DK202201081A1

DK202201081A1 - A method for identifying morphological abnormalities in heart rhythm data

Info

Publication number: DK202201081A1
Application number: DKPA202201081A
Authority: DK
Inventors: Jermiin Ravn Moll Jonas; Klitzing Hartmann Hansen Mikkel
Original assignee: Vital Beats Aps
Priority date: 2021-11-30
Filing date: 2022-11-28
Publication date: 2023-06-21

Abstract

A method for analysing, and detecting anomalies in, heart rhythm data is disclosed. A time series of heart rhythm data representing heartbeats from a person is analysed. A plurality of sub-sequences in the time series of heart rhythm data is identified, each sub-sequence corresponding to a single heartbeat, and the sub-sequences are isolated. The sub-sequences are fed to a mathematical model, and the sub-sequences are categorized and grouped, using the mathematical model, and each group of sub-sequences is labelled. The sub-sequences are superimposed on each other, based on one or more recognisable features in each heartbeat, and the superimposed and labelled sub-sequences collectively form the clinical decision support for the medical professional.

Description

A METHOD FOR IDENTIFYING MORPHOLOGICAL ABNORMALITIES IN HEART RHYTHM DATA

FIELD OF THE INVENTION

The present invention relates to a method for analysing heart rhythm data, which provides clinical decision support for a medical professional who is interpreting the heart rhythm data.

BACKGROUND OF THE INVENTION

The heart rhythm of persons sometimes needs to be measured or monitored, e.g. for diagnostic purposes or as a part of a general health check. To this end, a time series of heart rhythm data is measured, e.g., at a medical facility operated by the medical professional or at home using equipment operated by the person himself or herself. The measured heart rhythm data is then analysed and interpreted by a medical professional.

Since the measured heart rhythm data is presented as a time series of heartbeats, and anomalies and outliers are infrequent, it may be both time-consuming and difficult to detect the subtle differences, even for a trained medical professional, when performing the analysis and interpretation of the heart rhythm data.

DESCRIPTION OF THE INVENTION

It is an object of embodiments of the invention to provide a method for analysing heart rhythm data which assists a medical professional in quickly interpreting the heart rhythm data.

It is a further object of embodiments of the invention to provide a method for analysing heart rhythm data which reduces the time required by a medical professional for interpreting the heart rhythm data.

It is an even further object of embodiments of the invention to provide a method for analysing heart rhythm data which decreases the risk of missing early sighs of worsening or symptoms of deterioration in the heart rhythm data.

The invention provides a method for analysing heart rhythm data, the method comprising the steps of: — Interpreting a time series of heart rhythm data representing heartbeats from a person,

— identifying a plurality of sub-sequences in the time series of heart rhythm data, each sub-sequence corresponding to variances in a single heartbeat, and isolating the sub- sequences, — feeding the sub-sequences to a mathematical model, and categorizing and grouping the sub-sequences using the mathematical model, and labelling each group of sub- sequences, — superimposing the sub-sequences on each other, based on one or more recognisable features of a heartbeat, and — displaying the superimposed and labelled sub-sequences.

Thus, the method according to the invention is a method for analysing heart rhythm data, i.e. measured data relating to the heart rhythm of a person.

In the method according to the invention, a time series of heart rhythm data representing heartbeats from a person is initially measured. This could, e.g., be performed by medical professionals at a suitable medical facility, or at home using equipment or a device being operated by the person himself or herself. This will be described in further detail below. In any event, the measured heart rhythm data is in the form of a time series, i.e. the heartbeats are represented sequentially as a function of time.

Next, a plurality of sub-sequences is identified in the time series of heart rhythm data, in such a manner that each sub-sequence corresponds to variances in a single heartbeat.

Furthermore, the sub-sequences are isolated. Thus, the individual heartbeats of the person are identified and isolated from the time series, e.g. based on recognisable characteristic features of a heartbeat, which allows a specific variance pattern to be recognised as a heartbeat. The identification and isolation of the sub-sequences, and thereby of the individual heartbeats, allows the heartbeats to be easily analysed and compared. This will be described in further detail below.

The sub-sequences are then fed to a mathematical model, e.g. in the form of an Al model, such as a trained Al model. The mathematical model is preferably trained in identifying patterns in variances in heart rhythm data. The sub-sequences are categorized and grouped, using the mathematical model. Accordingly, the sub-sequences, and thereby the individual heartbeats, are analysed by means of the mathematical model, thereby identifying patterns in the variances in the heart rhythm data corresponding to the heartbeats. Based on these patterns, the sub-sequences are categorized and grouped, so that sub-sequences, and

2 DK 2022 01081 A1 thereby heartbeats, which are similar to each other are categorized and grouped together.

Thereby a number of groups of sub-sequences are formed. Furthermore, each group of sub- sequences are labelled, thereby allowing sub-sequences belonging to a specific group to be identified. The labelling could, e.g., be assigning a colour or a legend to each group.

Alternatively or additionally, the labelling could include flagging the data in a manner which allows only data relating to a given group of sub-sequences to be displayed or presented.

Next, the sub-sequences are superimposed on each other, based on one or more recognisable features of a heartbeat. A heartbeat has a number of well-defined features which are reflected in measured heart rhythm data, and which represents the various contractions of the heart which normally take place during a heartbeat. Such features include various waves, normally labelled P, Q, R, S, T and J waves. By superimposing the sub- sequences on each other based on such well-defined and recognisable features, it is ensured that the sub-sequences, when being superimposed, are aligned with respect to at least one of these features. Thereby the heartbeats corresponding to the sub-sequences are easily comparable to each other from the superimposition.

Finally, the superimposed and labelled sub-sequences are displayed. Accordingly, the measured heart rhythm data is displayed in a format in which the individual heartbeats are shown in a superimposed manner, i.e., ‘on top of each other’ in the sense that at least one of the phases during the heartbeats are aligned, thereby allowing easy and direct comparison among the heartbeats, instead of presenting the heart rhythm data as one long time series, where the individual heartbeats are separated from each other along a time axis.

Furthermore, the labelling of the sub-sequences is also displayed, thereby allowing sub- sequences belonging to a given group to be readily and easily recognised, thereby drawing the attention to specific features which are characteristic for a group of sub-sequences.

Thus, the displayed superimposed sub-sequences allow a medical professional who is interpreting the heart rhythm data to easily, reliably, and fast identify heart rhythm patterns which may be of concern, abnormalities, trends in the heart rhythm data, etc. Accordingly, the time required for the medical professional to interpret the heart rhythm data is reduced as compared to a situation where the medical professional interprets the raw time series data. Furthermore, the risk of missing details in the heart rhythm data which may be relevant is reduced. Thereby the displayed superimposed sub-sequences provide a valuable tool for the medical professional, which reduces the time used by the medical professional in performing the interpretation of the heart rhythm data, while improving the quality and reliability of the result, including minimising the risk of errors in terms of missing signals of concern in the data.

4 DK 2022 01081 A1

The step of measuring a time series of heart rhythm data may be performed by means of an implantable device and/or by means of a wearable device.

In the present context the term 'implantable device' should be interpreted to mean a device which has been introduced into the body of the person by means of surgery, either inside the heart or on an outer surface of the heart. The implantable device could, e.g., be in the form of a pacemaker, an implantable cardioverter defibrillator (ICD), or any other suitable kind of implantable device which is capable of measuring heart rhythm data. Since such an implantable device is always active, an extensive amount of heart rhythm data can be measured, including heart rhythm data corresponding to various activities and/or situations which the person is performing or experiencing.

In the present context the term 'wearable device' should be interpreted to mean a device which is worn by the person, i.e., which has not been implanted by means of surgery, whether invasive or non-invasive. The wearable device could, e.g., be in the form of a watch, a piece of garment, a pad, a plaster, etc., as long as it is in contact with the body of the person in a manner which allows it to measure heart rhythm data. Such wearable devices allow the person to measure the heart rhythm data without involving medical professionals, and e.g., while performing exercise or other tasks, while sleeping, etc., thereby obtaining an extensive amount of heart rhythm data, possibly reflecting various situations or activities which the person is experiencing or performing.

It should be noted that at least part of the analysis process described above may take place at the implantable or wearable device, e.g., the step of identifying a plurality of sub- sequences and possibly the step of categorizing and grouping the sub-sequences.

According to this embodiment, since the implantable or wearable device can be used for measuring heart rhythm data over long time periods and during various situations or activities, an extensive amount of heart rhythm data may be made available for the subsequent analysis, and the measured data forms a good representation of the normal life of the person, including diurnal patterns, activities normally performed by the person, etc.

Alternatively, or additionally, the heart rhythm data may be measured by means of standard

ECG equipment, including pads attached to the body of the person, at a suitable medical facility and by an appropriate medical professional.

The method may further comprise the step of identifying outliers and/or known diagnostic variances in the sub-sequences, and the step of displaying the superimposed sub-sequence may comprise highlighting identified outliers and/or known diagnostic relevant variances.

; DK 2022 01081 A1

According to this embodiment, the medical professional performing the interpretation of the heart rhythm data is provided further assistance in the sense that possible abnormalities and/or patterns in the heart rhythm data which are known to be an indication of a possible diagnosis are automatically identified and highlighted. Thereby the attention of the medical professional is pointed towards the parts of the heart rhythm data which may require specifically thorough investigation. Accordingly, the risk of missing small indications in the heart rhythm data which may require medical attention is minimised.

Outliers in the sub-sequences may, e.g., represent heartbeats which for some reason differ significantly from the average heartbeats of the person. This could, e.g., be an indication of atrial fibrillation, for example defined by distance between R peaks being irregular above a specified threshold, or ventricular tachycardia (VT), which could be determined by a heart rate (HR) above 150 beats per minute during 3 consecutive beats with a QRS under 120 milliseconds.

Relevant diagnostic variances in the sub-sequences may, e.g., include mean width of the

QRS complex, or mean distance between PQ and PR, or mean distance between the Q and T wave.

The step of superimposing the sub-sequences on each other may comprise recognising one or more waves of each heartbeat, and aligning features of the recognised waves of the sub- sequences.

As described above, a normal heartbeat comprises a sequence of contractions of the heart, which are reflected as a characteristic variance pattern in the heart rhythm data. Features of this variance pattern are normally referred to as waves. More particularly, the waves of a heartbeat are normally referred to as the P, Q, R, S, T and J waves, and for a given heartbeat, these waves can normally be recognised and identified.

The most significant part of a heartbeat is normally the R wave, which is preceded by the Q wave and succeeded by the S wave. This sequence is often referred to as the QRS complex.

Thus, the step of superimposing the sub-sequences on each other may advantageously comprise identifying the R wave or the QRS complex of each heartbeat, and aligning the sub- sequences based on their R waves or QRS complexes. For instance, the sub-sequences may be aligned in such a manner that the peak of their R waves coincide, the beginning of their R waves coincide, and/or the end of their R waves coincide.

Alternatively, or additionally, the sub-sequences may be aligned according to any of the other waves of the heartbeats, and/or according to any other suitable recognisable feature of the heartbeats.

The step of categorizing and grouping the sub-sequences may be performed based on known and diagnostically relevant features of the heartbeat variances.

According to this embodiment, the sub-sequences are not merely categorized and grouped based on similarity and patterns recognised by the mathematical model. The categorization and grouping are also based on features of the heartbeat variances which are known to be diagnostically relevant. Thereby the attention of the medical professional performing the interpretation of the heart rhythm data is directed specifically towards such diagnostically relevant features.

Examples of such diagnostically relevant features include but are not limited to the following.

Inverted T wave, i.e., a T wave which forms a valley rather than a peak in the heart rhythm data. An inverted T wave may, e.g., be diagnostically indicative of myocardial ischaemia.

Length of PR interval, i.e., the distance between the beginning, middle or end of the P wave and the beginning, middle or end of the Q wave (depending on the preference of the medical professional or clinic). Non-existent P waves may be diagnostically indicative of fibrillation and/or flutter.

Length of QT interval, i.e., the distance between the beginning, middle or end of the Q wave and the beginning, middle or end of the T wave (depending on the preference of the medical professional or clinic). A lengthening or extended length of the QT interval can be used to estimate Long QT syndrome.

ST segment abnormality (ST elevation or depression), i.e., elevated or depressed ST segment, i.e., the segment between the end of the S wave and the beginning of the T wave.

This is diagnostically indicative of myocardial infarction or myocardial ischaemia.

Width of QRS complex, i.e., the distance from the beginning, middle or end of the Q wave to the beginning, middle or end of the S wave (depending on the preference of the medical professional or clinic). A widening of the QRS over time may be diagnostically indicative of the medical condition called heart failure.

, DK 2022 01081 A1

The step of identifying a plurality of sub-sequences in the time series of heart rhythm data may comprise identifying a plurality of waves in the time series of heart rhythm data, and, for each identified wave, identifying a sub-sequence as a time interval from a predefined time before a start, and/or end, and/or middle of the wave to a predefined time after the start, and/or end, and/or middle of the wave.

The identified waves could be any of the P, Q, R, S, T and J waves described above. However, since the R wave is the one which is normally most significant, the step of identifying a plurality of sub-sequences may advantageously be based on an identification of a plurality of

R waves.

Thus, according to this embodiment, for each heartbeat, one or more recognisable and characteristic features, in the form of one or more of the well-known waves, are identified, and a sub-sequence representing a heartbeat is defined as a suitable time interval before and after this feature.

The method may further comprise the step of a user manually flagging a part of the time series of heart rhythm data, during the step of measuring the time series of heart rhythm data, and the step of displaying the superimposed sub-sequences may comprise highlighting sub-sequences corresponding to the flagged part of the time series of heart rhythm data.

According to this embodiment, the person whose heart rhythm data is being measured may manually flag a part of the time series of heart rhythm data while the heart rhythm data is being measured. This could, e.g., be relevant in the case that the person experiences abnormalities or discomfort.

By highlighting the sub-sequences corresponding to the flagged part of the time series of heart rhythm data, the attention of the medical professional performing the interpretation of the heart rhythm data is drawn towards parts of the heart rhythm data which corresponds to the time periods where the person felt that something abnormal was happening. This allows the medical professional to thoroughly investigate these parts of the data, in order to establish whether or not the measured data provides an explanation for the experience of the person.

This could, e.g., be relevant in the case of suspected extra beats such as PACs (pre-atrial contractions) and PVCs (pre-ventricular contractions) and/or SVEBs (Supra Ventricular Extra

Beats).

The method may further comprise the step of performing automated pre-diagnostics based on the displayed superimposed sub-sequences and/or on the categorization of the sub- sequences.

According to this embodiment, one or more automatically generated qualified guesses on possible diagnoses are offered to the medical professional performing the interpretation of the heart rhythm data, based on analysis of the data performed by means of the mathematical model.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention will now be described with reference to the accompanying drawings in which

Fig. 1 illustrates variances in heart rhythm data corresponding to a normal heartbeat,

Fig. 2 is a diagrammatic view of a system for performing a method according to an embodiment of the invention,

Fig. 3 illustrates a measured time series of heart rhythm data, and

Fig. 4 illustrates superimposed sub-sequences of heart rhythm data obtained by means of a method according to an embodiment of the invention.

DETAILED DESCRIPTION OF THE DRAWINGS

Fig. 1 illustrates variances in heart rhythm data corresponding to a normal heartbeat. More particularly, Fig. 1 shows the P, Q, R, S and T waves of a heartbeat. Furthermore, relevant distances between the features are illustrated, and the QRS complex is identified.

Fig. 2 is a diagrammatic view of a system 1 for performing a method according to an embodiment of the invention. An ECG recording apparatus 2, e.g., in the form of an implantable or a wearable device, measures heart rhythm data from a person, as a function of time, thereby obtaining a time series of heart rhythm data. The time series of heart rhythm data is supplied to an analysis engine 3. An analysis module 4 comprising a mathematical model identifies a plurality of sub-sequences in the time series of heart rhythm data, each sub-sequence corresponding to a heartbeat. This could, e.g., include identifying a plurality of waves of the heartbeats, such as a plurality of R waves, and defining each sub- sequence as a suitable time interval around each identified R wave.

2 DK 2022 01081 A1

The analysis module 4 further categorizes and groups the sub-sequences, using the mathematical model, preferably based on patterns in the sub-sequences which are recognised by the mathematical model. Thus, the sub-sequences are grouped in such a manner that sub-sequences which are similar to each other are grouped together.

Furthermore, the groups of sub-sequences are labelled.

A visualisation module superimposes the sub-sequences on each other, based on one or more recognisable features of a heartbeat, such as one or more of the well-known waves of a heartbeat. Accordingly, in the superimposition, the heartbeats are aligned with respect to one or more of the characteristic features. The visualisation module 5 further prepares a visual representation of the superimposed sub-sequences, and the visual representation is displayed 6 to a medical professional performing the interpretation of the heart rhythm data.

In an alternative embodiment, the sub-sequences may be superimposed by the analysis module 4.

A diagnosis module 7 analyses the sub-sequences, possibly in the form of the superimposed sub-sequences, in order to identify diagnostically relevant features in the heart rhythm data.

Based thereon, an alert module 8 may cause diagnostic information 9 and/or an early warning alert 10 to be displayed to the medical professional. The diagnostic information 9 may, e.g., include pre-diagnostics, such as one or more possible diagnoses, based on recognised patterns in the heart rhythm data. The early warning alert 10 may, e.g., include a warning that there is an increased risk of a condition occurring in the future, based on recognised patterns in the heart rhythm data.

Fig. 3 illustrates a measured time series of heart rhythm data, i.e., the kind of data which a medical professional is normally presented with. It can be seen that direct comparison between the individual heartbeats can not be readily performed, and it requires extensive experience and training to perform a correct interpretation of the heart rhythm data.

Furthermore, this process is time consuming.

Fig. 4 illustrates superimposed sub-sequences of heart rhythm data obtained by means of a method according to an embodiment of the invention. The sub-sequences have been superimposed by aligning the R waves or the QRS complexes of the heartbeats corresponding to the sub-sequences. The superimposed sub-sequences allow the heartbeats to be readily compared, and that anomalies, outliers, trends, etc., can easily be identified. Thereby the medical professional is able to perform a thorough, accurate and reliable interpretation of the heart rhythm data in a fast manner.

Claims

1. A method for identifying morphological abnormalities in heart rhythm data, the method comprising the steps of: — analysing a time series of heart rhythm data representing heartbeats of a person, — identifying a plurality of sub-sequences in the time series of heart rhythm data, each sub-sequence corresponding to a single heartbeat, and isolating the sub-sequences, — feeding the sub-sequences to a mathematical model, categorizing, and grouping the sub-sequences using the mathematical model, and labelling each group of sub- sequences, — superimposing the sub-sequences on each other, based on one or more recognisable features of a heartbeat, and — collectively form the clinical decision support for the medical professional.

2. A method according to claim 1, wherein the step of measuring a time series of heart rhythm data is performed by means of an implantable device and/or by means of a wearable device.

3. A method according to claim 1 or 2, further comprising the step of identifying outliers and/or known diagnostic variances in the sub-sequences, and wherein the step of displaying the superimposed sub-sequence comprises highlighting identified outliers and/or known diagnostic relevant variances.

4. A method according to any of the preceding claims, wherein the step of superimposing the sub-sequences on each other comprises recognising one or more waves of each heartbeat, and aligning features of the recognised waves of the sub-sequences.

5. A method according to any of the preceding claims, wherein the step of categorizing and grouping the sub-sequences is performed based on known and diagnostically relevant features of the heartbeat variances.

6. A method according to any of the preceding claims, wherein the step of identifying a plurality of sub-sequences in the time series of heart rhythm data comprises identifying a plurality of waves in the time series of heart rhythm data, and, for each identified wave,

identifying a sub-sequence as a time interval from a predefined time before a start, and/or end, and/or middle of the wave to a predefined time after the start, and/or end, and/or middle of the wave.

7. A method according to any of the preceding claims, further comprising the step of a user manually flagging a part of the time series of heart rhythm data, during the step of measuring the time series of heart rhythm data, and wherein the step of displaying the superimposed sub-sequences comprises highlighting sub-sequences corresponding to the flagged part of the time series of heart rhythm data.

8. A method according to any of the preceding claims, further comprising the step of performing automated pre-diagnostics based on the displayed superimposed sub-sequences and/or on the categorization of the sub-sequences.