WO2010000009A1

WO2010000009A1 - Improved detection of cardiac dysfunction

Info

Publication number: WO2010000009A1
Application number: PCT/AU2008/000972
Authority: WO
Inventors: Aaron Ginzburg; Andrew Maxwell Tonkin; Alexander Tsintsiper
Original assignee: Cardanal Pty Ltd
Priority date: 2008-07-02
Filing date: 2008-07-02
Publication date: 2010-01-07

Abstract

A method of measuring heart function of a patient comprises the step of performing an electrocardiogram (ECG) measurement of the patient over a continuous time interval having a specified duration, to obtain a corresponding sequence of ECG data. The sequence of ECG data is divided into a plurality of subsequences, each of which is processed to obtain a corresponding plurality of subsequence index values. In particular, the processing comprises extracting data corresponding with specified ECG signal segments, determining a time derivative of the extracted data, and generating a subsequence index value which is representative of energy in said time derivative. A statistical mean value and a statistical variation value of the plurality of subsequence index values are calculated, and used to compute a heart function index value which is a measure of heart function of the patient. Systems and apparatus implementing the method are also provided. The invention advantageously enables a preliminary assessment of possible abnormality of heart function of a patient based upon a single ECG measurement.

Description

IMPROVED DETECTION OF CARDIAC DYSFUNCTION FIELD OF THE INVENTION

The present invention relates to diagnosis of possible heart dysfunction. More particularly, the invention is directed to providing improved methods and apparatus for detecting possible abnormality of heart function of a patient by processing electrocardiogram (ECG) signals gathered over a single continuous monitoring session. BACKGROUND OF THE INVENTION

A method and system for processing ECG signals in order to identify possible abnormality of heart function of a patient is disclosed in commonly assigned US patent application serial no. 10/865,985, filed on 6 November 2004, and published under publication no. US 2005/0027202 on 2 March 2005. In accordance with this prior art method, data is extracted from an ECG signal of the patient, and a time derivative of the data is determined. A normalised index value, representative of energy in the time derivative, is then generated.

While the normalised index value computed in accordance with this prior art method has been shown to provide a useful indication of possible abnormality of heart function of the patient, the prior art methods and systems suffer from a number of limitations. In particular, it has previously been considered necessary to monitor the patient over an extended period of time. Specifically, the normalised index value computed in accordance with the above-described method provides only a relative indication of patient health. That is, a "normal" value of the index for a particular individual may be very different from a corresponding "normal" value for a different individual. It was accordingly not considered possible to identify potential problems based upon a single period of monitoring of a patient's ECG waveform. Therefore, the normalised index was proposed to be used as part of an ongoing monitoring process of the patient, wherein ECG measurements of the patient would be recorded at different times, enabling a historical average of the value of the normalised index value to be computed for each individual patient. Under these conditions, it was demonstrated in animal trials, and in clinical trials including comparative tests relative to the diagnosis of expert cardiologists, that the normalised index value is a useful and potentially highly sensitive indicator of abnormality of heart function, detectable by a deviation, in a subsequent measurement, of the index value from the historical average.

This method of diagnosis is obviously extremely useful for patients who are subject to ongoing monitoring. However, a patient will only enter into a program of ongoing monitoring after a potential problem has been identified, such as following a heart attack or other heart-related episode. The prior art system and method is therefore not so effective for identifying possible abnormalities of heart function in new patients who have not previously shown any relevant symptoms, and therefore been subject to ongoing monitoring. It is therefore clearly desirable to provide improved methods, systems and apparatus that are capable of alerting health professionals, and particularly health professionals such as nurses and general practitioners who are not expert cardiologists, to the possible abnormality of heart function of patients presenting for general health checkups, or for diagnosis in relation to symptoms that may or may not be indicative of heart problems. As noted in the background section of the abovementioned prior art patent publication, the early identification of heart dysfunction has the potential to save many thousands of lives in developed countries annually.

It is accordingly an object of the present invention to provide improved methods, systems and apparatus for detecting possible abnormality of heart function, which are capable of generating an immediate indication of potential heart dysfunction based upon a single ECG measurement of the patient over a continuous time interval of specified, and/or reasonable, duration. SUMMARY OF THE INVENTION In one aspect, the present invention provides a method of measuring heart function of a patient comprising the steps of: performing an electrocardiogram (ECG) measurement of the patient over a continuous time interval having a specified duration, to obtain a sequence of ECG data corresponding with an ECG signal over said time interval; dividing the sequence of ECG data into a plurality of subsequences; processing each subsequence of ECG data to obtain a corresponding plurality of subsequence index values; computing a statistical mean value and a statistical variation value of the plurality of subsequence index values; and using said statistical mean value and said statistical variation value to compute a heart function index value which is a measure of heart function of the patient, wherein said processing step comprises, for each subsequence of ECG data: extracting data corresponding with specified ECG signal segments from the subsequence of ECG data; determining a time derivative of said extracted data; and generating a subsequence index value which is representative of energy in said time derivative.

Advantageously, therefore, embodiments of the present invention enable a preliminary assessment of possible abnormality of heart function of a patient based upon a single ECG measurement. The heart function index value, computed in accordance with the aforementioned method, may be immediately indicative of potential problems, and a past history of patient monitoring is not necessary in order to identify any change. Accordingly, while the method does not provide a diagnosis of the cause of any identified abnormality, and therefore does not replace the expertise of a qualified cardiologist, it has the advantage that it may be implemented in an entirely automated system, such that the necessary measurements may be conducted by health care professionals, such as general practitioners or nurses, who do not have particular expertise in cardiology. The heart function index value output from such a system may be utilised by a health care professional in order to recommend any possible further investigation or treatment for the patient. The invention therefore advantageously facilitates improved early detection of possible cardiac problems, even when a patient may not be exhibiting any symptoms of abnormal heart function. Embodiments of the invention therefore have the potential to improve patient outcomes through early diagnosis, and even to save many thousands of lives as a result.

Advantageously, all data processing is performed digitally. In this context, it will be understood that the ECG data is a time-ordered sequence of numerical values corresponding with a sampled and digitised ECG signal waveform. Preferably the step of generating a subsequence index value comprises: determining frequency components of the time derivative; and calculating a subsequence index value representative of the energy in said frequency components. In accordance with preferred embodiments, the specified duration of the

ECG measurement is at least 120 seconds. Shorter or longer durations of measurements are, however, possible. In particular, it may be useful to perform longer measurements, such as over 3, 4 or 5 minutes, in order to gather additional information for use in computing the heart function index value. However, much longer measurement intervals, for example in excess of 5 minutes, are less desirable, due to the need for the patient to remain connected to the electrocardiograph for the full specified duration.

Advantageously, the heart function index value is computed as a ratio of said statistical variation value to said statistical mean value. In accordance with this preferred embodiment, the heart function index value is lower when there is less variation amongst the subsequence index values in different subsequences, and is greater when there is more variation amongst the subsequence index values in different subsequences. It has been discovered that, even over relatively short measurement periods, such as those suggested above, this measure of variation can be a sensitive and practical indicator of possible abnormality of heart function. While there is no clear dividing line between "normal" and "abnormal", it is envisaged that a sliding scale may be developed, in accordance with which suitable recommendations may be made for further investigation and/or treatment, based upon the heart function index value of the patient.

In preferred arrangements, the ECG measurement is performed over a plurality of standard ECG leads, and a separate sequence of ECG data is obtained and processed for each lead, in order to compute a plurality of heart function index values, each of which corresponds with a single ECG lead. In particular, the standard 12-lead ECG may be employed. As will be known to persons skilled in the application and interpretation of ECG data, the 12-lead standard comprises leads I, II, III, v1 to v6, aVF, aVL and aVR. Furthermore, it will be appreciated that the term "standard 12-lead ECG" refers to a well- established set of measurements familiar to persons skilled in the art, but that the 12 ECG waveforms obtained by these measurements do not necessarily correspond with 12 physical wires and/or electrodes. More particularly, the word "lead" has two meanings in the art of electrocardiography, and may refer either to a physical wire that connects an electrode to an electrocardiograph, or (more commonly) to a combination of electrodes that form an imaginary line in the body along which electrical signals are measured. In practice, a standard 12-lead ECG commonly uses only 10 wires and electrodes. Within the present specification, the word "lead" is exclusively intended to have the latter of the forgoing two meanings.

The method preferably includes computing a global heart function index value which is an average of the heart function index values corresponding with each of said plurality of standard ECG leads. Additionally, the method may include computing one or more cluster heart function index values, each of which corresponds with a selected group of said plurality of standard ECG leads. In particular, the group of standard ECG leads is preferably selected from the set of groups comprising: leads II, III and aVF (inferior group); leads I, aVL, v5 and v6 (lateral group); and leads v1 to v6 (anterior group).

Since the foregoing preferred groups of leads correspond with activity in particular regions of the heart, the associated cluster heart function index values are more preferably termed "regional heart function index values".

The cluster/regional heart function index values may be computed as an average of the heart function index values corresponding with each ECG lead of the selected group.

In preferred embodiments, the step of processing each subsequence of ECG values includes automatically identifying specified points and/or intervals of individual beats of the ECG signal. Preferably the specified points and/or intervals are selected from the group comprising: Q-peak; R-peak; S-peak; P-wave; T-wave; J-point; QRS complex; PR-segment; ST-segment; PR-interval; and QT-segment. The specified signal segments from which data is extracted are preferably selected from the group comprising: QRS complexes, ST-segments, T-waves and QT-segments. However, this group is not intended to be limiting, and it is reasonably foreseeable that with continuing investigation and analysis various segments of the ECG waveform may be found to provide useful and relevant information in relation to patient heart function.

In a preferred implementation, automatically identifying specified points and/or intervals includes the steps of: filtering or smoothing the ECG signal in order to reduce noise; computing one or more time derivatives of the noise-reduced ECG signal; identifying Q-peak, R-peak and S-peak as points having the highest values amongst all data values of the computed derivatives; identifying J-point as a point having the largest curvature amongst the noise-reduced ECG signals following the S-peak in each beat; and identifying P-wave and T-wave by identifying points having the highest curvature of the noise-reduced signal values between identified QRS complexes of adjacent beats.

Advantageously, the method additionally includes performing automated quality control by excluding from further computations data corresponding with beats of the ECG signal that do not satisfy specified quality criteria. One preferred form of quality control includes computing a noise value and a signal-to- noise ratio for each specified signal segment, and excluding from further computations any segment for which the signal-to-noise ratio is less than a predetermined value. An additional preferred form of quality control includes determining the number of beats or complexes successfully identified for each ECG lead, and excluding from further computations data corresponding with any lead for which the number of successfully identified beats or complexes is less than a predetermined value.

In another aspect, the present invention provides a system for measuring heart function of a patient. The system includes: an electrocardiograph having a plurality of ECG leads, the electrocardiograph being configured to monitor the electrical activity of the patient's heart over a continuous time interval having a specified duration, and to output at least one sequence of ECG data comprising digitised representations of the electrical activity monitored by each ECG lead; a processor configured to receive the sequence of ECG data output by the electrocardiograph, wherein said processor includes: means for dividing the sequence of ECG data into a plurality of subsequences; means for processing each subsequence of ECG data to obtain a corresponding plurality of subsequence index values; means for computing a statistical mean value and a statistical variation value of the plurality of subsequence index values; and means for using said statistical mean value and said statistical variation value to compute a heart function index value which is a measure of heart function of the patient, wherein said means for processing each subsequence of ECG data is adapted to: extract data corresponding with specified ECG signal segments from the subsequence of ECG data; determine a time derivative of said extracted data; and generate a subsequence index value which is representative of energy in said time derivative.

In a particularly preferred embodiment, the electrocardiograph is located remotely from the processor, and the ECG data are transmitted from the electrocardiograph to the processor via a data communications network.

In a further aspect, the present invention provides an apparatus for measuring heart function of a patient, including: a central processing unit; at least one memory device operatively associated with the central processing unit; and a communications interface operatively associated with the central processing unit, the interface providing a channel for communication with an electrocardiograph having a plurality of ECG leads, and which is configured to monitor the electrical activity of the patient's heart over a continuous time interval having a specified duration, and to output at least one sequence of ECG data comprising digitised representations of the electrical activity monitored by each ECG lead; the memory device having computer_^executable instructions stored therein which, when executed by the central processing unit, cause the apparatus to effect the steps of: receiving, via the communications interface, a sequence of ECG data output by the electrocardiograph; dividing the sequence of ECG data into a plurality of subsequences; processing each subsequence of ECG data to obtain a corresponding plurality of subsequence index values; computing a statistical mean value and a statistical variation value of the plurality of subsequence index values; and using said statistical mean value and said statistical variation value to compute a heart function index value which is a measure of heart function of the patient, wherein said processing comprises, for each subsequence of ECG data: extracting data corresponding with specified ECG signal segments from the subsequence of ECG data; determining a time derivative of said extracted data; and generating a subsequence index value which is representative of energy in said time derivative.

Further preferred features and advantages of the invention will be apparent to those skilled in the art from the following description of preferred embodiments of the invention, which should not be considered to be limiting of the scope of the invention as defined in the preceding statements, or in the claims appended hereto.

Comprises/comprising and grammatical variations thereof when used in this specification are to be taken to specify the presence of stated features, integers, steps or components or groups thereof, but do not preclude the presence or addition of one or more other features, integers, steps, components or groups thereof. BRIEF DESCRIPTION OF THE DRAWINGS

Preferred embodiments of the invention are described with reference to the accompanying drawings, in which like reference numerals refer to like features, and wherein: Figure 1 is a schematic diagram of an Internet-based system embodying the present invention;

Figures 2A and 2B are flowcharts illustrating a method of detecting possible abnormality of heart function of a patient according to a preferred embodiment of the invention; Figures 3A and 3B are schematic representations of ECG waveforms corresponding with typical heartbeats, illustrating methods of identifying specified points and/or intervals of the ECG signals, according to a preferred embodiment of the invention;

Figures 4A and 4B are flowcharts illustrating quality control methods according to a preferred embodiment of the invention; and

Figures 5 to 10 are screenshots illustrating features of a graphical user interface (GUI) of a software application for accessing and processing patient records according to a preferred embodiment of the invention. DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS Figure 1 illustrates schematically an Internet-based system 100 embodying the present invention. The system 100 is interconnected via the Internet 102.

The system 100 includes an electrocardiograph 104 and a user computer 106, both of which would typically be located at a doctor's office, clinic, hospital, or the like, generally being a location at which patient health may be assessed and monitored. A remote server computer 108 is accessible, to both the electrocardiograph 104 and the user computer 106, via the Internet 102. It should be noted that while the electrocardiograph 104 and user computer 106 are depicted in the system 100 as having separate connections to the Internet 102, it is equally possible that the electrocardiograph 104 may be connected to the Internet 102 via the user computer 106, or alternatively via some intermediate communications device. The general principal, however, is that ECG data recorded using the electrocardiograph 104 may be transferred, via the Internet 102, to the server computer 108. While the system 100 represents an exemplary, and particularly preferred implementation of the invention, this embodiment is not limiting of the scope of the invention. Alternative arrangements, for example in which the electrocardiograph 104 is directly connected to the computer 108, rather than being remotely accessible via the Internet 102, are equally within the scope of the invention. The particular advantage of the system 100 is that it enables diagnosis and monitoring of patients at one or more remote locations, with patient data subsequently being made available, potentially on a global basis, from the server 108 via the Internet 102. This is considered to be a particularly advantageous embodiment, and accordingly the invention is described with reference to this specific system.

The server computer 108 includes at least one processor 110, which is interfaced, or otherwise associated, with a high-capacity, non-volatile memory/storage device 112, such as one or more hard-disk drives. The storage device 112 is used to maintain a database of patient records, the contents and use of which are described below, particularly with reference to Figures 5 to 10. The storage device 112 may also contain other programs and data required for the operation of the server 108, and for the implementation and operation of various software components embodying portions of the present invention. The means by which this may be achieved are well-known in the art, and accordingly will not be discussed in detail herein.

The server computer 108 further includes an additional storage medium 114, typically being a suitable type of volatile memory, such as random access memory, for containing program instructions and transient data relating to the operation of the computer 108. Additionally, the computer 108 includes a network interface 116, accessible to the central processor 110, facilitating communications via the Internet 102 with other devices, such as electrocardiograph 104 and user computer 106.

The memory device 114 contains a body of program instructions 118 implementing various software-implemented features of the present invention, as are described in greater detail below, with reference to the remaining drawings. In general, these features include analysis and processing functions, for detecting possible abnormality of heart function of a patient, as described with reference to Figures 2 to 4. Additionally, a database server application is implemented, which enables patient records held in a database on the storage device 112 to be accessed remotely, for example from user computer 106. The database server application may include commercially available and/or open source database systems, or a proprietary database system, and may be accessed either via a dedicated client application (such as is described below with reference to Figures 5 to 10), or alternatively a Web-based interface may be provided. All such options and variations, as well as others that will be apparent to persons skilled in the art, are within the scope of the present invention. Turning now to Figure 2A, there is shown a flowchart 200 illustrating a method of detecting possible abnormality of heart function of a patient in accordance with the preferred embodiment of the invention. At step 202, an ECG measurement is performed on the patient, using the electrocardiograph 104, over a continuous time interval having a specified duration, resulting in a corresponding sequence of ECG data. In accordance with a presently preferred embodiment, the measurement is performed over a time interval of 120 seconds, using an electrocardiograph that provides digital samples of the ECG waveform at a frequency of 1 kHz. This measurement is performed in general over a full standard 12 leads, commonly derived from 10 electrodes attached in a conventional manner to corresponding points on the patient's body. That is, the output of the electrocardiograph 104, which is provided to the method at step 202, consists in all of 12 sequences of data values, each including 120,000 data points, representing 2 minutes of measurements at 1 ,000 Hz. If desired, a lesser or greater number of leads may be utilised, a different sampling rate may be employed, and/or the patient may be monitored over a shorter or longer time interval. It is presently believed, however, that a measurement interval of at least 2 minutes is desirable, in order to obtain sufficient data for subsequent processing. Longer intervals may provide additional information, such as intervals of 3 minutes, 4 minutes, or 5 minutes. Further extended intervals are also possible, however are not considered to be necessary and may be disadvantageous due to the requirement for the patient to remain connected to the electrocardiograph 104 for longer periods of time. It should be noted that the method of electrocardiograph monitoring according to the presently preferred embodiment exhibits some differences from conventional monitoring processes. For example, monitoring is typically performed over much shorter time intervals, for example 15 seconds, in order to generate ECG waveforms for review by specialist experts, such as cardiologists. Furthermore, the raw ECG data is not generally utilised in such an analysis. Indeed, many electrocardiograph units do not make raw data available for external processing, and accordingly methods such as those embodied in the present invention have generally been discouraged by manufacturers of prior art equipment. Nonetheless, electrocardiograph equipment, and add-on hardware and/or software devices enabling the extraction and processing of raw ECG data, are becoming increasingly available, thereby enabling more sophisticated and specialised processing to be performed.

Returning now to Figure 2A₁ at step 204 the ECG waveform data is segmented, by dividing the sequence of ECG data into a plurality of subsequences. According to presently preferred embodiments, the overall ECG waveform for each of the leads utilised in the measurement 202 is divided into 12 segments, each of which therefore corresponds with 10 seconds of measurement in the preferred case, or equivalently 10,000 sample values. At step 206 each subsequence of ECG values, on each of the leads utilised in the measurement, is processed to obtain corresponding subsequence index values. The method by which the subsequence index values are calculated will now be described, with reference to the flow chart shown in Figure 2B.

The ECG index value calculation corresponds closely with the mathematical methods described in detail in the commonly assigned US patent application serial no. 10/865,985, published under publication no. US

2005/0027202, which is hereby incorporated herein in its entirety by reference, in order to avoid the need to repeat details of the calculation methods and formulae.

The first step in the process 206, is to extract data corresponding with specified ECG signal segments, at step 212. The methods by which the ECG signal segments of interest are identified, such that the data may be extracted, are described in greater detail below with reference to Figures 3A and 3B. For present purposes, it is sufficient to observe that the result of the extraction is a sequence of consecutive samples of the ECG data corresponding with a particular segment or interval of the ECG waveform. It is also particularly to be noted that this extraction is performed for each "beat" that is identifiable within the ECG signal subsequences, utilising the methods described below. That is, there is no averaging performed over multiple beats, and all of the available data is utilised in the analysis.

More specifically, for each individual sequence of extracted data, corresponding with a desired segment or interval of a single beat, a time derivative is calculated, at step 214. In the preferred method of computation, a Fourier Transform 216 is then performed on the time derivative data, in order to obtain a corresponding sequence of frequency domain samples. If necessary, the time derivative data may be "extended", for example by zero-padding, in order to produce a sequence including a total number of samples that is a power of two (eg 512, 1024 etc), in order to take advantage of efficient Fast Fourier Transform algorithms. Then, at step 218, a subsequence index value is computed as a normalised sum of squares of frequency components in the transformed data sequence. As noted above, full details of the computational steps are provided in the above-referenced prior patent specification.

Returning now to Figure 2A, it should be understood that the computation step 206 is repeated for all subsequences within the full sequence of ECG data obtained from each individual ECG lead. At step 208, statistical properties of the subsequence index values are calculated, on a lead-by-lead basis. That is, for each lead a statistical mean value of the subsequence index values is calculated, and a statistical variation value of the ECG index value is also calculated. In the preferred arrangement, the statistical mean value is simply the average of the subsequence index values for the lead, and the statistical variation value is the (sample) standard deviation of the subsequence index values for the lead. Accordingly, in accordance with the presently preferred parameters, there will be, for each lead, up to 12 subsequence index values, which are averaged and their standard deviation computed.

At step 210, an overall heart function index value is computed for each lead. In particular, the heart function index value is based upon the statistical mean value and the statistical variation value previously computed for each lead, and in presently preferred embodiments consists of the ratio of the statistical variation value (ie standard deviation) divided by the statistical mean value (Ze the average). This quantity is commonly known as the "coefficient of variation". This is a quantity which has a smaller value when there is less variation amongst the subsequence index values computed in each of the subsequences of the ECG data on a particular lead. It has been discovered by the present inventors that this is a very sensitive and effective measure of possible abnormality of heart function. In particular, it has surprisingly been found that, even over an interval as short as 2 minutes, which has been divided into 12 sub-intervals of 10 seconds apiece, the above-described heart function index value is strongly indicative of potential heart dysfunction. Specifically, as the value becomes larger, the likelihood of patient heart abnormality, requiring further consideration and diagnosis by a specialist, such as a cardiologist, becomes greater. Accordingly, the invention is able to provide a fully automated early indication of patients in need of further study, diagnosis and treatment. No particular expertise in cardiology is required in order to obtain this early indication. There is accordingly great potential that the invention may be utilised to identify patients for early referral to cardiologists, leading to better patient outcomes, and potentially many lives saved. It is anticipated that further studies and trials will enable specific thresholds of the heart function index value to be identified, with which may be associated corresponding specific guidelines to practitioners for diagnosis and treatment of patients. For example, a particular threshold may be established at which it is recommended that the patient return within a specific interval of time, such as six months, for a follow-up test. At a higher threshold value, the recommendation may be that the patient be referred, at the earliest convenient time, to a cardiologist for further analysis, diagnosis and/or treatment.

As previously explained, the calculations described with reference to Figures 2A and 2B are performed for every individual lead utilised in the ECG measurement. Accordingly, a heart function index value is made available corresponding with each lead. Thus up to 12 heart function index values are provided, corresponding with the 12 standard leads utilised in typical ECG measurement. The index values may be utilised in a number of ways, as will now be described.

Firstly, each individual heart function index value may be considered separately. It has been found that, in the case of some patients, the heart function index value may be relatively low for a majority of leads, but may be relatively high for one or more individual leads. Thus, though an average of the heart function index value may be low over all leads, an abnormality of heart function relating to the measurement performed by a particular lead having a high index value may be indicated. Accordingly, the provision of individual heart function index values corresponding with each individual lead to a practitioner operating the test, or analysing the results, may be advantageous.

Secondly, a global heart function index value may be calculated, being an average of the heart function index values corresponding with each individual ECG lead. In many cases, the global heart function index value will provide a strong overall indication of patient health, however as noted above there may be some cases in which this value is relatively low despite one or more individual leads providing indications of possible abnormality. It may generally be concluded from this, however, that a relatively high value of the global heart function index value is a particularly strong indication of a potential problem. Finally, one or more cluster heart function index values may be computed, corresponding with selected groups of ECG leads. That is, additional index values may be calculated which correspond with specific lead clusters. In this regard, any meaningful or interesting cluster of leads may be selected, however there are particular clusters of leads that are established as having an association with particular regions of the heart, and the detection of corresponding abnormalities of function. Specifically, the standard leads II, III and aVF, known as the "inferior group", the group of standard leads I, aVL, v5 and v6, known as the "lateral group", and the cluster of leads v1 to v6, known as the "anterior group", are of particular interest. In each case, a corresponding "regional heart function index" may be computed as an average of the heart function index values corresponding with each ECG lead of the cluster. As with the individual leads, review of the cluster/regional heart function index values may enable abnormalities associated with corresponding regions of the heart to be identified, even if the overall (global) heart function index is relatively low.

Some preliminary studies of healthy individuals have indicated, with a high degree of statistical significance, that there may be "normal" ranges of global and/or regional heart function index values, corresponding with different analysed segments or complexes, and with male or female subjects. While further verification and quantification is required, it is presently therefore believed that heart function index values falling outside suitably determined ranges may be utilised as a strong indicator of potential abnormality of heart function of a patient. Turning now to Figures 3A and 3B, there is shown schematic representations of ECG waveforms corresponding with typical heartbeats. Methods of identifying specified points and/or intervals of the ECG signal, enabling data corresponding with specified ECG signal segments to be automatically extracted, will now be described with reference to these drawings. In particular, Figure 3A represents a schematic representation 300 of a single beat of a conventional ECG waveform. The most commonly observed features are represented, in an idealised form, within the waveform 300. As will be known to those knowledgeable in the art of ECG interpretation, the beat commences with a P-wave (labelled P), followed by a so-called QRS complex, including the Q-, R- and S-peaks (labelled accordingly) and nominally terminating at the point labelled J. The beat concludes with a T-wave (labelled T).

In Figure 3B, two adjacent beats 302, 304 are shown. Nominal demarcation lines, 306, 308 are also depicted, which represent, respectively, a time shortly following the QRS complex of beat 302, and a time shortly prior to the QRS complex of beat 204. Appropriate time values 306, 308 may be identified once the QRS complexes have been identified, in the manner shortly to be described.

The general process of feature identification employed in preferred embodiments of the invention proceeds as follows. Firstly, the ECG signal is filtered or smoothed in order to reduce noise. In a particularly preferred embodiment, a least squares cubic spline fit is performed in order to generate a smoothed ECG signal waveform. Further processing, as now described, is then conducted using the noise-reduced spline fit. Firstly, one or more time derivatives of the spline fit are computed. In the presently preferred embodiments, first and second time derivatives are computed. The Q-peak, R-peak, and S-peak are then determined, by identifying within the first and second derivatives those points having the highest values. As can be seen in the schematic diagram 300, the Q-, R- and S-peaks exhibit the greatest rate of change in all derivatives of the ECG waveform.

Subsequently, the J-point is identified, by searching the time derivative data for the point exhibiting the largest curvature following the identified S-peak. Once the Q-, R- and S-peaks have been identified, and optionally also the J-point (which may sometimes be difficult to identify with accuracy), the location of the QRS complex within the beat 300, and indeed within adjacent beats 302, 304, is known. The points 306, 308 may then be determined, and a search for the T- and P-waves appearing between these time instants is then conducted. In particular, the T-wave and P-wave are identified by points of highest curvature, corresponding with the start and/or end points of each wave, lying between the points 306 and 308.

Once the above-identified features of the waveform 300 have been identified, it is then possible to specify the location of particular signal segments, or sub-intervals, of interest. For example, the QRS complex may be identified. A PR-segment may be identified, lying between the end of the P-wave, and the start of the QRS complex. Additionally, a PR-interval may be identified, lying between the beginning of the P-wave, and the beginning of the QRS complex. A QT-interval may be identified, lying between the start of the QRS complex and the end of the T-wave. An ST-segment may be identified, lying between the end of the QRS complex (Ze the J-point) and nominally the beginning of the T-wave. In this latter regard, however, it is common for cardiologists to consider a specific time period following the J-point as corresponding with the ST-segment, since in some cases the T-wave may be hard to identify with accuracy. In particular, intervals such as 60 milliseconds, 80 milliseconds, 100 milliseconds, 120 milliseconds, and so forth, following the J-point, are utilised by various practitioners in analysis of the ST-segment. Accordingly, preferred embodiments of the present invention enable a user-specified period, such as the foregoing, to be entered for the purpose of identifying and analysing ST-segments. In a further variation, an "extended" ST-segment may be identified which further includes waveform data preceding the J-point. For example, the extended ST-segment may include the portion of the waveform 300 within a 30 ms interval prior to the J-point. This interval may include a so-called "late potential" of the QRS complex, which is known by cardiologists to be significant in the assessment of patient heart function. Other interval lengths prior to the J-point, such as 10 ms, or 20 ms, may also be utilised, and in a preferred embodiment the extended interval may be specified by the user.

While any desired sub-interval of each recorded beat 300 may be utilised in the calculation of the heart function index values, presently favoured embodiments of the invention preferably analyse the QRS complex and/or the ST-segment. Other sub-intervals of the waveform 300 are being studied, in relation to their effectiveness in producing heart function index values that are reliable and sensitive indicators of potential abnormality. It is therefore envisaged that any particularly advantageous intervals that are subsequently identified will be readily incorporated into embodiments of the invention.

A further feature of presently preferred embodiments is the use of quality control techniques, in order to exclude from the calculations any data corresponding with beats of the ECG signals that do not satisfy specified quality criteria. It should be appreciated that factors such as noise, interference, patient movement, and various other external factors, may result in some portions of the ECG waveforms being very unreliable indicators of underlying heart function. It is therefore desirable that any such portions of the waveforms be excluded from critical calculations. Two possible methods of quality control, employed in preferred embodiments of the invention, are illustrated by the flowcharts in Figures 4A and 4B. The flowchart 400 of Figure 4A is illustrative of a method based on a signal-to-noise ratio (SNR) calculation, whereas the flowchart 412 in Figure 4B is illustrative of a method based upon a check of likely quality of data on individual ECG leads.

Turning first to the flowchart 400, a method is depicted for performing quality control on individual beats. In particular, at step 402, a noise value is calculated associated with an individual beat, and a corresponding SNR value. More specifically, noise is calculated based upon the "flat" regions of the waveform, such as those intervals prior to the P-wave, or following the T-wave, as illustrated in Figure 3A. Noise is determined using the difference between the raw data and the noise-reduced spline fit, described previously in relation to Figure 3A. A signal value is estimated as a peak-to-peak amplitude, ie between the Q- and/or S-peaks, and the R-peak at the opposing extreme. The SNR value is a ratio of the aforementioned signal amplitude to the aforementioned noise measure. At step 404 a comparison is performed between the computed SNR and an acceptable SNR minimum threshold. The value of this threshold is selected based upon experience so as to reject a majority of beats that would be considered to be excessively noisy. This value may potentially be made available, in some embodiments, as a parameter that can be varied by the user. If the SNR does not exceed the minimum threshold, then the beat is rejected at step 408, and data is not extracted for inclusion in subsequent computations. However, if the SNR exceeds the minimum threshold, then the beat is accepted at step 406, and is utilised in subsequent calculations. At step 410, a check is performed to see whether all beats have been subjected to the SNR quality control, and if not then the process returns to step 402 for analysis of a subsequent beat. The flowchart 412 illustrates a method employed to assess the quality of information available on individual ECG leads. More particularly, it will be appreciated from the foregoing description that through the process of beat identification, discussed with reference to Figures 3A and 3B, and beat rejection, discussed with reference to Figure 4A, there will subsequently be a known number of identified and usable beats detected over the full measurement interval (eg 120 seconds) on each ECG lead. At step 414 this number of identified and usable beats is counted for a particular lead. At step 416, the counted value is compared with a minimum threshold. For example, it might be expected for a patient having a heart rate of 80 beats-per-minute, measured over a 2-minute interval, that approximately 160 beats should be recorded. However, if on a particular lead the actual counted number of usable beats is substantially less than this value, it is likely that the lead is particularly badly affected by noise, or other factors, and is therefore of questionable validity. Additionally, if the number of identifiable and usable beats is too low, the statistical significance of the final calculated heart function index value may be doubtful. Accordingly, if the number of counted beats is too low, then the entire lead is rejected at step 420. Otherwise the lead is accepted at step 418. At step 422, it is determined whether all leads have been checked, and if not the method returns to step 414 to perform the quality control operation on a subsequent lead. Rejected leads are excluded from calculations of a global heart function index value, and from calculations of any corresponding cluster/regional heart function index values. A user is preferably informed of the rejection of a lead, so that this may be taken into account in analysing the results, and/or appropriate corrective action taken.

The discussion now turns to a description of user interface aspects of a preferred embodiment of the invention. In particular, Figures 5 to 10 are screenshots illustrating a graphical user interface (GUI) of a software application for accessing and processing patient records. In particular, all patient data is transferred from the electrocardiograph 104 and the user computer 106 to the server computer 108, where it is stored in a database on storage 112. This database is made remotely accessible, for example via the user computer 106, by database server software executing on the computer 108, and corresponding client software executing on the user computer 106. This client/server software may be proprietary, may be based upon commercial and/or open source products, or may be Web-based. Various such implementation options will be readily apparent to persons skilled in the relevant art.

Figure 5 is a screen display 500 illustrating a window of a particular database client application, executing on the user computer 106. The window includes a patient search form 502, corresponding patient search results 504, a list of corresponding patient consultation records held within the database, shown in the region 506, and information associated with a selected consultation in region 508. It is also noted that, for each consultation a button 512 is provided enabling the user to view a corresponding stored ECG record. In preferred embodiments, the server system 108 employs a user identification and authentication system, whereby each user of the system, utilising the client application from a computer, eg 106, is required to log in using at least a user identifier (such as a user name, email address or the like), and a password. Advantageously, relatively high levels of security and authentication are provided, in order to ensure the privacy and integrity of patient records. For example, individual computers, eg 106, may be registered with the server 108 such that access from unknown and/or unauthorised computers may be denied. Individual computers may be identified using globally unique information such as network addresses (eg IP and/or Ethemet/MAC addresses), disk volume labels, and/or a "signature" derived from the specific hardware and software configuration of the computer. Additionally, all data transfers between the server 108 and a client computer 106 should be encrypted, eg using SSL protocols, or the like.

Different levels of access may be provided to data within the database according to an associated level of user authorisation. For example, a user may be a doctor associated with a particular clinic, and will thus be provided only with access to records corresponding with patients of that clinic. Alternatively, the user may be a doctor associated with particular patients, regardless of clinic, and will be provided with access to the records of those patients. In general, it sill be possible for patients to attend various clinics, doctors' offices, hospitals, and so forth, for heart rate monitoring via associated electrocardiograph equipment, eg 104, and those records will be transferred back for central storage at the server computer 108, ie in the database held on storage 112. Advantageously, this provides patients with greater mobility, while enabling doctors and other health care professionals associated with the patients to gain access to relevant records at any appropriate time and place.

Figure 6 shows a more detailed view of the patient search window 502. In particular, there is shown a dropdown menu 602, from which a user is able to load or refresh a clinic list maintained at the server computer 108. Once a clinic list has been loaded, it is possible to access patient records associated with that clinic.

This functionality is illustrated in greater detail in Figure 7. In particular, a dropdown list 702 enables a clinic to be selected. Search boxes 704, corresponding with patient first and last names, enable the user to identify corresponding details of a patient of interest. Activating the search button 706 results in a corresponding search being conducted, and the search results being displayed in the search results region 504.

From the search results pane, the user may select a specific patient, for example by double-clicking on the appropriate result, and patient record information will be loaded into the region 506. Selection of a particular consultation record will result in the consultation information being displayed in the region 508. Information displayed in this region includes patient data including name, date of birth, age, gender, recorded symptoms, body measurements and so forth. Beneath this, there is a scrolling text window in which various additional information may be displayed. This information may include automatically-generated information, such as results from ECG measurement and processing of ECG data. Additionally, consultation notes entered by the doctor at the time of consultation may also be displayed in this region. Selecting the ECG view button 512 associated with a particular consultation results in a further window opening to display all of the stored ECG data. Exemplary contents 800 of this window are depicted in Figure 8. As can be seen, the initial display includes the full set of data /e waveforms, recorded for all 12 ECG leads over the recording interval. In Figure 9, there is depicted an exemplary display 900, wherein the user has requested a 300 percent zoom, and has activated a "magnifying glass" feature 902. These features of the client application enable more detailed examination and analysis of ECG waveforms. A further function (not shown in the drawings) is the ability for the user to display only a subset of selected ECG leads. Finally, there is illustrated in Figure 10 a further display 1000 in which two separate sets of ECG data 1002, 1004 have simultaneously been open for viewing by the user. These may correspond, for example, with two different consultations of the same patient. The windows may similarly be viewed side-by-side, rather than one above the other. This feature enables a rapid comparison of differences in recorded data for the same patient at different times. While the foregoing description has covered various exemplary features of a preferred embodiment of the invention, it will be appreciated that this is not intended to be exhaustive of all possible functions provided within various embodiments of the invention. It will be understood that many variations of the present invention are possible, and the overall scope is as defined in the claims appended hereto.

Claims

CLAIMS:

1. A method of measuring heart function of a patient comprising the steps of: performing an electrocardiogram (ECG) measurement of the patient over a continuous time interval having a specified duration, to obtain a sequence of ECG data corresponding with the ECG signal over said time interval; dividing the sequence of ECG data into a plurality of subsequences; processing each subsequence of ECG data to obtain a corresponding plurality of subsequence index values; computing a statistical mean value and a statistical variation value of the plurality of subsequence index values; and using said statistical mean value and said statistical variation value to compute a heart function index value which is a measure of heart function of the patient, wherein said processing step comprises, for each subsequence of ECG data: extracting data corresponding with specified ECG signal segments from the subsequence of ECG data; determining a time derivative of said extracted data; and generating a subsequence index value which is representative of energy in said time derivative.

2. The method of claim 1 wherein the step of generating a subsequence index value comprises: determining frequency components of the time derivative; and calculating a subsequence index value representative of the energy in said frequency components.

3. The method of claim 1 wherein the specified duration of the ECG measurement is at least 120 seconds.

4. The method of claim 1 wherein the heart function index value is computed as a ratio of said statistical variation value to said statistical mean value.

5. The method of claim 1 wherein the ECG measurement is performed over a plurality of standard ECG leads, and a separate sequence of ECG data is obtained and processed for each lead, in order to compute a plurality of heart function index values, each of which corresponds with a single ECG lead.

6. The method of claim 5 wherein said plurality of ECG leads includes 12 standard ECG leads.

7. The method of claim 5 which includes computing a global heart function index value which is an average of the heart function index values corresponding with each of said plurality of standard ECG leads.

8. The method of claim 5 which includes computing one or more cluster heart function index values, each of which corresponds with a selected group of said plurality of standard ECG leads.

9. The method of claim 8 wherein at least one regional heart function index value is computed from a group of standard ECG leads selected from the set of groups comprising: leads II, III and aVF (inferior group); leads I, aVL, v5 and v6 (lateral group); and leads v1 to v6 (anterior group).

10. The method of claim 8 wherein the cluster heart function index values are computed as an average of the heart function index values corresponding with each ECG lead of the selected group.

11. The method of claim 1 wherein the step of processing each subsequence of ECG data includes automatically identifying specified points and/or intervals of individual beats of the ECG signal.

12. The method of claim 11 wherein the specified points and/or intervals are selected from the group comprising: Q-peak; R-peak; S-peak; P-wave; T-wave; J-point; QRS complex; PR-segment; ST-segment; PR-interval; and QT-segment.

13. The method of claim 1 wherein the specified ECG signal segments from which data is extracted are selected from the group comprising: QRS complexes; ST-segments; T-waves; and QT-segments.

14. The method of claim 11 wherein automatically identifying specified points and/or intervals includes the steps of: filtering or smoothing the ECG data in order to reduce noise; computing one or more time derivatives of the noise-reduced ECG data; identifying Q-peak, R-peak and S-peak as points having the highest values amongst all data values of the computed derivatives; identifying J-point as a point having the largest curvature amongst the noise-reduced ECG signals following the S-peak in each beat; and identifying P-wave and T-wave by identifying points having the highest curvature of the noise-reduced signal values between identified QRS complexes of adjacent beats.

15. The method of claim 11 further including performing quality control by excluding from further computations data corresponding with beats of the ECG signal that do not satisfy specified quality criteria.

16. The method of claim 15 including the step of computing a noise value and a signal-to-noise ratio of data within each specified signal segment, and excluding from further computations any segment for which the signal-to-noise ratio is less than a predetermined value.

17. The method of claim 15 including the step of determining the number of beats or complexes successfully identified for each ECG lead, and excluding from further computations data corresponding with any lead for which the number of successfully identified beats or complexes is less than a predetermined value.

18. A system for measuring heart function of a patient, the system including: an electrocardiograph having a plurality of ECG leads, the electrocardiograph being configured to monitor the electrical activity of the patient's heart over a continuous time interval having a specified duration, and to output at least one sequence of ECG data comprising digitised representations of the electrical activity monitored by each ECG lead; a processor configured to receive the sequence of ECG data output by the electrocardiograph, wherein said processor includes: means for dividing the sequence of ECG data into a plurality of subsequences; means for processing each subsequence of ECG data to obtain a corresponding plurality of subsequence index values; means for computing a statistical mean value and a statistical variation value of the plurality of subsequence index values; and means for using said statistical mean value and said statistical variation value to compute a heart function index value which is a measure of heart function of the patient, wherein said means for processing each subsequence of ECG data is adapted to: extract data corresponding with specified ECG signal segments from the subsequence of ECG data; determine a time derivative of said extracted data; and generate a subsequence index value which is representative of energy in said time derivative.

19. The system of claim 18 wherein the electrocardiograph is located remotely from the processor, and the ECG data are transmitted from the electrocardiograph to the processor via a data communications network.

20. An apparatus for measuring heart function of a patient, including: a central processing unit; at least one memory device operatively associated with the central processing unit; and a communications interface operatively associated with the central processing unit, the interface providing a channel for communication with an electrocardiograph having a plurality of ECG leads, and which is configured to monitor the electrical activity of the patient's heart over a continuous time interval having a specified duration, and to output at least one sequence of ECG data comprising digitised representations of the electrical activity monitored by each ECG lead; the memory device having computer-executable instructions stored therein which, when executed by the central processing unit, cause the apparatus to effect the steps of: receiving, via the communications interface, a sequence of ECG data output by the electrocardiograph; dividing the sequence of ECG data into a plurality of subsequences; processing each subsequence of ECG data to obtain a corresponding plurality of subsequence index values; computing a statistical mean value and a statistical variation value of the plurality of subsequence index values; and using said statistical mean value and said statistical variation value to compute a heart function index value which is a measure of heart function of the patient, wherein said processing comprises, for each subsequence of ECG data: extracting data corresponding with specified ECG signal segments from the subsequence of ECG data; determining a time derivative of said extracted data; and generating a subsequence index value which is representative of energy in said time derivative.