WO2009103108A1 - Diagnostic device and method for detecting an acute coronary episode - Google Patents

Abstract

A device, system and method for diagnosing the occurrence of an acute coronary episode of an acute coronary syndrome in a patient by processing data from a photoplethysmography (PPG) instrument, an electrocardiogram (ECG) signal or a correlation of data from a PPG and ECG.

Description

"Diagnostic device and method for detecting an acute coronary episode"

Cross-Reference to Related Applications

5 The present application claims priority from Australian Provisional Patent

Application TSfo 2008900823 filed on 20 February 2008, the content of which is incorporated herein by reference.

Field of the Invention 10

The present invention relates to an apparatus and method for detecting acute coronary syndromes.

Background Art

15

Ischaemic heart diseases is reported to be the leading cause of mortality in Australia. Health expenditure classified by disease or injury group is highest for cardiovascular diseases at an estimated $5.48 billion in 2000-01. According to the WHO, there were 7.1 million deaths from coronary heart disease globally m 1999. By

20 2002, this number had risen to 7.22 million.

Acute coronary syndromes (ACS) form the diagnostic and pathophysiological continuum from unstable angina to myocardial infarction. It is the goal of the emergency physician to accurately diagnose and treat a patient with potentially life- 25 threatening ACS, while avoiding the misdiagnosis and discharge of these patients. Due to the variability of clinical manifestations of ACS between unstable angina and acute myocardial infarction, the diagnosis of ACS can be challenging.

Risk stratification exists in emergency department (ED) guidelines for the 30 diagnosis of ACS in patients with chest pain. Low risk patients, even with a suspected cardiac cause for their chest pain, will often be discharged after serial electrocardiograms (ECG) and troponins, and high-risk patients will be admitted for further testing and appropriate treatment. However, it is an intermediate risk group of patients between these.which present a problem for testing and treatment. 35 Due to the misdiagnosis of non-cardiac chest pain as ACS, nearly $US600 million per year in unnecessary inpatient care costs are generated in the US, but in addition to this, 1-2% of patients sent home from the ED after presenting with chest pain actually had a myocardial infarction (Gibler WB, Blomkalns AL, Diagnosis of acute coronary syndromes in the emergency department Topol EJ, ed. Acute Coronary Syndromes. 2nd ed. New York: Marcel Dekker; 2001).

The probability of proportionately high costs as a consequence of these problems exists in countries other than the USA.

The relationship between the autonomic nervous system and cardiovascular disease is well recognised in the literature. (Task Force of the European Society of

Cardiology and the North American Society of Pacing and Electrophysiotogy. Heart

Rate Variability: Standards and Measurement, Physiological Interpretation and Clinical Use. Circulation 1996;93(5):1043-65).

The advent and availability of very fast microprocessors, together with the portability of modern computers means that frequency domain analysis (FDA) of the ECG is. readily available. This principle has been recently extended to the analysis of photoplethysmography/pulse oximetry waveforms (PPG-POW). .

The present invention aims to provide a non-invasive, reliable apparatus and method for diagnosing acute coronary syndromes.

Summary of the Invention

In a first aspect, the present invention provides an apparatus for detecting an acute coronaτy episode of an acute coronary syndrome in a patient comprising: a receiving means to receive one or more input signals from a photoplethysmography (PPG) instrument; processing means to process said one or more input signals and provide at least one output signal; output means to present said at least one output signal; characterised in that said at least one output signal is representative of whether or not said patient is experiencing said acute coronary episode. In a second aspect, the present invention provides an apparatus for detecting an acute coronary episode of an acute coronary syndrome in a patient said apparatus comprising:

. a photoplethysmography unit to receive a photoplethysmography signal from a LED/photodetector unit attached to a patient and to convert said photoplethysmography signal to an input signal; processing means to process said one or more input signals and provide at least one output signal; output means to present said at least one output signal; characterised in that said at least one output signal is representative of whether or not said patient is experiencing said acute coronary episode.

In a third aspect, the present invention provides an apparatus for detecting an acute coronary episode of an acute coronary syndrome in a patient comprising: a receiving means to receive one or more input signals from a photoplethysmography (PPG) instrument and one or more ECG input signals ; processing means to process said input signals and provide at least one output signal; output means to present said at least one output signal; characterised in that said at least one output signal is representative of whether or not said patient is experiencing said acute coronary episode.

Acute coronary syndromes is a term used to represent the diagnostic and pathophysiological continuum from unstable angina to myocardial infarction. ' ^•

The photoplethysmography (PPG) instrument of the first aspect may comprise a pulse oximetry instrument (POI). In this embodiment, the POI measures light transmission as a function of time and may present a signal indicative of the tissue blood volume. Changes in the blood volume during systole and diastole may therefore be measured. The one or more input signals may comprise measurements of the blood volume over a period of time. Typically said one or more input signals comprises a waveform signal comprising a series of systolic peaks and diastolic troughs measured over said time period.

. Either the pulse oximetry instrument- of the first aspect or the pulse oximtery unit of the second aspect comprise or may receive data from an LED/photodetector unit. The LED/photodetector unit may comprise two LEDs, a red LED and an infrared LED that alternately illuminate a peripheral blood sample with two wavelengths of electromagnetic radiation. The photodetector may convert the incident radiation to a varying electrical signal which may comprise said one or more input signals.

In another embodiment, the one or more input signals from either the pulse oximetry instrument or the pulse oximetry unit are received by said receiving means wherein said receiving means converts the one or more input signals into a modified signal suitable for further processing.

The processing means is typically a central processing unit (CPU) of a computer system. Typically, the processing means includes frequency domain analysis software.

The processing means may also include filtering means to filter the input signal and remove any spikes or trough sequences caused by artefacts.

The processing means may further comprise an interpolating means. In one embodiment, the interpolating means interpolates the systolic peak and diastolic trough sequences to evenly spaced samples in time. Methods of interpolation may include staircase-type interpolation, linear interpolation or spline interpolation.

The at least one output signal may include a number of frequency components determined by frequency domain analysis of the processed at least one input signal. The frequency components may range between very low frequency oscillations and relatively high frequency oscillations.

In one embodiment, at least one component of the output signal comprises a very low frequency (VLF) component. The frequency range of this component may be between 0.015 Hz and 0.035Hz.

A further component of the output signal may comprise a first low frequency component (LF). The frequency range of this component may be in the order of 0.035Hz to 0.08Hz. Another component of the output signal may be a second low frequency component (MF). The frequency Tange of this component may he in the order of 0.08Hz to 0.150Hz.

In another embodiment, the output signal may comprise a high frequency (HF) component typically in the range of 0.150Hz to 0.450Hz.

The HF component is considered to relate to respiration and the LF component or MF component that occurs at a lower frequency than the HF component ie less than 0.15Hz relate to the autonomic nervous system.

The output signal may be presented in a number of forms. In one embodiment, the output signal may be presented as a spectral plot. The plot may depict spectral peaks of some or all of the abovementioned frequency components. In one embodiment of the invention, the output signal may comprise a spectral peak having relatively elevated power spectral density at certain frequencies and relatively low power spectral density at other frequencies.

Ia one embodiment, the spectral plot may comprise a relatively elevated power spectral density for a low density component ie LF and/or MF. Such elevation may he indicative of an acute coronary episode including myocardial infarction.

Further, the spectral plot may include a representation of the high frequency component of the output signal that is in the range of 0.150Hz and 0.450Hz wherein the power spectral density of this component is relatively depressed.

While the apparatus may present the data as a spectral plot, in another embodiment, the processing means further processes the signal to a positive or negative format ie a signal for "yes there is an acute coronary episode" or "no, there is no acute coronary episode detected". In this embodiment, the processing means may process an elevated LF component either alone or in addition to a depressed HF component to a signal that is receivable by the output means said output means presenting said signal as a positive or negative signal.

The output means may comprise at least one indicia means including a light or a series of lights. ^' Alternatively, the output signal maybe processed to an auditory signal.

The output signal may be processed to a graded signal depending upon the severity of the coronary episode. For example, a mild episode may be presented as a light of one colour, a moderate episode presented as a light of a different colour and a severe episode presented as a light of a still different colour. Similarly, in the embodiment wherein the output signal is processed to an auditory signal, the severity of the episode may be presented by different types or volume of auditory signal.

The latter embodiment may be useful in determining between various syndromes which make up the spectrum of acute coronary syndromes. For example, unstable angina may be detected by the apparatus of the present invention which present a suitable indicator of said syndrome. Similarly, the apparatus may detect a more severe episode such as a myocardial infarction and provide an indicator to this effect.

In a further aspect, the present invention provides a diagnostic apparatus for detecting an acute coronary episode of an acute coronary syndrome in a patient comprising: a receiving means to receive one or more input signals from an ECG; processing means to process said one or more input signals and provide at least one output signal; output means to present said at least one output signal; characterised in that said at least one output signal is representative of heart rate variability (HRV) of the patient and wherein said output signal is indicative of whether or not said patient is experiencing said acute coronary episode.

In a further aspect of the invention there is provide a system for diagnosing an acute coronary episode of an acute coronary syndrome, said system including

(i) obtaining at least one input signals from a photoplethysmography (PPG) instrument;

(ii) processing said at least one input signals and providing at least one output signal; and (iii) wherein said at least one output signal is representative of whether or not said patient is experiencing said acute coronary episode. In a still further aspect, the present invention provides a system for diagnosing an acute coronary episode of an acute coronary syndrome in a patient said system including: receiving one or more input signals from an ECG; processing said one or more input signals and providing at least one output signal; characterised in that said at least one output signal is representative of heart rate variability (HRV) of the patient and wherein said output signal is indicative of whether or not said patient is experiencing said acute coronary episode -

Brief Description of the Drawings

Figure 1 is a schematic representation of one embodiment of an apparatus of the present invention;

Figure 2 is a schematic representation of a further embodiment of the present invention;

Figure 3 is a schematic spectral plot representing an output signal indicative of an acute coronary episode;

Figures 4 and 5 show the detection of RR from ECG traces;

Figures 6 provides sample traces of finger PPG waveforms; Figure 7 provides sample traces of ear PPG waveforms;

Figure 8a shows a box and whisker plot and corresponding ROC curve depicting the difference in mean heart rate between negative and positive Troponin 1 ;

Figure 8b shows a box and whisker plot and corresponding ROC curve depicting the difference in mean heart rate between negative and positive Troponin 2 measurements: Figure 9 details normalised LF in Troponin 1 positive and negative and Troponin 2 positive and negative ECG and PPG (ear and finger) analyses;

Figure 10a^" depicts box and whisker plot and associated ROC curve to show differences in ear PPG MF% between negative and positive Troponin 1;

„ Figure 10b depicts box and whisker plot and associated ROC curve to show differences in ear PPG MF% between negative and positive Troponin 2;

Figure 11 depicts box and whisker plot and associated ROC curve to show differences in ear PPG MF/HF between negative and positive Troponin 1 ;

Figure 12 depicts box and whisker plot and associated ROC curve to show differences in ear PPG MF/HF between negative and positive Troponin 2;

Figure 13a depicts box and whisker plot and associated ROC curve to show the differences in finger (F) PPG LF% between negative and positive Troponin 1 ;

Figure 13b depicts box and whisker plot and associated ROC curve to show the differences in finger (F) PPG HF% between negative and positive Troponin 1;

Figure 14a depicts box and whisker plot and associated ROC curve to show the differences in finger (F) PPG MF/HF% between negative and positive Troponin 1;

Figure 14b depicts box and whisker plot and associated ROC curve to show the differences in finger (F) PPG MF/HF% between negative and positive Troponin 2;

Figure 15 depicts differences in cross correlation linking RR + E PPG PK, VOL and TR in the LF region between negative and positive Troponin 1 and Troponin 2;

Figure 16 depicts ROC curves indicating power of cross correlation linking RR + E PPG PK, VOL and TR in the LF region to discriminate between negative and positive Troponin 1; Figure 17 depicts the differences in cross correlation linking RR + E PPG PK, VOL and TR in the MF region between negative and positive Troponin 1 and Troponin

2;

Figure 18 depicts ROC curves indicating power of cross correlation linking RR

+ E PPG MF% PK, VOL and TR in the MF region to discriminate between negative and positive Troponin 1;

Figure 19a shows output from ECG analysis both in the time domain and associated frequency spectra in a patient with a negative troponin test on admission and subsequent negative testing for myocardial infarction;

Figure 19b shows output from ECG analysis both in the time domain and associated frequency spectra in a patient with a positive troponin test on admission and subsequent positive testing for myocardial infarction;

Figure 20a shows output from PPG analysis both in the time domain and associated frequency spectra in a patient with a negative troponin test on admission and subsequent negative testing for myocardial infarction; and

Figure 20b shows output from PPG analysis variability in the time domain and associated frequency spectra in a patient with a positive troponin test on admission and subsequent positive testing for myocardial infarction.

Detailed Description of the Drawings

The apparatus 10 of the present invention was developed as a result of research into the relationship of data from pulse oximetry and ECG and acute coronary episodes. The study involved monitoring both pulse oximtery data (processed in the frequency domain and presented as a signal having various frequency components) and heart rate variability taken from ECG data when a subject was admitted to hospital with chest pain. The data was then compared to troponin levels in the patients studied. Troponin is a key biomarker of cardiac injury.

Study Study setting

Eligible for the study were all adult patients presenting with complaints of chest pain at the Emergency Department (ED) of the Prince of Wales Hospital (POWH)₅ an academic hospital in Sydney, Australia over a 4 month period. The POWH emergency department treats approximately 40,000 patients per annum. There were no age or sex exclusions. Patients were selected by convenience sampling.

Study Protocol

Patients identified at triage with chest pain were allocated a study number to de- identify data. All patients received routine clinical care, including clinical history and examination. A data sheet was completed for each patient with details including:

• Age

• Gender • Date and day of arrival

• Mode of arrival

• Time of arrival

• Triage category

• Time since onset • Pain scale (0-10) at onset and on presentation

• Pain character and location (including radiation)

• Other symptoms

• Cardiac risk factors

• Medications • Medical history

• Treatment received in ED or pre-arrival

• Vital signs (including blood pressure, temperature, heart and respiratory rates, oxygen saturation, and Glasgow Coma Scale)

• The results of other investigations including serial troponin I₅ D dimer, ECG, and chest x-ray were also noted.

Troponin I results come from venous blood samples taken from patients on presentation and a later collection for the 8 hour Troponin. Troponin 1 in the present results refers to the initial Troponin and Troponin 2 refers to the 8 hour Troponin. An elevated Troponin result was considered to be above O.lng/mL. Each patient was connected to the Powerlab 16/30 using three standard ECG electrodes on the chest, and pulse oximeter probes on an earlobe and fingertip, with data collected via a BioAmp and saved onto a laptop computer. Each recording lasted approximately ten minutes, with a minimum duration of five minutes. Patients were encouraged not to move during the recording to minimise movement artefact. Sampling of heart rate and pulse oximetry waveform took place at 200Hz.

Data Analysis

Data were entered in an Excel (Microsoft Corp., Redmond, WA) database for analysis. Data analysis was performed using Analyse-It (Analyse-It Software Ltd, Leeds, UK.) Data are described using descriptive statistics, 95% confidence intervals and p values. P-values <0.05 were considered significant. Data are presented as median + interquartile range as the distributions were not symmetrical, and Wilkinson's Signed Rank and Mann Whitney U tests, Kruskal-Wallis ANOVA and Receiver Operating Characteristic curve analysis are used where appropriate. Calculations were made of sensitivity, specificity, positive and negative predictive values using contingency tables.

Waveform Analysis

All signal processing and feature extraction was implemented in Matlab (Natick, MA, USA). Zero-phase lowpass filtering (18Hz cutoff) of the PPG signal was performed prior to feature detection. The systolic peak (PK) and diastolic trough (TR) were identified from the PPG signal, and pulse volume amplitude (VOL) corresponds to the difference between peak and trough values after trend removal (by subtraction of a 2 s moving average). The R-wave peaks were detected from the ECG signal using a set of programming routines involving lowpass filtering, differentiation, and threshold- based peak detection, and the R-R interval (RR) and HR were computed accordingly. Poor quality beats (e.g. those affected by motion artifacts) and abnormal beats (e.g. ectopics) were removed and replaced by linear interpolation from the preceding to the following beats. The detection of RR from ECG is illustrated in Figures 4 and 5, and sample traces of finger and ear PPG waveforms are shown in Figures 6 and 7 respectively. The beat-to-beat sequences of RR and finger PPG waveform features in the four minute segments were converted into variability signals by interpolating to evenly spaced, samples in time then downsampling to 2 Hz after appropriate lowpass filtering. The very low frequency trend (75 s moving average smoothed by a Hanning window) was subtracted from the downsampled signal. The power spectra of RR and PPG features were computed by a 2048-pt Fast Fourier Transform (FFT) of the windowed autocorrelation of RR, based on the Blackman Tukey method. The cross-power spectra of RR and PPG features were obtained by a 2048-pt FFT of the windowed cross- correlation between the peak and the pulse volume variabilities, also based on the Blackman Tukey method. The coherence-weighted cross-power spectrum was computed by the product of the cross-power spectrum and the coherence function, which aimed at emphasizing the highly correlated frequency components that were believed to represent common physiological mechanisms behind the two variability signals (for example, sympathetic modulation and respiratory fluctuation).

The frequency bands used as definitions in tins study were:

• VLF - 0.015 -0.035Hz

• LF = 0.035 - 0.085Hz

• MF = 0.085 -0.150Hz • HF = 0.150 -0.450Hz

In this study the diagnostic utility of ECG RR and PPG variability was addressed by specifically investigating LF power, MF power, HF power and MF/HF ratio. Traditionally the LF/HF ratio is suggested as a measure of autonomic balance however as the usual LF band was split into an LF and a MF band, both of which may be seen as components of the usual band; the MF region accorded more precisely with that used as the LF definition in previous studies in this study. Furthermore, to deal with the large amount of inter-individual variation in total PPG amplitudes, LF, MF and HF were normalised by dividing them by the total power in the spectrum and expressed as a percentage. Thus normalised LF is quoted as LF% and normalised MF is quoted as MF%.

Examples of Analysis.

Cross correlation and coherence weighting

The power spectra of HRV and PPG features and the coherence-weighted cross- power spectra were divided into a LF band (0.04-0.15 Hz) and a HF band (0.15-0.45

Hz). For HRV, the LF band is influenced by both sympathetic, and vagal nerve activities, and the HF band reflects vagal modulation to some extent. In contrast, the LF band of PPG waveform reflects mainly sympathetic influences on peripheral vessels, whereas the HF band represents predominantly the mechanical effect of respiration. Part of the LF band was defined as the mid frequency (MF) band (0.08-0.15 Hz), which has been regarded as a more specific representation of sympathetic vascular modulation by autonomic mechanisms ,

The power in each band was calculated by integration of the power spectrum over the specified frequency range. The powers in the LF and MF bands were expressed in normalized units (nu) after division by the total power in 0.04-0.45 Hz (excluding the very low frequency (VLF) band at < 0.04 Hz) then multiplied by 100, and denoted as LF% and MF% for spectral analysis and LFnu and MFnu for cross- spectral analysis. Outcomes

Comparator outcomes for study were defined by biochemical testing. They were defined as -

• Measured troponin I >O.lμg L^"1 on admission to emergency department - referred to as troponin 1

• Measured troponin I X).lμg L^"1 at 8 hours from the onset of symptoms - referred to as troponin 2

Statistical Analysis

All the information was initially analysed in terms of morphological classification using Matlab^® 7.7 (The MathWorks Inc., Natick, MA₅ USA), and the data sorting and initial statistical analysis was performed in Excel. This was followed by subsequent statistical testing using Analyse-It (^Analyse-It Software Ltd, Leeds, UK), SPSS 9.0 (^®SPSS Inc., Chicago, IL, USA) and SAS (SAS Institute Inc., Gary, NC, USA).

Data is described using descriptive statistics, 95% confidence intervals and p values. P-values < 0.05 were considered significant. Data is presented as median + interquartile range as the distributions were not symmetrical, and Wilkinson's Signed Rank and Mann Whitney U tests, Kruskal-Wallis ANOVA and Receiver Operating Characteristic curve analysis are used where appropriate. Calculations were made of sensitivity, specificity, positive and negative predictive values using contingency tables. The classification of area under the ROC curve (AUROC) in this study was that given by The University of Nebraska detailed below:

• 0.90 - 1.00 = excellent • 0.80 - 0.90 = good

• 0.70 - 0.80 = fair

• 0.60 - 0.70 = poor

• 0.50 - 0.60 - fail The sample comprised 41 males and 27 females. The age of the patients ranged from 20 to 90 years old with a mean of 57.4 years (SD= 18.5). In terms of caτdiao risk factors, 35% of patients were over the age of 65 years (n= 24). 53% of patients smoked

(n= 36), and another 53% had a past or current medical history of hypertension (n= 36).

41% of patients had a past or current history of hypercholesterolemia (n= 28), 46% of patients were identified as having a BMI >25 (n= 31), 34% of patients had a family history of ischaemic heart disease (n= 23), and 18% of patients have a medical history of diabetes mellitus (n= 12).

Due to technical artefacts three ECG traces, 26 finger PPG waveforms, and 1O- ear PPG waveforms were unable to be used. This left 65 ECG traces, 42 finger PPG waveforms, and 58 ear PPG waveforms for analysis. Most of the artefacts involved noisy signals and/or movement artefacts.

Calculated components within each of the ECG and PPG (ear and finger) traces were compared against each other. The Kruskal-Wallis ANOVA test was used to look for differences between similar components in multiple independent cardiovascular variables. There were significant differences in LF (p= 0.0101), MF (p= 0.0414), HF (p= 0.0004), LF% (p= <0.0001), HF% (p= <0.0001), LF/HF (p= O.0001). There was a trend towards significance in MF% (p= 0.0550).

. Abbreviation Meaning

Mean HR mean heart rate

SDNN standard deviation of signal

RMSSD square root of the mean squared differences of successive intervals

PSD power spectral density

VLF PSD of very low frequency component (0.015-0.035Hz)

LF PSD of low frequency component (0.035-0.15Hz)

MF PSD of mfd frequency component (0.08-0.15Hz)

HF PSD of high frequency component (0.150-0.450Hz) LF%, MF%, HF% PSD in normalised units expressed as percentage

LFnu Low frequency in normalised units for cross correlation

MFnu Mid frequency in normalised units for cross correlation

Spectral Component (%) (Spectral Component/(LF + HF)XlOO

RAHO LF/(LF+HF)

The components of the two PPG modalities (finger and ear) were compared against each other using the Mann-Whitney U test. There were significantly different results between these two modalities in LF (p= 0.0058), LF% (p= <0.0001), and MF% (p= 0.0284). A trend towards significance was shown in MF (p= 0.0589).

The results of clinical investigations for each patient were divided into normal and abnormal, and statistically compared to each time and frequency domain component. These investigations included serial troponin I measurement, ECG, echocardiography, and angiography. Significant relationships were found between several time and frequency domain components of ECG and PPG and elevated troponin results. There were no statistically significant relationships between variability components and other investigations.

From ECG analysis, significant relationships were found between mean HR vs troponin 2 (p= 0.0385), EMSSD vs troponin 2 (p= 0.0227), LF vs troponin 2 (p- 0.0297)(not shown in Table 1), HF vs troponin 2 (ρ= 0.0102), and MF% vs troponin 1 (p = 0.0170) as shown in Table 1.

Receiver Operator Characteristic (ROC) curves were plotted for these significant results and showed good discrimination for mean HR vs troponin 2 (AUC= 0.817)- (see Figure 8b for graphical depiction); RMSSD vs troponin 2 (AUC= 0.849), HF vs troponin 2 (AUC= 0.856), MF% vs troponin 1 (AUC= 0.859). The following tables show these comparisons along with cut-off, sensitivity, specificity, positive and negative predictive values, medians, interquartile range (IQR), and 95% confidence intervals (CI).

Figure 9 depicts the differences in ECG LF between negative and positive troponin I.

Table 1: Significant results and good discrimination in ROC curve from ECG traces (AUC: area under ROC curve; PPV: positive predictive value; NPV: negative predictive value; bpm: beats per minute; ms: milliseconds).

Table 2: Comparison of medians, IQR, and 95% CI between negative and positive Troponin results for significant FDA components from ECG traces (bpm: beats per minute; ms: -milliseconds;; -ve: negative; +ve: positive; IQR: interquartile range; CI: confidence interval).

PPG Data

From the PPG data, spectral analyses were performed separately for peak and trough values.

For peak Values in ear (E) PPG, significant results were achieved for MF vs troponin 1 (p= 0.0381), MF vs troponin 2 (p= 0.0202), MF% vs troponin 1 (P=³0.0148), and MF% vs troponin 2 (p= 0.0179), MF% between negative and positive troponin 1 and troponin 2.

Differences in E PPG MF% between negative and positive troponin 1 are depicted in Figure 10a and for troponin 2 in Figure 10b.

Differences in E MF/HF between negative and positive troponin 1 and troponin 2 are shown in Figures 11 and 12 respectively.

ROC curves showed good discrimination for MF vs troponin 2 (AUC= 0.802),

MF% vs troponin 1 (AUC= 0.833), and MF% vs troponin 2 (AUC= 0.807) - see ROC curves shown in Figures 10a and 10b. ROC curves indicating power of E PPG MF% to discriminate between negative and positive troponin 1; and differences in MF/HF between negative and positive troponin 1 and 2 are depicted in Figure 12.

The following tables show these comparisons along with cut-off, sensitivity, specificity, positive and negative predictive values, medians, interquartile range (IQR), and 95% confidence intervals (CI).

Table 3^': Significant results and good discrimination in. ROC curve from peak values in ear PPG traces (AUC: area under ROC curve; PPV: positive predictive value; NPV: negative predictive value; mV: millivolts).

Table 4: Comparison of medians between negative and positive Troponin results for significant FDA components of peak values in ear PPG (mV: millivolts; -ve: negative; +ve: positive; IQR: interquartile range; CI: confidence interval).

Finger PPG Results

For peak values in finger PPG, significant results were achieved for LF vs troponin 1 (p= 0.0215), LF% vs troponin 1 (p= 0.0495), LF% vs troponin 2 (p= 0.0366), HF% vs troponin 1 (ρ= 0.0495), HF% vs troponin 2 (p= 0.0366), and LF/HF ys troponin 2 (p= 0.0398). There was a trend towards significance for LF/HF vs troponin 1 (p= 0.0553). Differences in F PPG LF and HF% between negative and positive troponin 1; and differences in MF / HF between negative and positive troponin 1 and 2 are shown in box and whisker plots and ROC curves in Figures 13a through to 14b.

ROC curves showed good discrimination for LF vs troponin 1 (AUC= 0.854),

LF% vs troponin 1 (AUC= 0.802), LF nu vs troponin 2 (AUC= 0.830), HF% vs troponin 1 (AUC= 0.802), HF% vs troponin 2 (AUC= 0.830), and LF/HF vs troponin 2 (AUC= 0.825). Results from trough values in finger PPG were the same as these for peak values. The following tables show these comparisons along with cut-off, sensitivity, specificity, positive and negative predictive values, medians, interquartile range (IQR), and 95% confidence intervals (CI).

Table 5: Significant results and good discrimination in ROC curve from peak values in finger PPG traces (AUC: area under ROC curve; PPV: positive predictive value; NPV: negative predictive value; mV: millivolts).

Table 6: Comparison of medians, IQR, and 95% CI between negative and positive Troponin results for significant FDA components of peak values in finger PPG (mV: millivolts; -ve: negative; +ve: positive; IQR: interquartile range; CI: confidence interval).

Discussion of PPG results

Finger Although the MF% components of F PPG did not show statistically significant differences between negative and positive tests, LF% showed a significant increase of 6% (95%CI 0 to 14%, p=0.05) in patients with a positive troponin 1 compared with those patients with a negative test.

LF% also showed a significant increase of 7% (95%CI 1 to 14%, p=0.04) in patients with a positive troponin 2 compared with those patients with a negative test.

HF% showed a statistically significant decrease in spectral power of 7% (95%CI ^• 0 to 14%, p=0.05)m patients with a positive troponin 1 compared to those patients with a negative test. Similarly again, HF% showed a statistically significant decrease in spectral power of 7% (95%CI 1 to 14%, p=0.04) in patients with a positive troponin 2 compared to those patients with a negative test.

The ratio of MF / HF spectral powers showed a significant increase of 1.2

(95%CI 0.1 to 3.0, p=0.03) in patients with a positive troponin 1 compared to those patients with a negative test.

MF / HF similarly showed a significant increase of 1.10 (95%CI 0.00 to 3.00, p=0.05) in patients with a positive troponin 2 compared to those patients with a negative test.

ROC curve analysis showed that LF% was a good discriminator between positive and negative troponin 1 (AUROC 0.80, 95%CI 0.55 to 1.00), and between positive and negative troponin 2 (AUROC 0.83, 95%CI 0.57 to 1.00).

ROC curve analysis also showed that HF% was a good discriminator between positive and negative troponin 1 (AUROC 0.80, 95%CI 0.55 to 1.00), and between positive and negative troponin 2 (AUROC 0.83, 95%CI 0.57 to 1.00).

Finally, ROC curve analysis showed that MF / HF was a good discriminator between positive and negative troponin 1 (AUROC 0.83, 95%CI 0.63 to 1.00), and between positive and negative troponin 2 (AUROC 0.81, 95%CI 0.60 to 1.00).

No other F PPG measure showed significantly different results between tests.

Bar

MF% showed a statistically significant increase in spectral power of 16% (95%CI 3 to 30%, p=0.02) in patients with a positive troponin 1 compared to those patients with a negative test.

MF% also showed a statistically significant increase in spectral power of 14% (95%CI 4 to 28%, p=0.02) in patients with a positive troponin 2 compared to those patients with a negative test. MF / HF showed a statistically significant increase in spectral power of 1.90 (95%CI 0.80 to 3.13, p=0.003) in patients with a positive troponin 1 compared to those patients with a negative test.

5 MF / HF similarly showed a statistically significant increase in spectral power of

1.67 (95%CI 0.18 to 2.92, p=0.03) in patients with a positive troponin 2 compared to those patients with a negative test.

ROC curve analysis showed that MF% was a good discriminator between 0 positive and negative troponin 1 (AUROC 0.83, 95%CI 0.67 to 1.00), and also between positive and negative troponin 2 (AUROC 0.81, 95%CI 0.60 to 1.00).

ROC curve analysis showed that MF / HF was an excellent discriminator between positive and negative troponin 1 (AUROC 0.90, 95%CI 0.82 to 0.99), and was 15 a fair discriminator between positive and negative troponin 2 (AUROC 0.78, 95%CI 0.49 to 1.00).

Cross correlation between HRV and PPG - Low Frequency 0 Figure 15 depicts differences in cross correlation linking RR + E PPG PK, VOL and TR in the LF region between negative and positive troponin 1 and troponin 2.

Figure 16 depicts ROC curves indicating power of cross correlation Unking RR + E PPG PK, VOL and TR in the LF region to discriminate between negative and 5 positive troponin 1.

Discussion of Cross Correlation - Low Frequency

There were statistically significant differences in the cross correlation linking 30 RR + E PPG PK in the LF region, with an increase of 16 LFnu (95%CI 2 to 40 LFnu, p=0.02) in the troponin 1 positive group over the troponin 1 negative group.

There were also statistically significant differences in the cross correlation linking RR + E PPG VOL in the LF region, with an increase of 24 LFnu (95%CI 2 to 35 45 LFnu, p=0.02) in the troponin 1 positive group over the troponin 1 negative group. Finally, there were differences in the cross correlation linking RR + E PPG TR in the LF region, which did not reach statistical significance, with an increase of 9 LFnu (95%CI -1 to 30LFnu, p=0.09) in the troponin 1 positive group over the troponin 1 negative group.

5 ^{' ■} .

ROC curve analysis showed that the cross correlation linking RR + E PPG PK in the LF region was a good discriminator between positive and negative troponin 1

(AUROC 0.84, 95%CI 0.69 to 1.00), as was the cross correlation linking RR + E PPG

. VOL in the LF τegion (AUROC 0.84, 95%CI 0.67 to 1.00). The cross correlation_Q linking RR + E PPG TR in the LF region was a fair discriminator with an AUROC of

.0.75 (95%CI 0.57 to 0.93),

Cross correlation between HRV and PPG - Mid Frequency 5 . Figure 17 depicts the differences in cross correlation linking RR + E PPG PK,

VOL and TR in the MF region between negative and positive troponin 1 and troponin 2.

Figure 18 depicts ROC curves indicating power of cross correlation linking RRO ⁺ E PPG MF% PK, VOL and TR in the MF region to discriminate between negative and positive troponin 1

. Discussion of Correlation - Mid Frequency 5 There were statistically significant differences in the cross correlation linking

RR + E PPG PK in the MF region, with an increase of 22 MFnu (95%CI 8 to 33 MFnu, p=Q.OO4) in the troponin positive group over the troponin negative group.

There were statistically significant differences in the cross correlation linking0 RR + E PPG VOL in the MF region, with an increase of 21 MFnu (95%CI 8 to 32 MFnu, p=0.007) in the troponin positive group over the troponin negative group.

There were statistically significant differences in the cross correlation linking RR + E PPG TR in the MF region, with an increase of 20 MFnu (95%CI 8 to 33 MFnu,5 p=0.007) in the troponin positive group over the troponin negative group. ROC curve analysis showed that the cross correlation linking RR + E PPG PK in the MF region was an excellent discriminator between positive and negative troponin 1 (AUROC 0.93, 95%CI 0.91 to 1.0O)₅ as were the cross correlations linking RR + E 5 PPG VOL in the MF region (AUROC 0.91, 95%CI 0.81 to 1.0O)₇ and the cross correlation linking RR + E PPG TR in the MF region (AUROC 0.91, 95%CI 0.82 to 1.00).

Summary of results 0

In the ECG analysis, heart rate was increased in patients with both elevated first and second troponin measurements, however this was only statistically significant with troponin 2 as the 95% confidence intervals for the difference in the troponin 1 group included zero. Time domain measures were also statistically different between troponin5 negative and positive groups, with the square root of the mean squared differences of successive RR intervals (RMSSD) lower in the troponin 2 positive group. MF% was significantly higher in troponin 1 positive patients and had good discrimination between the groups. 0 In F PPG analysis, there were no significant differences in MF% components of the total LF band between troponin positive and negative groups, however there was a statistically significant increase in LF for positive compared with negative patients for both troponin 1 and troponin 2. F PPG HF% was significantly decreased in troponin 1 and troponin 2 positive patients, and the ratio of MF/HF spectral powers also showed5 statistically significant increases in patients positive for both troponin 1 and troponin 2. ROC curve analysis revealed that all these tests were classed as having good discriminatory power to identify troponin and troponin 2 positive patients.

E PPG analysis showed dissimilarities to F PPG analysis. E PPG MF% wasø significantly increased in both troponin 1 and troponin 2 patients, compared with test negative patients. This was also true of the MF / HF ratio which also was significantly increased for both troponin tests. ROC curve analysis described E PPG MF% as a good , discriminator between positive and negative tests, and MF / HF ratio as an excellent discriminator for troponin 1 and a fair discriminator for troponin 2. 5 The coherence-weighted cross-correlation was computed at both low frequency (LF) and mid-frequency (MF) between the variability in the RR interval and the variability in E PPG peak (PK)₅ pulse amplitude (VOL) and baseline (or trough) (TR) variabilities. There were statistically significant increases in the LF cross-correlations between RR and E PPG PK and VOL, but a non-significant increase in the RR - E PPG TR cross-correlation. ROC curve analysis showed the first two cross-correlations to be good discriminators and the RR - E PPG TR cross-correlation to be a fair discriminator. MF cross-correlations showed superior results — there were statistically very significant increases in RR - E PPG PK₅ RR - E PPG VOL and RR - PPG TR in the troponin 1 positive groups, and ROC curve analysis gave values of 0.93, 0.91 and 0.91 respectively, all defined as having excellent discriminatory power to identify patients with a positive test.

Variable Elevated troponin 1 >0.1 μg L"¹ Elevated troponin 2 >0.1 μg L^'1

Mean HR -

RMSSD -

ECG MF% * -

F PPG LF_(ot ^■t- -t

F PPG HF%

F PPG MF / HF

E PPG MF%

E PPG MF / HF ♦

RR-E PPG PK LF •*• -

RR-E PPG VOL LF -

RR-E PPG TR LF -

RR-E PPG PK MF -

RR-E PPG VOL MF -

RR-E PPG TR MF -

Table 7- summary of statistically significant differences between test positive and test negative patients These results show that several frequency domain analysis (FDA) components are useful in discriminating between acute myocardial ischaemia or infarction and other causes of chest pain. The strong significant relationships are seen between the FDA components and initial and 8-hour Troponins, but not with other investigations such as echocardiograph or angiogram.

Both mean heart rate and time domain measures based on standard deviations are statistically different in chest pain patients with a first positive troponin test. It was also demonstrated that frequency domain analysis of heart rate variability and pulse oximeter waveform variability can give important information to discriminate between troponin positive and troponin negative patients, or those with evidence of probable myocardial necrosis and those without. It was demonstrated that HRV frequency domain measures were good discriminators between troponin positive and negative patients, as were spectral measures of finger PPG₃ however it was also demonstrated that some measures derived from the ear PPG had excellent discriminatory power to identify troponin positive patients. Finally, the results show that a cross-correlation between central and peripheral cardiovascular variabilities had improved power to discriminate troponin positive patients over all other measures.

Figure 19a shows output from ECG analysis both in the time domain and associated frequency spectra in a patient with a negative troponin test on admission and subsequent negative testing for myocardial infarction. Low amplitude spectral peaks are seen in the LF region and high amplitude peaks are seen in the HF region.

Figure 19b shows output from ECG analysis both in the time domain and associated frequency spectra in a patient with a positive troponin test on admission and subsequent positive testing for myocardial infaτction. In this patient higher amplitude spectral peaks are seen in the LF region and peaks are seen in the HF region are attenuated or missing.

Figure 20a both in the time domain and associated frequency spectra in a patient with a negative troponin test on admission and subsequent negative testing foτ myocardial infarction. Low amplitude spectral peaks are seen in the LF region and high amplitude peaks are seen in the HF τegion. Figure 20b shows output from PPG analysis variability in the time domain and associated frequency spectra in a patient with a positive troponin test on admission and subsequent positive testing for myocardial infarction. Ia this patient higher amplitude spectral peaks are seen in the LF region and peaks are seen in the HF region are attenuated or missing

The Tesults provide a technique to assess interaction between HRV and PPG variability. The correlation between simultaneous changes in the frequency domain was computed, correlating LF and MF frequency components in HRV with frequency components in the peak, waveform area and baseline PPG variability. This was reinforced by the addition of a coherence-weighted technique which emphasised the frequencies at which there was correlation to allow the most efficient identification of synchronised change. This technique gave results that showed excellent discrimination between patients with positive and negative troponin tests on admission, with a positive RR - E PPG PK correlation suggesting that 93% of patients would have a raised troponin.

Based on the clear trends found between frequency domain measures of autonomic variability and elevated troponin levels, the present invention provides a diagnostic apparatus 10 to receive and process data from either a pulse oximeter unit or an ECG unit.. The processed data is presented in a form which is representative of whether a patient has suffered or is suffering an acute coronary episode.

In one form, the apparatus 10 comprises a receiving means 11 to receive one or more input signals 12 from a photoplethysmography (PPG) instrument 13. A processing means 14 to process said one or more input signals 12 and provide at least one output signal 15 is also provided. An output means 16 presents the output signal

15.

As depicted in Figure 2, the device 10 may include an in-built photoplethysmography unit 13 a.

The photoplethysmography (PPG) instrument 13 or unit 13a comprises a pulse oximetry. photodetector 13b. The detector 13b receives data from an LED unit 17. The LED unit comprises two LEDs, a red LED 18a and an infrared LED 18b that alternately illuminate a peripheral blood sample with two . wavelengths of electromagnetic radiation. The photodetector 13a converts the incident radiation to a varying electrical signal comprising the input signals 12.

The input signals are received by the receiving means 11 which processes the signal into a modified signal 19 suitable for further processing.

The processing means 14 is a central processing unit (CPU) of a computer system. The signal is processed to provide a processed signal 20. The processed signal is then converted by output means 16 to an output signal 15.

The output signal 15 may comprise a spectrum of frequency oscillations determined by frequency domain analysis of the processed input signal. The frequency oscillations may range between very low frequency oscillations and relatively high frequency oscillations. The spectral plot depicted in Figure 3 is indicative of an acute coronary episode including a myocardial infarction (MI). The frequency peak in the LF component is elevated while the HF component is depressed.

While the data may be displayed in this manner for interpretation by a physician or other healthcare worker, the signal may also be converted into a positive or negative signal, for example, a light or sound to indicate the presence of a coronary episode.

It will be appreciated by persons skilled in the art that numerous variations and/or modifications may be made to the invention as shown in the specific embodiments without departing from the spirit or scope of the invention as broadly described. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive.

Claims

CLAMS:

1. An apparatus for detecting an acute coronary episode of an acute coronary syndrome in a patient comprising: a receiving means to receive one or more input signals from a pliotoplethysmography (PPG) instrument; processing means to process said one or more input signals and provide at least one output signal; output means to present said at least one output signal; characterised in that said at least one output signal is representative of whether or not said patient is experiencing said acute coronary episode.

2. An apparatus for detecting an acute coronary episode of an acute coronary syndrome in a patient said apparatus comprising: a photoplethysmography unit to receive a photoplethysmography signal from a

LED/photodetector unit attached to a patient and to convert said photoplethysmography signal to an input signal; processing means to process said one or more input signals and provide at least one output signal; output means to present said at least one output signal; characterised in that said at least one output signal is representative of whether or not said patient is experiencing said acute coronary episode.

3. An apparatus for detecting an acute coronary episode of an acute coronary syndrome in a patient comprising: a receiving means to receive one or more input signals from a photoplethysmography (PPG) instrument and one or more BCG input signals ; processing means to process said input signals and provide at least one output signal; output means to present said at least one output signal; characterised in that said at least one output signal is representative of whether or not said patient is experiencing said acute coronary episode.

4. The apparatus of any one of claims 1 to 3 wherein said acute coronary syndromes comprises any presentation in the diagnostic and pathophysiological continuum, from unstable angina to myocardial infarction.

5. The apparatus of any one of the preceding claims wherein the photoplethysmography (PPG) instrument comprises a pulse oximetry instrument (POI).

5 6. The apparatus of any one of the preceding claims wherein said receiving means converts the one or more input signals into a modified signal suitable for further processing.

7. The apparatus of any one of the preceding claims wherein the processing means 10 is a central processing unit (CPU) of a computer system.

8. The apparatus of any one of the preceding claims wherein the processing means includes frequency domain analysis software.

15 9. The apparatus of any one of the preceding claims wherein said processing means includes filtering means to filter the input signal and remove any spikes or trough sequences caused by artefacts.

10. The apparatus of any one of the preceding claims wherein the processing means 20 further comprises an interpolating means.

.

11. The apparatus of claim 10 wherein the interpolating means interpolates the systolic peak and diastolic trough sequences to evenly spaced samples in time by staircase-type interpolation.. 25

12. The apparatus of any one of the preceding claims wherein the at least one output signal includes a plurality of frequency components determined by frequency domain analysis of the processed, at least one input signal.

30 13. The apparatus of claims 12 wherein the frequency components range from very low frequency (VLF) oscillations to relatively high frequency (HF) oscillations.

14. The apparatus of claim 12 or claim 13 wherein at least one component of the output signal comprises a very low frequency (VLF) component.

^'35

15. The apparatus of claim 14 wherein the frequency range of the VLF component is in the order of 0.015 Hz and 0.035Hz;

16. The apparatus of any one of claims 12 to 15 wherein the output signal comprises a low frequency (LF)component.

17. The apparatus of claim 16 wherein the frequency range of the LF component is in the order of 0.035Hz to 0.08Hz.

18. The apparatus of claim 16 wherein the frequency range of the LF component is less than 0.15Hz.

19. The apparatus of any one of claims 12 to 17 the output signal comprises a mid frequency component (MF).

20. The apparatus of claims 18 wherein the frequency range of the MF component is in the order of 0.08Hz to 0.150Hz.

21. The apparatus of claim 19 wherein the frequency range of the MF component is less than 0.15Hz.

22. The apparatus of any one of claims 12 to 21 wherein the output signal comprises a high frequency (HF) component.

23. The apparatus of claim 22 wherein the range of the HF component is in the order of 0.15Hz to 0.45Hz.

24. The apparatus of any one of claims 1 to 12 wherein the output signal comprises fluctuations in PPG waveform having a lower frequency relative to heart beat frequency.

25. The. apparatus of claim 22 wherein said HF component relates to respiration of a subject.

26. ^• The apparatus of claim 19 wherein the MF component relates to the autonomic nervous system of a subject.

27. The apparatus of claim 16 wherein the LF component relates to the autonomic nervous system of a subject.

28. The apparatus of any one of the preceding claims wherein the output signal is presented as a spectral plot.

29. The apparatus of any one of claims 12 to 28 wherein the output signal is presented as a spectral plot depicting spectral peaks of one or more frequency component.

30. The apparatus of claim 29 wherein spectral plot presents a sum of the spectral power over a pre-specified band of frequencies.

31. The apparatus of claim 29 wherein the spectral plot comprises an elevated power spectral density for a low density component including LF and/or MF,

32. The apparatus of claim 31 wherein said elevated spectral density for LF and/or MF is indicative of an acute coronary episode.

33. The apparatus of claim 32 wherein said acute coronary episode is myocardial infarction.

34. The apparatus of claim 3 wherein the input signals from the ECG and the PPG instrument are processed and cross correlated by a correlation means to provide a single output signal indicative of said acute coronary episode.

35. A diagnostic apparatus for detecting an acute coronary episode of an acute coronary syndrome in a patient comprising: a receiving means to receive one or more input signals from an ECG; processing means to process said one or more input signals and provide at least one output signal; output means to present said at least one output signal; characterised in that said at least one output signal is representative of heart rate variability (HRV) of the patient and wherein said output signal is indicative of whether or not said patient is experiencing said acute coronary episode.

36. A system for diagnosing an acute coronary episode of an acute coronary syndrome, said system including

(i) obtaining at least one input signals from a photoplethysmography (PPG) 5 instrument;

(ii) processing said at least one input signals and providing at least one output signal; and

(iii) wherein said at least one output signal is representative of whether or not said patient is experiencing said acute coronary episode. 10

37. A system for diagnosing an acute coronary episode of an acute coronary syndrome in a patient said system including: receiving one or more input signals from an ECG; processing said one or more input signals and providing at least one output 15 signal; characterised in that said at least one output signal is representative of heart rate variability (HRV) of the patient and wherein said output signal is indicative of whether or not said patient is experiencing said acute coronary episode

20 38. A system for diagnosing an acute coronary episode of an acute coronary syndrome, said system including

(i) obtaining at least one input signals from a photoplethysmography (PPG) instrument arid at least one signal from an ECG;

(ii) processing said signals;

25 (iii) cross correlating said processed signals; and

(iv) providing at least one output signal; wherein

(v) said at least one output signal is representative of whether or not said patient is experiencing said acute coronary episode.