Method and system for detecting heartbeat irregularities

Abstract

There is a method and system for detecting heartbeat irregularities comprising the steps of receiving a dataset representative of at least one waveform, the at least one waveform indicative of a subject's heart activity over a predetermined period of time; identifying from the data representative of at least one waveform, a plurality of peaks, each peak corresponding to a heartbeat; identifying from the predetermined period of time the time occurrence of each peak; calculating the difference (duration) between the time occurrence of each peak with its adjacent peak; determining the difference between each duration; classifying the absolute value of the difference into one of at least three intermediate categories; wherein each intermediate category comprises a specified range such that the absolute value is categorized into the intermediate category if it falls between the range; the intermediate categories further providing an indication of whether the subject has heartbeat irregularity.

Claims

1. A method for detecting heartbeat irregularities comprising the steps of: a. receiving a dataset representative of at least one waveform, the at least one waveform indicative of a subject's heart activity over a predetermined period of time; b. identifying from the data representative of at least one waveform, a plurality of peaks, each peak corresponding to a heartbeat; c. identifying from the predetermined period of time the time occurrence of each peak; d. calculating the duration difference (P.sub.n−(n+1)) between the time occurrence of each peak with its adjacent peak; e. determining the difference (D.sub.n)between each duration difference (P.sub.n −(n+1)) calculated in step d; and f. classifying the absolute value (|D.sub.n|) of the difference (D.sub.n) into one of at least three intermediate categories; wherein the first intermediate category of the at least three intermediate categories has a specified range of between 0 to 5; the second intermediate category of the at least three intermediate categories has a specified range of between 6 to 11; and the third intermediate category of the at least three intermediate categories has a specified range of 12 and above; and wherein if all of the classified absolute values |D.sub.n|occurs in the first intermediate category, with no occurrences in the second and third categories, the at least one waveform is categorized as ‘Regular heartbeat’; wherein if there are a number of occurrences |D.sub.n| in all three Intermediate categories, the at least one waveform is categorized as ‘Irregularly irregular’ heartbeat; and wherein in all other cases the at least one waveform is categorized as ‘Regularly irregular’ heartbeat.

2. A method according to claim 1, wherein the waveform indicative of the subject's heart activity over a period of time is an arterial pulse waveform, an ECG signal or a time series of obtained camera frames captured based on variations in finger skin colour and brightness that occur due to blood pulsation.

3. A method according to claim2, wherein where the at least one waveform is an arterial pulse waveform, the peaks are determined and identified based on the identification of dicrotic notches as well as the gradient of the upstroke and downstroke identified on the at least one waveform.

4. A method according to claim 2, wherein where the at least one waveform is a time series of obtained camera frames, the method further comprises a conversion step before step c.

5. A method according to claim 4, wherein the conversion step includes the step of accounting for the discrepancy in sampling rates across different mobile devices using the following mathematical expression:
f(HR.sub.n)=(60 seconds×S)/Δt.sub.n−(n+1) wherein f(HR.sub.n) denotes beats per minute of each heartbeat, S denotes the sampling rate of the captured waveform; and t.sub.n−(n+1) denotes time units in milliseconds between each peak.

6. A mobile device having camera and flash capabilities, the mobile device operable to obtain a time series of obtained camera frames captured based on variations in finger skin colour and brightness that occur due to blood pulsation when a subject's finger is positioned against the camera lens and flash; and upon obtaining the time series detects whether the subject has heartbeat irregularity according to the method of claim 2.

7. A system for detecting heartbeat irregularities comprising a measurement device for receiving a dataset representative of at least one waveform, the at least one waveform indicative of a subject's heart activity over a predetermined period of time; a processor arranged to receive the dataset and further arranged to: identify from the at least one waveform, a plurality of peaks, each peak corresponding to a heartbeat; identify from the predetermined period of time the time occurrence of each peak; calculate the duration difference (P.sub.n−(n+1)) between the time occurrence of each peak with its adjacent peak; determine the difference (D.sub.n)between each duration difference (P.sub.n−(n+1)); and classify the absolute value (|D.sub.n|) of the difference (D.sub.n) into one of at least three intermediate categories; wherein the first intermediate category of the at least three intermediate categories has a specified range of between 0 to 5; the second intermediate category of the at least three intermediate categories has a specified range of between 6 to 11; and the third intermediate category of the at least three intermediate categories has a specified range of 12 and above; and wherein if all of the classified absolute values |D.sub.n,| occurs in the first intermediate category, with no occurrences in the second and third categories, the at least one waveform is categorized as ‘Regular heartbeat’; wherein if there are a number of occurrences |D.sub.n| in all three Intermediate categories, the at least one waveform is categorized as ‘Irregularly irregular’ heartbeat; and wherein in all other cases the at least one waveform is categorized as ‘Regularly irregular’ heartbeat.

8. A system according to claim 7, wherein the waveform indicative of the subject's heart activity over a period of time is an arterial pulse waveform, an ECG signal or a time series of obtained camera frames captured based on variations in finger skin colour and brightness that occur due to blood pulsation.

9. A system according to claim 8, wherein where the at least one waveform is an arterial pulse waveform, the peaks are determined and identified based on the identification of dicrotic notches as well as the gradient of the upstroke and downstroke identified on the at least one waveform.

10. A system according to claim 8, wherein where the at least one waveform is a time series of obtained camera frames, the system further comprises a conversion step before the processor identify from the predetermined period of time the time occurrence of each peak.

11. A system according to claim 10, wherein the conversion step includes the step of accounting for the discrepancy in sampling rates across different mobile devices using the following mathematical expression:
f(HR.sub.n)=(60 seconds×S)/Δt.sub.n−(n+1) wherein f()HR.sub.n) denotes beats per minute of each heartbeat, S denotes the sampling rate of the captured waveform; t.sub.n−(n+1) denotes time units in milliseconds between each peak.

12. A system according to claim 8, wherein the measurement device is a real time beat to beat blood pressure monitoring device.

13. A system according to claim 8, wherein the measurement device is a mobile device with camera and flash capabilities.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The invention will now be described, by way of example only, with reference to the accompanying drawings, in which:

(2) FIG. 1 is a flow chart of a method of determining the presence of arrhythmia in accordance with a first embodiment of the present invention.

(3) FIGS. 2a, 2b, and 2c are illustrative pulse waveforms showing the various types of heartbeats.

(4) FIGS. 3a, 3b, and 3c are illustrative pulse waveforms with additional segmentation to show how the peaks and peak differences are calculated for the various types of heartbeats.

(5) FIGS. 4a, 4b, and 4c are illustrative plots showing the differences in peak to peak duration for the various types of heartbeats;

(6) FIG. 5 is a system diagram of the invention in accordance with another embodiment of the invention.

(7) FIGS. 6a to 6c are illustrations of another embodiment of the invention wherein a mobile device is used for the acquisition of data and the detection of arrhythmia; and

(8) Other arrangements of the invention are possible and, consequently, the accompanying drawings are not to be understood as superseding the generality of the preceding description of the invention.

PREFERRED EMBODIMENTS OF THE INVENTION

(9) Particular embodiments of the present invention will now be described with reference to the accompanying drawings. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of the present invention. Additionally, unless defined otherwise, all technical and scientific terms used herein have the same meanings as commonly understood by one or ordinary skill in the art to which this invention belongs.

(10) In accordance with an embodiment of the invention there is a method 110 of detecting heartbeat irregularities and in particular (but not limited to), determining the presence of arrhythmia and differentiating between different types of arrhythmia such as between Atrial Fibrillation (Irregularly irregular heartbeat′) and ‘regularly irregular heartbeat’. The method 110 comprises steps as illustrated in flow chart form in FIG. 1. FIGS. 2a, 2b and 2c illustrate the different types of heartbeats types, namely ‘regular’, ‘irregularly irregular’ and ‘regularly irregular’ respectively.

(11) As shown in FIG. 1, the method 110 commences by capturing a block of waveform in step 112. In this embodiment the waveform block is a pulse waveform. More preferably, the waveform block is an arterial pulse waveform which is a mechanical manifestation of one or more electrical heartbeat signal(s). The block of waveform may be captured by a suitable heart rate monitor or real time beat-to-beat blood pressure monitoring device 202 such as the BPro® device of Healthstats International Pte Ltd. Each waveform block should preferably comprise at least enough heartbeats deemed necessary for determining whether the heartbeats are ‘regular’, ‘irregularly irregular’ and ‘regularly irregular’. As such, the block of waveform should be captured for a period of time as predetermined by a user or medical practitioner. The predetermined period of time may be any period of time between 5 to 60 seconds for example, as long as the period of time is adequate for purpose of analysis and classification. Preferably the block of waveform is between 8 to 30 seconds to capture a block of waveform comprising enough heartbeats for determining regularity or irregularity. Longer or shorter periods of time may be utilized or determined by the user or medical practitioner depending on the measurement device used to capture the block of waveform.

(12) A factor for determining the predetermined period of time is based on the sampling rate of the measurement device 202 used to obtain the block of waveform. For example, a measurement device 202 having sampling rate for obtaining arterial waveform data at 60 Hertz (Hz) require around 5 to 8 seconds to obtain the necessary data to determine a pulse waveform. A measurement device 202 having sampling rate of less than 60 Hz may require longer time to obtain the pulse waveform necessary for analysis.

(13) A measurement device having a sampling rate of 60 Hz at 5 to 8 seconds will obtain about five heart beats, which is deemed suitable for determining whether the heart beats are ‘regular’, ‘irregularly irregular’ or ‘regularly irregular’.

(14) After the block of waveform is captured, the method 110 may check whether the waveform is valid based on step 114. An exemplary valid waveform should typically have up-slopes 114a and down-slopes 114b signifying the pumping of the ventricular valves. There are existing methods of determining whether a block of waveform is valid and any one of them may be used.

(15) Upon determining that the block of waveform is a valid waveform, the method goes on by determining and identifying the number of peak positions on the obtained pulse waveform within the predetermined block of time in step 116. Each peak position corresponds to a heartbeat. For purpose of naming convention, the peak positions are labelled P.sub.1, P.sub.2, P.sub.3, . . . P.sub.n. A suitable method for determining and identifying the peak positions and heartbeats on the pulse waveform is based on the identification of dicrotic notches as well as the gradient of the upstroke and downstroke as described in WO/2002/030277 and will not be further elaborated.

(16) Once the peak positions are determined and identified, the method goes on in step 118 to determine the peak to peak duration (duration between each heartbeat) using equation (1):
P.sub.n−(n+1)=P.sub.n+1−P.sub.n  (1)

(17) Wherein P.sub.n−(n+1) denotes the total number of sampling points (or relative time units in milli-seconds) between each peak. Typically, a 10 second block is preferably used because under normal conditions a healthy human heart beat should not deviate by more than 5 beats per minute.

(18) Upon determining the peak-to-peak duration, the difference between each peak to peak duration, D.sub.n (typically measured in terms of normalized or relative time scale in milliseconds), can then be calculated in step 120 using equation (2).
D.sub.n=P.sub.(n+1)−(n+2)−P.sub.n−(n+1)  (2)

(19) For example in FIG. 3a; D.sub.1=P.sub.2-3−P.sub.1-2=−2; D.sub.2=P.sub.3−4−P.sub.2-3=−1;

(20) in FIG. 3b; D.sub.1=−10; D.sub.5=24; etc.

(21) The obtained D.sub.n will be next categorized into at least three intermediate categories based on its absolute value i.e. |D.sub.n|. (step 122). Each intermediate category comprises a range of the absolute value. The three sub-categories (A, B, and C) may be:— A—where |D.sub.n| is between 0 to 5; B—where |D.sub.n| is between 6 to 11; C—where |D.sub.n| is equals or greater than 12.

(22) Using the D.sub.1=−10 and D.sub.5=24 of FIG. 3b as examples, D.sub.1 has an absolute value of 10 and will thus be categorized under category B and D.sub.5 has an absolute value of 24 and will thus be categorized under category C.

(23) The method 110 will also collate the number of occurrences in categories A, B and C.

(24) The method 110 then determines whether Arrhythmia or Atrial Fibrillation was detected based on the treatment of the intermediate categories according to the following rules:—

(25) If all the number of occurrences |D| occurs in Intermediate category A only (i.e. with no occurrences in intermediate categories B and C, the pulse waveform block is categorized as ‘Regular heartbeat’;

(26) If there are a number of occurrences |D| in all Intermediate categories A, B and C, the pulse waveform block is categorized as ‘Irregularly irregular’ heartbeat;

(27) For all other cases, the pulse waveform block is categorized as ‘Regularly irregular heartbeat.

(28) (see step 126)

(29) Optionally, to improve accuracy, steps 112, 114, 116, 118, 120, 122 may be repeated for additional blocks of waveform for the same subject (step 128). Typically steps 112, 114, 116, 118, 120, 122 may be repeated for preferably at least 3 waveform blocks, and optionally recommended at around 5 to 10 waveform blocks.

(30) FIGS. 3a, 3b and 3c show how the method determines peak positions 116 and determines the peak to peak duration 118 and displaying the differences between peak to peak 120, D.sub.n, below the respective graphs.

(31) It can be seen from plotting out the differences between peak to peak duration 120, D.sub.n, for the various heartbeats show different graphs in FIGS. 4a, 4b and 4c for the regular heartbeat, irregularly irregular heartbeat (atrial fibrillation) and regularly irregular heartbeat (arrhythmia) respectively. From the trends of these plots, the method can then determine what type of heartbeat for each block of waveforms and thereafter determine one of three scenarios: 1) No arrhythmia detected—this conclusion, is reached if all the blocks of waveforms are classified as regular heartbeats. 2) Atrial fibrillation detected—this conclusion is reached if any irregularly irregular heartbeats are found in the blocks of waveforms. 3) Arrhythmia detected—this conclusion is reached if there are no irregularly irregular heartbeats found in the blocks of waveforms, and there are instances of regularly irregular′ heartbeats found.

(32) In accordance with another embodiment of the invention there is a system 200 of determining the presence of arrhythmia. The system 200 comprises a pulse waveform measurement device 202, preferably arterial pulse waveform measurement device for obtaining pulse waveform from a subject (typically a person). The measurement device may be invasive or non-invasive, as long as the measurement device is capable of obtaining real time beat-to-beat pulse waveform. This is to be distinguished from pulse waveforms obtained based on extrapolation or averaging methods.

(33) The arterial pulse waveform measurement device 202 may preferably be a non-invasive real-time beat-to-beat blood pressure monitoring device such as the BPro® device of Healthstats International Pte Ltd.

(34) System 200 further comprises a processing device 204 for obtaining and storing the arterial pulse waveform measurements obtained. The processing, device may be a computer or mobile device as known to a skilled person in the art. The mobile computing device may optionally be integrated with the non-invasive real-time beat-to-beat blood pressure monitoring device 202 as described above.

(35) Upon receipt of the arterial pulse waveform measurements, the processing device is operable to perform the method 110 to determine if Arrhythmia is present or absent, and if present, whether it is atrial fibrillation.

(36) In accordance with another embodiment of the invention, where like reference numerals designate like features, there is a mobile device 400 for determining the presence of arrhythmia. Mobile device 400 is preferably a smart phone having camera 402 and flash 404 capabilities. Mobile device 400 is capable of installing thereon a dedicated software application 406 (colloquially known as ‘apps’) suitable for download on an Android™ platform, for example. Dedicated software application 406 is operable to access and activate the camera flash function to detect the heart rate of a person and thereby obtain a waveform. In addition, dedicated software application 406 is capable of implementing the method 110 as described in the earlier embodiment(s) to determine the presence of heartbeat irregularity/arrhythmia.

(37) Instead of a pulse waveform as described in the earlier embodiment(s), the block of waveform as mentioned in step 112 is a time series of obtained camera frames captured based on variations in finger skin colour and brightness that occur due to blood pulsation when a person's finger (preferably but not limited to an index finger of a person) is lightly placed against the camera lens 402 and flash 404.

(38) The detection of variations in finger skin colour and brightness is based on analysing average red component values of the frames or part of the frames taken by the camera. It is to be appreciated that other colour components (blue, green) are generally discarded.

(39) The time series of average red component values of the obtained frames is considered as the captured block of waveform 408 for heart rate measuring. The time series of average red component values of the obtained signal comprises “sharp”—local maxima, each sharp local maxima corresponding to a single heartbeat. It is to be appreciated that the number of heart beats and length of the measurement are the two variables required to calculate the heart rate.

(40) An optional filtering step may be used to filter any noise from the time series if required.

(41) After filtering, the time series signal 408 is converted from the ‘obtained frames’ form into a form suitable for analysis by the method 110. The time series signal 408 is typically converted into a form where each peak corresponds to a heartbeat.

(42) An example of the conversion comprises three (3) steps, wherein the first and second steps are similar to the concept of step 118 and step 120:—

(43) First step: to determine the time between each peak positions calculated from the first to the n.sup.th peak (t.sub.1 . . . t.sub.n); i.e. and

(44) Second step: to determine the peak-to-peak duration (i.e. duration between each heartbeat using the following equation (2a)
Δt.sub.n−(n+1)=t.sub.n+1−t.sub.n  (2a)
wherein t.sub.n−(n+1) denotes time units in milliseconds between each peak.

(45) It is to be appreciated that the sampling rate of the captured waveform may differ for different waveform(s) captured on different mobile devices 400. For example, thirty (30) frames may be the norm for some mobile devices, while for others it may be higher or lower. To account for the discrepancy in sampling rates across different mobile devices 400, a mathematical equation (3) is used to account for the sampling rate to a form suitable for analysis by method 110.

(46) Third step: Find the number of beats per minute of each heartbeat using the equation (3):
f(HR.sub.n)=(60 seconds×S)/Δt.sub.n−(n+1)  (3)

(47) Wherein f(HR.sub.n) denotes beats per minute of each heartbeat, and S denotes the sampling rate of the captured waveform.

(48) Once the number of beats per minute of each heartbeat is determined, the method 110 is then used to calculate and tabulate the number of occurrences of ‘regular’, ‘regularly irregular’ and ‘irregularly irregular’ heartbeats in accordance with steps 118, 120, 122, 124 accordingly.

(49) Based on the frame sampling number, a suitable predetermined period of 10-15 seconds for may be chosen.

(50) The current embodiment is advantageous in that it is highly mobile and is targeted at the public, particularly for those who are in the high risk group, such as patients suffering from hypertension, diabetes mellitus, heart disease or have a family history of stroke or sudden death. The App is also useful for patients who know of their AF condition and are being medically treated as it will be able to show the effectiveness of control.

(51) Clinical Trials

(52) To test the efficiency and accuracy of the method 110 in various described embodiments, the algorithm is tested on thirty (30) subjects based on the following parameters:—

(53) time period of measurement—10 seconds;

(54) sampling rate of 60 Hz; and

(55) the non-invasive beat-to-beat blood pressure monitoring device 202—BPrO™.

(56) The clinical results are tabulated in the form of Table 1 below:—

(57) TABLE-US-00001 TABLE 1 Clinical Data for Arrhythmia & Atrial Fibrillation (AF) detection algorithm against actual detection Algorithm Result Result File Actual detected (True (True No. Name detection results Result Positive) Negative) 1 Subject AF AF True 1 001 positive 2 Subject AF AF True 1 002 positive 3 Subject AF AF True 1 003 positive 4 Subject AF AF True 1 004 positive 5 Subject AF AF True 1 005 positive 6 Subject AF AF True 1 006 positive 7 Subject AF AF True 1 007 positive 8 Subject AF AF True 1 008 positive 9 Subject AF AF True 1 009 positive 10 Subject AF AF True 1 010 positive 11 Subject AF AF True 1 011 positive 12 Subject AF AF True 1 012 positive 13 Subject AF AF True 1 013 positive 14 Subject Arrhyth- Arrhyth- True 1 014 mia mia positive 15 Subject Arrhyth- Arrhyth- True 1 015 mia mia positive 16 Subject Arrhyth- Arrhyth- True 1 016 mia mia positive 17 Subject Arrhyth- Arrhyth- True 1 017 mia mia positive 18 Subject Arrhyth- Arrhyth- True 1 018 mia mia positive 19 Subject Sinus Sinus True 1 019 Rhythm Rhythm negative 20 Subject Sinus Sinus True 1 020 Rhythm Rhythm negative 21 Subject Sinus Sinus True 1 021 Rhythm Rhythm negative 22 Subject Sinus Sinus True 1 022 Rhythm Rhythm negative 23 Subject Sinus Sinus True 1 023 Rhythm Rhythm negative 24 Subject Sinus Sinus True 1 024 Rhythm Rhythm negative 25 Subject Sinus Sinus True 1 025 Rhythm Rhythm negative 26 Subject Sinus Sinus True 1 026 Rhythm Rhythm negative 27 Subject Sinus Sinus True 1 027 Rhythm Rhythm negative 28 Subject Arrhyth- Arrhyth- True 1 028 mia mia positive 29 Subject Sinus Sinus True 1 029 Rhythm Rhythm negative 30 Subject Sinus Sinus True 1 030 Rhythm Rhythm negative

(58) Summary of Results: True Positive (TP): 19 False Negative (FN): 0 True Negative (TN): 11 False Positive (FP): 0 Sensitivity: TP/(TP+FN)=100% Specificity: TN/(TN+FP)=100%

(59) It is to be appreciated that for the ‘actual detection’ labelled in the second column, the actual detection may be based on any currently established methods used to determine arrhythmia; sinus rhythm (i.e. normal heartbeat); and Atrial Fibrillation (AF).

(60) It can be appreciated by a person skilled in the art that the above invention is not limited to the embodiments described. In particular, the following modifications and improvements may be made without departing from the scope of the present invention: As there is a correlation between ECG waveform and arterial pulse waveform, an ECG waveform may be used in place of the pulse waveform (subject to the mechanical-electrical dissociation), so long as the algorithm is able to detect, tabulate and classify the various forms of heartbeats and thereafter make a determination whether arrhythmia not detected, arrhythmia detected or atrial fibrillation detected. Similar methods for converting the obtained ECG waveform to extract salient parameters including:—number of peaks; duration between peaks; etc for usage of the method 110 may be used. The device can also record and log the date and time whenever arrhythmia or atrial fibrillation is detected. Instead of obtaining measurements via the camera 402 and flash 404, the integrated mobile device 400 may obtain measurements from the arterial pulse waveform measurement device 202 for analysis. The data from the pulse waveform measurement device 202 may be sent to the mobile device 400 via wireless means such as (but not limited to) infra-red or Bluetooth.

(61) The above is a description of embodiments of a system and method of detecting heartbeat irregularities in accordance with this invention. It is envisioned that those skilled in the art can and will design an alternative embodiment of this invention that infringe on this invention as set forth in the followings claims. It is also to be further appreciated that various aspects of the embodiments as described may be combined to form further embodiments without departing from the scope of the invention.