Observe squares on the graph and find no deviations.

Remark: the box count in the graph varies between two intervals R R

Cell count on graph between P and R indicates> 200 ms

ECG analyzer

R-R interval value: 100
P-R interval value: 50

R-R interval value: -100
P-R interval value: 50

R-R interval value: 100
P-R interval value: -50

YRISH Technologyh

ECG Analyzer Algorithm

Data sets generated for normal ECG data:

YRISH Technology

Normal ECG Data: Decoded RR interval and PR interval data are always 100 and 50 for normal ECG data.

ECG Data Sets Generated for Atrial Fibrillation

YRISH Technology

Atrial Fibrillation: Deviation between the previous RR interval and the current RR interval, the RR interval data will be reduced to -100 for one cycle.

ECG data sets for first degree heart block

YRISH Technology

First Degree Heart Block: Whenever the P to R interval exceeds 200 ms, the PR interval data will be reduced to -50 for one cycle.

As a result we have developed a pocket-sized mini-diagnostic ECG analyzer device that can independently diagnose heart disease without the Internet.

© Copyright 2021 YRISH Technology, S.C.A. - All Rights Reserved