JP7562338B2 - Computer-assisted method for non-invasive detection of carbohydrate metabolic disorders from electrocardiograms. - Google Patents

Description

The present invention relates to medicine, and more particularly to a computer-assisted method for detecting carbohydrate metabolism disorders from the electrocardiogram (ECG) of the heart, considered as an open nonlinear self-oscillating system characterized by a Fermi-Pasta-Ulam (FPU) autoregression, the ECG being an autoregressive representation of the FPU.

The method can be used to diagnose disorders of carbohydrate metabolism and to follow up patients to diagnose the presence or absence of diabetes. The method can be used in a variety of settings, including clinical hospitals, emergency care centers, and can also be used to follow up the condition of patients outside the hospital in outpatient care.

Type 2 diabetes (T2D) is a chronic endocrine disease that develops as a result of insulin resistance and pancreatic beta cell dysfunction.

Type 2 diabetes results from a disturbance in carbohydrate metabolism caused by increased cellular resistance to insulin (insulin resistance). The ability of tissues to accept and utilize glucose is reduced, resulting in elevated plasma glucose, a state of hyperglycemia.

The difficulty in detecting non-insulin-dependent diabetes mellitus is explained by the absence of noticeable symptoms in the early stages of the disease. In this regard, people at risk and all people over 40 years of age are recommended to have their plasma sugar levels screened. The most informative is laboratory diagnosis, since it allows the detection of not only the early stages of diabetes, but also the prediabetic state.

It is extremely urgent to improve the quality of medical care for patients with various diseases at an acceptable cost for the development of the health care system. The most important direction of this work is the development of various telemedicine methods for diagnosis, screening, monitoring and delivery of care to patients, regardless of their geography. One of these diseases is diabetes.

A method and device for non-invasive diagnosis of patients is disclosed, for example, in RU2487655C2 published on 20 July 2013. The method includes measuring biophysical parameters of patients at a diagnostic center, transferring the parameters to an integrated system for processing and storing information, registering the patient's measurement results in the system together with the patient's identification data, analyzing and diagnosing, and converting the diagnosis into a feedback signal that is sent to the diagnostic center. The diagnostic center measures a fixed list of biophysical parameters for each patient and forms an information vector of the patient's parameters resulting from the current diagnostic test.

Diagnostic tests are performed periodically, and from the difference between the information vectors of the patient's parameters obtained from the current and previous diagnostic tests, an information vector of parameter deviations is determined to assess the degree of change in the patient's condition, then a diagnosis is made using the information vector reflecting the degree of change, the patient's identification data, as well as medical history and genetic factors. If the information is insufficient for a clear diagnosis at the diagnostic center, additional measurements of biophysical parameters not included in the fixed list are made, an additional information vector of the patient's parameters is formed and a diagnosis is made taking into account the additional information vector. The results of the measurements of biophysical parameters are recorded in a unified register of all patients included in the medical support system.

The system includes m diagnostic centers housing diagnostic equipment for non-invasive measurements, a communication channel with an information processing and storage unit, a diagnostic evaluation unit, a monitoring, control and information display unit, a unit for automatically determining methods and means of treatment. All diagnostic equipment is connected to the respective inputs of an output adapter at each diagnostic center, the adapter output is connected to the input adapter of the system via a duplex communication channel, the input adapter is connected by a first output to a first input of the diagnostic evaluation unit and to a first input of the information processing and storage unit and by a second output to a first input of a unit for further follow-up of the patient's condition. Diagnostic equipment for additional measurements of biophysical parameters is connected via a communication channel to the respective inputs of the units for further follow-up of the patient's condition. The invention increases the objectivity of diagnosing a patient through computerization of the decision-making process.

The disadvantages of this device include the need to visit a diagnostic center, on the basis of which a diagnosis is made and taking into account genetic factors, repetition of diagnostic tests to detect deviations in parameters to determine the degree of changes in the patient's condition.

Most closely related to the present invention is a device for electrocardiographic monitoring of a patient's condition, disclosed, for example, in RU2615721C2 published on July 4, 2017. The device includes a monitor, an interface, ECG electrodes for acquiring electrocardiographic signals from the patient's body, which are connected to an input of a primary signal processing unit, another input of which is connected to an output of a time sampling unit, the output of which is connected to a channel switching unit. The output of the channel switching unit is connected to a discrete Fourier transform unit, the output of which contains the values of the amplitude, frequency and phase of the harmonics of the signal under study, and to a patient data input unit. The harmonics are processed in an electrocardiogram recorder, which stores and outputs the amplitudes of the harmonics of the signal under study for the required amount of time. The amplitudes of the harmonics are sent to an electrocardiogram image identifier, which compares the images received from the ECG electrodes with those from an electrocardiogram image base, taking into account the confidence interval and the specified confidence level.

The output of the identifier is connected to the input of a unit for registering the state and analyzing its evolution, and a diagnosis of the patient's disease is formed from the data of the electrocardiogram images from all ECG electrodes by comparing the set of electrocardiogram images from the ECG electrodes with a set characterizing the diagnosis of the disease from the cardiac diagnostic base, taking into account the confidence interval and the specified confidence level. The same unit determines the reliability of the diagnosis, the evolution of the diagnosis depending on the patient's previous examinations, and the time to make the diagnosis. The data are displayed on the monitor and transferred to the interface for storage and examination in other technical means, and to the patient data input unit, where they are stored in the respective patient archive.

The device is a complex design, and the disclosure does not address the criteria used to assess carbohydrate metabolic disorders based on the patient's cardiac condition.

The object of the present invention is to provide a computer-assisted method for non-invasive detection of carbohydrate metabolic disorders from the electrocardiogram (ECG) of the heart, considered as an open nonlinear self-oscillating system characterized by the autoregression of Fermi-Pasta-Ulam (FPU), the ECG being a representation of the autoregression of the FPU. A further object of the present invention is to provide a computer-assisted method for screening a population to identify those with signs of carbohydrate metabolic disorders, and a computer-assisted method for monitoring the level of carbohydrate metabolic disorders in patients.

The present invention aims to improve the accuracy of assessing a patient's cardiac condition to better detect carbohydrate metabolic disorders from electrocardiograms, and to reduce the cost and time required for diagnosis, population screening, and monitoring the level of carbohydrate metabolic disorders in patients.

The object is to provide a computer-assisted method for the non-invasive detection of carbohydrate metabolism disorders from an electrocardiogram (ECG) of the heart, considered as an open nonlinear self-oscillating system characterized by an autoregression of Fermi-Pasta-Ulam (FPU), the ECG being a representation of the autoregression of the FPU, comprising the steps of taking at least one electrocardiogram (ECG) of at least I lead , performing a Fourier transform on said at least one ECG to obtain a Fourier amplitude spectrum of said at least one ECG, discretizing the Fourier amplitude spectrum of said at least one ECG to obtain a discretized Fourier amplitude spectrum of the original Fourier amplitude spectrum of said at least one ECG. This is achieved by providing a computer-assisted method comprising the steps of: acquiring a delta ECG (DECG); and using the acquired DECG for computer processing, the computer processing including analyzing the DECG to detect the presence or absence of cardiac signs of carbohydrate metabolism disorder (CHMD) by determining similarities/differences between the DECG of a patient with carbohydrate metabolism disorder (CHMD) from the set of annotated ECGs and the DECG of a healthy subject from the set of annotated ECGs, wherein similarity between the analyzed DECG and the DECG of a CHMD patient and lack of similarity with the DECG of a healthy subject are indicative of carbohydrate metabolism disorder.

Preferably, the discretizing comprises constructing at least one ECG envelope of at least lead I that has been Fourier transformed in an averaged form.

Preferably, the discretizing includes constructing a harmonic peak envelope of at least one ECG of at least Lead I that has been Fourier transformed.

The object is to provide a computer-assisted method for screening a population to identify persons with indications of carbohydrate metabolism disorders from the electrocardiogram (ECG) of the heart, considered as an open nonlinear self-oscillating system characterized by an autoregression of Fermi-Pasta-Ulam (FPU), the ECG being an autoregressive representation of the FPU, comprising the steps of specifying target sensitivity and specificity values for the screening, preparing the screening, said preparing comprising determining from a table the minimum required number of electrocardiograms (ECGs) to be taken for one subject, taking into account the specified target sensitivity and specificity values, and a threshold value for the number of ECGs with indications of CHMD for each age-specific study group, above which the presence of CHMD is recognized in a patient of the group, and screening the population to identify persons with indications of carbohydrate metabolism disorders, said screening comprising determining from a table the minimum required number of I- leads of one subject, determined from the table, performing a Fourier transform on each ECG taken from the determined number of ECGs to obtain a Fourier amplitude spectrum of each taken ECG; discretizing the Fourier amplitude spectrum of each taken ECG to obtain a discrete form of the original Fourier amplitude spectrum (DECG) of each taken ECG; using the DECGs for computer processing, the computer processing including analyzing the DECGs to detect the presence or absence of cardiac signs of carbohydrate metabolism disorder (CHMD) by determining similarity/difference between DECGs of patients with carbohydrate metabolism disorder (CHMD) from the set of annotated ECGs and DECGs of healthy subjects from the set of annotated ECGs, where similarity between the analyzed DECGs and the DECGs of CHMD patients and lack of similarity with the DECGs of healthy subjects are indicative of carbohydrate metabolism disorder; and performing a statistical analysis of the screening results, the statistical analysis determining the number of ECGs with signs of CHMD from a minimum number of ECGs of one subject; This is further achieved by providing a computer-assisted method comprising the steps of: determining that the subject has a carbohydrate metabolism disorder based on a statistical analysis of the screening results , including comparing the number of ECGs with signs of CHMD with a predetermined threshold.

Preferably, the method further comprises determining the severity of the carbohydrate metabolism disorder based on the number of ECGs with signs of CHMD in a subject that exceed a threshold value, and making a preliminary diagnosis of diabetes based on the severity.

The object is to provide a computer-assisted method for monitoring the level of carbohydrate metabolism disorders in a patient based on the electrocardiogram (ECG) of the heart, considered as an open nonlinear self-oscillating system characterized by an autoregression of Fermi-Pasta-Ulam (FPU), the ECG being a representation of the autoregression of the FPU, comprising the steps of: collecting in a database, for each patient from a plurality of monitored patients, a series of recorded electrocardiograms (ECGs) of at least lead I of the patient, marked with a patient identifier, said series of ECGs being generated by adding each newly recorded patient ECG marked with said identifier to a series of ECGs of a plurality of patients previously collected throughout an observation period; performing a Fourier transform on each of the recorded ECGs to obtain a Fourier amplitude spectrum of each of the recorded ECGs; discretizing the Fourier amplitude spectrum of each of the recorded ECGs to obtain a discrete form of the original Fourier amplitude spectrum (DECG) of each of the recorded ECGs; and, This is also achieved by providing a computer-assisted method comprising the steps of: using the ECG for computer processing, the computer processing including detecting the presence or absence of cardiac signs of carbohydrate metabolism disorder (CHMD) by analyzing the DECG to determine similarities/differences between the DECG of a patient with carbohydrate metabolism disorder (CHMD) from the set of annotated ECGs and the DECG of a healthy subject from the set of annotated ECGs, wherein a similarity/difference between the analyzed DECG and the DECG of a CHMD patient and a lack of similarity with the DECG of a healthy subject are indicative of carbohydrate metabolism disorder; calculating the degree of difference between the analyzed DECG and the DECG of the healthy subject and using the degree as an integrated index of carbohydrate metabolism disorder; evaluating the temporal variation of the level of carbohydrate metabolism disorder of the monitored patient based on the temporal variation of the integrated index of carbohydrate metabolism disorder; and predicting the level of carbohydrate metabolism disorder of the monitored patient based on a "follow-up history" including recorded temporal variations in previous values of the integrated index of carbohydrate metabolism disorder.

The object is to provide a computer-assisted method for monitoring the level of carbohydrate metabolism disorders in a patient based on the electrocardiogram (ECG) of the heart, considered as an open nonlinear self-oscillating system characterized by an autoregression of Fermi-Pasta-Ulam (FPU), the ECG being a representation of the autoregression of the FPU, comprising the steps of: collecting in a database, for each patient from a plurality of monitored patients, a series of recorded electrocardiograms (ECGs) of at least lead I of the patient, marked with a patient identifier, said series of ECGs being generated by adding each newly recorded patient ECG marked with said identifier to a series of ECGs of a plurality of patients previously collected throughout an observation period; performing a Fourier transform on each of the recorded ECGs to obtain a Fourier amplitude spectrum of each of the recorded ECGs; discretizing the Fourier amplitude spectrum of each of the recorded ECGs to obtain the original Fourier amplitude spectrum of each of the recorded ECGs. and using the acquired DECG for computer processing, the computer processing including detecting the presence or absence of cardiac signs of carbohydrate metabolism disorder by analyzing the DECG to determine similarities/differences between the DECG of a patient with carbohydrate metabolism disorder (CHMD) from the set of annotated ECGs and the DECG of a healthy subject from the set of annotated ECGs, where a similarity/difference between the analyzed DECG and the DECG of a CHMD patient and a lack of similarity with the DECG of a healthy subject is an indication of carbohydrate metabolism disorder; evaluating the temporal variation of the level of carbohydrate metabolism disorder of the monitored patient based on the incidence of ECGs with signs of CHMD; and predicting the level of carbohydrate metabolism disorder of the monitored patient based on a "follow-up history" including recorded temporal variations in previous values of the incidence of ECGs with signs of CHMD.

The present invention provides a reliable method for non-invasive detection of disorders of carbohydrate metabolism from the electrocardiogram.

The invention allows the patient to perform a preliminary self-diagnosis of the condition and quickly and clearly register carbohydrate metabolism disorders based on the electrocardiogram.

The present invention will now be described by way of a preferred embodiment with reference to the accompanying drawings.

FIG. 1 shows the Fourier spectrum of an electrocardiogram (ECG) of a healthy subject, with the X-axis showing frequency and the Y-axis showing amplitude of harmonics. FIG. 1 shows the Fourier spectrum of an electrocardiogram (ECG) with indications of carbohydrate metabolism disorders. FIG. 1 shows the DECG of healthy individuals with known, confirmed absence of signs of carbohydrate metabolism disorders. FIG. 1 shows the DECG of hospitalized patients with signs of impaired carbohydrate metabolism. FIG. 1 shows the evolution of the ratio between ECGs with and without signs of CHMD (73 year old patient diagnosed with T2D over a 25 year period). FIG. 1 shows the combined proportion of ECGs with and without signs of CHMD during the entire observation period (73-year-old patient diagnosed with T2D over 25 years). FIG. 1 shows the evolution of the ratio between ECGs with and without signs of CHMD (73 year old patient without T2D). FIG. 1 shows the combined proportion of ECGs with and without signs of CHMD during the entire observation period (73-year-old patient without T2D). FIG. 1 shows the evolution of the ratio between ECGs with and without signs of CHMD (44 year old patient without T2D). FIG. 1 shows the combined proportion of ECGs with and without signs of CHMD during the entire observation period (44 year old patient without T2D). FIG. 1 shows the evolution of the ratio between ECGs with and without signs of CHMD (34 year old patient without T2D). FIG. 1 shows the combined proportion of ECGs with and without signs of CHMD during the entire observation period (34-year-old patient without T2D).

Rationale for the method
In 1954, scientists Fermi, Pasta, and Ulam attempted to numerically demonstrate that a nonlinear system necessarily transitions from a state of single-mode excitation to a state in which the energy is uniformly distributed.

The inventors proposed a version of the Fermi-Pasta-Ulam regression model (FPU autorecurrence), mathematically explained it, and hypothesized a suitable description of the cardiac electrodynamics within the detected FPU autoregressive phenomenon. Autoregression is a pattern in which the states of a nonlinear system are repeated in time. The evolution of the solutions of the FPU autoregression, which correspond to the electrodynamics of a normally functioning heart in the form of an electrocardiogram (ECG), was obtained by computer modeling studies (see Schmid A., Berezin A., Novopashin M. "Fermi-Pasta-Ulam auto recurrence in the description of the electrical activity of the heart", Journal Medical Hypothesis, 101 (2017) pp. 17-22).

The heart is an open nonlinear self-oscillating system that can be modeled, subject to certain restrictions, by a system of two coupled van der Pol equations.

A large-scale study using big data technology of at least lead I ECGs from the CardioQwark Internet electrocardiogram collected from a group of patients over several months of constant monitoring has revealed some regularities that do not fit the above model and are characteristic of both healthy people and patients with some kind of heart disease.

To test the hypothesis, over 20,000 ECGs from both healthy individuals and patients with carbohydrate metabolism disorders were examined.

By processing the ECG arrays of healthy subjects, the characteristic types of Fourier spectra of healthy subject ECGs were determined.

To simulate the ECG, it is taken into account that the heart works in a self-oscillating mode, which means that there are similar principle evolutions in the recursive structure of the FPU used to simulate cardiac activity.

The results show that the electrical activity of a normally functioning heart can be interpreted by the autoregressive phenomenon of the FPU, while carbohydrate metabolism disorders are characterized by the appearance of noise when energy flows between the low- and high-frequency harmonics of the autoregressive FPU.

Mathematical simulations have shown that the electrical activity of the heart represents an autoregression of the FPU, originally formulated by the inventor, which is of an individual nature and whose structure contains a picture of the physiological state of the heart and pathological processes in the body, the patterns of which appear periodically in the Fourier spectrum of the autoregression.

The autoregression of the Fourier spectrum of the ECG of a healthy subject contains both a low-frequency part in the range of 1-5 Hz, corresponding to the regression of a standard FPU, and a high-frequency part in the range of 20-35 Hz, corresponding to the regression of a high-frequency FPU. Thus, the ECG is a biological example of a complete FPU autoregression.

Based on the simulation results, it is suggested that the sets of frequency patterns obtained from different electrocardiograms of one or more patients do not have a random distribution, and therefore some sets of frequency patterns can be distinguished by some degree of proximity.

This means that the clustering algorithm can be applied to multiple ECG frequency patterns and stable frequency patterns can be obtained for both the condition of a particular patient and for groups of patients selected for common features.

For the purposes of the study, the inventors used data from 1,997 electrocardiograms of 128 patients with a stay of 5 days or more at the Endocrinology Center of the State Budgetary Medical Institution "City Clinical Hospital No. 52" and the State Budgetary Medical Institution "Moscow Clinical Scientific Center named after A. Loginov", and as a reference group, data from 741 electrocardiograms of 59 patients of various age groups without carbohydrate metabolism disorders (CHMD).

The inventors used the CardioQVARK system (cardio monitor + software) to study the deviations of ECG waveforms against the background of treatment administered during inpatient follow-up of patients with carbohydrate metabolism disorders (CHMD), as well as the differences between ECG waveforms of subjects without CHMD (reference group) and ECG waveforms of CHMD patients.

Fourier spectra of electrocardiograms (ECGs) of healthy subjects (Figure 1) and subjects with signs of carbohydrate metabolic disorder (CHMD) (Figure 2) were obtained.

The Fourier spectrum of an ECG with signs of CHMD is characterized by a significant intensity of frequency harmonics, whereas the Fourier spectrum of an ECG from a healthy subject is characterized by a lack of significant intensity of frequency harmonics.

The electrocardiogram is a repetitive signal characterized by harmonics in the Fourier spectrum. The energy transfer between the low frequency (LF) and high frequency (HF) ranges is characterized by the amount of noise between them. If the level of noise is high, it means that a process of energy flowing back and forth is ongoing. If a violation of the synchronization of the energy flow process is observed, the inventors consider this to be a sign of carbohydrate metabolism disorders.

Autoregression is the flow of energy back and forth between the high and low frequency spectra.

You don't need to have several ECGs to compare that in one ECG the energy flows one way and in another ECG it flows in a somewhat different way. Once you have an ECG, it's possible to determine by noise level whether energy is flowing or not.

It is impossible to determine where the flow is coming from and where it is going. Even when comparing multiple ECGs, you always only see a static image.

In multiple ECGs, it is seen successively that the LF part of one ECG decreases and the HF part increases. But in reality they increase differently, i.e. some of the LF harmonics may increase and some decrease.

For each spectrum, the percentage of chaoticity can be calculated by the harmonic distribution. If a person is healthy, his harmonic distribution curve will be convex (axis Y is the percentage of chaoticity, axis X is the frequency), while for diabetics, a dip (concave) of the curve is obtained, and for prediabetics, the dip is smaller.

Not all states are traced, but often processes that change states (transition states) are traced. To trace multiple states, many ECGs must be taken.

One ECG may show that the patient is in this state and that this state is generalizing. Another ECG may show that the patient is passing from one state to another. Whether the high frequencies increase or decrease, this is just another frame of an entire process evolving in time.

The functional activity of the brain is closely linked to the use of glucose, and therefore, when the glucose concentration in the bloodstream of the cerebral vessels decreases, it is necessary to increase the amplitude of the impulses triggering the heart from the brain in order to increase the blood flow intensity. This is due to a significant increase in the amplitude of the brain-generated impulses entering the sinus node, which leads to the formation of a more rigid shape of the Fourier harmonics of the ECG spectrum with a narrow pedestal, which is seen in patients with carbohydrate metabolism disorders. This, in turn, reduces the percentage of chaoticity in the spectrum, which may serve as a diagnostic sign of carbohydrate metabolism disorders. This mode of cardiac activity corresponds to a certain solution of the order-disorder type of Ginzburg-Landau string. It should be noted that the magnitude of the feedback delay with the decrease in the glucose concentration in the cerebral bloodstream may be comparable to the reaction time of the brain at the onset of hypoglycemic coma.

In addition to the circadian rhythm of function, the pancreas also has a functional rhythm synchronized with the main frequency of cardiac contraction, which is equal to 1-2 Hz. Violations of this rhythm of the pancreas, as shown by computer studies of the evolution of the autoregressive spectra of FPUs, lead to a decrease in the energy of the lowest frequency harmonics of the spectrum, which can be seen in the spectra of patients with carbohydrate metabolism disorders.

Description of the Preferred Embodiments According to the present invention, a computer-assisted method is provided for non-invasive detection of carbohydrate metabolism disorders from a cardiac electrocardiogram (ECG). As described above, the heart is considered as an open nonlinear self-oscillating system characterized by a Fermi-Pasta-Ulam (FPU) autoregression, and the ECG is a representation of the FPU autoregression. The method includes the following steps:

At least one ECG is taken, at least in lead I. A single ECG is sufficient to detect the presence of CMHD.

A Fourier transform is performed on each of the at least one ECG to obtain a Fourier amplitude spectrum of the at least one ECG (Figures 1 and 2).

The Fourier amplitude spectrum of the at least one ECG is discretized (thick solid line in Fig. 1 or Fig. 2) to obtain a discrete form of the original Fourier amplitude spectrum (DECG) of the at least one ECG. The obtained DECG is used for computer processing, which includes analyzing the DECG to detect the presence or absence of cardiac signs of carbohydrate metabolism disorders (CHMD) by determining similarities/differences between the DECG of CHMD patients from the set of annotated ECGs (Fig. 3B) and the DECG of healthy controls from the set of annotated ECGs (Fig. 3A).

The similarity between the analyzed DECG and that of CMHD patients, and the lack of similarity with that of healthy controls, is an indication of a carbohydrate metabolism disorder.

In this method, the discretization is performed, for example , by constructing at least one envelope of the ECG that has been Fourier transformed in averaged form (thick solid line in Figs. 1 and 2).

The discretization can be performed by constructing the Fourier transformed harmonic peak envelope of at least one ECG (bold dots in Figures 1 and 2).

According to the present invention, there is also provided a computer-assisted method for screening a population to detect individuals with indications of carbohydrate metabolism disorders based on the electrocardiogram (ECG) of the heart, considered as an open nonlinear self-oscillating system characterized by the autoregression of Fermi-Pasta-Ulam (FPU), the ECG being a representation of the autoregression of the FPU. The screening method is carried out as follows:

Screening sensitivity and specificity targets required to evaluate the efficiency of the method are specified.

The sensitivity of a study is its ability to distinguish between two states of the analyzed individuals, healthy individuals and those with pathology, and sensitivity quantitatively reflects the ratio between the positive positive conclusion of the method and the total number of final diagnoses in a group of patients.

Specificity is the ability to dismiss a disease, i.e., to state the absence of a disease when in fact it is not present. Specificity quantitatively reflects the proportion of negative conclusions in a group of disease-free individuals.

To determine sensitivity and specificity, a group of people is studied where each person is known to be healthy or diseased.

Sensitivity, specificity, and predictive value for a positive result (PPR) are calculated using the following formula:

In the formula, A is the number of true positive conclusions, B is the number of false positive conclusions, C is the number of false negative conclusions, and D is the number of true negative conclusions.

To calculate the sensitivity and specificity for detecting ECG disorders in patients with carbohydrate metabolism disorders, two groups were formed.
Patients - Inpatients at City Clinical Hospital No. 52 MKSC with type 2 diabetes who had an ECG taken at least twice (127 patients).
Healthy subjects-staff of Municipal Clinical Hospital No. 52 MKSC with non-diabetic and non-diabetic subjects who had ECG taken at least twice (59 people).

Table 1 shows the results of applying the method on a single ECG.

The values of sensitivity and specificity can be significantly increased by basing conclusions on several ECGs of a patient, rather than on a single one. Based on the results of the study, thresholds for the specified number of ECGs, the minimum number of ECGs containing signs of CMHD, as well as the corresponding values of sensitivity and specificity are created in a table (Table 2).

Preparation for screening is then carried out, which involves determining from Table 2 the minimum required number of electrocardiograms (ECGs) per subject, taking into account the specified target values of sensitivity and specificity, and the thresholds for the number of ECGs with signs of CMHD for each age group to be monitored.

The population is then screened to identify individuals with symptoms of impaired carbohydrate metabolism.
Obtaining the minimum required number of electrocardiograms (ECGs) of at least lead I for each subject, as determined from a table;
performing a Fourier transform on each taken ECG to obtain a Fourier magnitude spectrum of each taken ECG;
discretizing the Fourier magnitude spectrum of each taken ECG to obtain a discrete form of the original Fourier magnitude spectrum (DECG) of each taken ECG;
and using the DECG for computer processing, the using including detecting the presence or absence of cardiac signs of carbohydrate metabolism disorder (CHMD) by analyzing the DECG to determine similarities/differences between the DECG of a patient with carbohydrate metabolism disorder (CHMD) from the set of annotated ECGs and the DECG of a healthy subject from the set of annotated ECGs, wherein similarity between the analyzed DECG and the DECG of a CHMD patient and lack of similarity with the DECG of a healthy subject is an indication of carbohydrate metabolism disorder.

A statistical analysis of the screening results is then performed, which involves determining the number of ECGs with signs of CHMD from the minimum required number of ECGs per subject, comparing it with the thresholds specified during the preparation of the screening (from Table 2), and determining, based on the statistical analysis, that the subject has a carbohydrate metabolism disorder.

The method may further include determining the severity of carbohydrate metabolic disorder from the number of ECGs with signs of CHMD in a subject that exceed a threshold, and making a preliminary diagnosis of diabetes based on the determined result.

According to the present invention, there is also provided a computer-assisted method for monitoring the level of carbohydrate metabolism disorders in a patient based on the electrocardiogram (ECG) of the heart, considered as an open nonlinear self-oscillating system characterized by the autoregression of Fermi-Pasta-Ulam (FPU), the ECG being a representation of the autoregression of the FPU. The monitoring method is carried out as follows:

For each patient from the plurality of monitored patients, a series of recorded electrocardiograms (ECGs) of at least lead I marked with a patient identifier is stored in a database. The series of ECGs is generated by adding each newly recorded ECG of the patient to a series of ECGs of the patient previously collected throughout an observation period marked with the identifier.

A Fourier transform is performed on each of the patient's collected ECGs to obtain a Fourier magnitude spectrum of each of the collected ECGs.

The Fourier amplitude spectrum of each collected ECG is discretized (thick lines in Figures 1 and 2) to obtain a discrete form of the Fourier amplitude spectrum (DECG) of each ECG of the collected ECGs.

The obtained DECGs are used for computer processing, which includes detecting the presence or absence of cardiac signs of carbohydrate metabolism disorder (CHMD) by analyzing the DECGs to determine similarities/differences between the DECGs of patients with carbohydrate metabolism disorder (CHMD) from the set of annotated ECGs and the DECGs of healthy subjects from the set of annotated ECGs, where similarity between the analyzed DECGs of CHMD patients and lack of similarity with the DECGs of healthy subjects is an indication of carbohydrate metabolism disorder.

The patient's condition can then be monitored over time, in one embodiment, by the temporal variation of an integrated indicator of carbohydrate metabolism disorder (CMHD), or in another embodiment, by the variation of the incidence of CMHD signs in the patient's electrocardiogram.

When monitoring a patient's condition based on the time-dependent changes in the integrated DECG index, the degree of difference between the analyzed DECG and that of healthy individuals is calculated and used as an integrated index of carbohydrate metabolism disorders.

Based on the time variation of the integrated indicator of carbohydrate metabolic disorder, the time variation of the level of carbohydrate metabolic disorder in the monitored patient is evaluated.

Based on the "follow-up history", which includes the recorded temporal variations in previous values of the integrated indicators of carbohydrate metabolic disorders, the level of carbohydrate metabolic disorders in the monitored patients is predicted.

Based on the variation in the incidence of CMHD signs in the patients' electrocardiograms, all series of ECGs are divided into two classes: those with CMHD signs and those without CMHD signs (Figure 4A, Figure 5A, Figure 6A, Figure 7A).

The results of the analysis of the presence of signs of CHMD in a series of N ECGs of a patient are statistically summarized. Depending on the incidence of signs of CHMD in a series, a conclusion is made about the severity of the carbohydrate metabolism disorder already present in the patient, up to the point of preliminarily diagnosing the presence of diabetes mellitus.

The progression of increase or decrease in the incidence of ECGs with signs of CMHD in patients is monitored based on observation of the entire array of previously accumulated ECGs of patients by comparing the incidence of ECGs with signs of CMHD for previous and subsequent observation periods of days, weeks, months, and years.

In diabetes, coronary heart disease progresses rapidly and often leads to heart attacks, so it is important for patients to non-invasively detect carbohydrate metabolic disorders themselves from electrocardiograms in various conditions.

ECG diagnosis helps in emergency situations to find the cause of a heart attack within minutes and provide targeted assistance without delay.

Below is an example of monitoring the level of carbohydrate metabolic disorder in a patient using the cardiac electrocardiogram (ECG), which is considered as an open nonlinear self-oscillating system characterized by the autoregression of a Fermi-Pasta-Ulam (FPU), where the ECG is a representation of the autoregression of the FPU.

In particular, patients with a proven diagnosis of type 2 diabetes and patients without a proven diagnosis of diabetes, i.e. healthy controls, were studied.

Example 1
Patient A, 73 years old. Diagnosis: type 2 diabetes mellitus for 25 years. No associated chronic diseases. Examination results show good general medical condition.

Electrocardiogram (ECG) monitoring was performed. ECGs were taken 5 to 20 times per week from September 2017 to August 2019 (Figure 4A).

A series of recorded electrocardiograms (ECGs) of at least lead I of a patient, marked with a patient identifier, was stored in a database. A series of ECGs was generated by adding each newly recorded ECG of a patient to a series of ECGs previously stored throughout the observation period.

A Fourier transform was performed on each of the recorded ECGs to obtain a Fourier amplitude spectrum for each of the recorded ECGs.

The Fourier amplitude spectrum of each of the recorded ECGs was discretized to obtain a discrete form (DECG) of the original Fourier amplitude spectrum of each of the recorded ECGs (examples in Figures 1 and 2).

Finally, all ECGs in the series were divided into two classes: those with signs of CMHD and those without signs of CHMD (Figure 4A).

A computational procedure was performed to analyze the DECGs to detect the presence or absence of cardiac manifestations of carbohydrate metabolism disorders (CHMD) by determining the similarities/differences between the DECGs of patients with carbohydrate metabolism disorders (CHMD) from the set of annotated ECGs and the DECGs of healthy controls from the set of annotated ECGs.

The degree of difference between the analyzed DECG and that of healthy subjects was calculated and used as an integrated index of carbohydrate metabolism disorders.

Based on the temporal variation of the integrated index of carbohydrate metabolic disorder, the temporal variation of the level of carbohydrate metabolic disorder in patients was evaluated.

Based on the "follow-up history" including the registered temporal variations in previous values of the integrated index of carbohydrate metabolic disorders, the level of carbohydrate metabolic disorders in patient A was predicted.

Over the entire observation period, the total proportion of ECGs with cardiac signs of CMHD was 99%, and ECGs without cardiac signs of CMHD was 1% (Figure 4B).

The evolution of the increase or decrease in the incidence of ECGs with signs of CMHD in patients was monitored based on observation of the entire array of previously collected ECGs of patients by comparing the incidence of ECGs with signs of CMHD for the previous and subsequent observation periods of days, weeks, months and years.

The apparent trend (dashed line) for patient A with proven type 2 diabetes shows no shift in the incidence of CMHD symptoms (Figure 4A), which may indicate a positive shift toward normalization of carbohydrate metabolism.

From the above, we can see that the incidence of CMHD symptoms in patients with proven type 2 diabetes has not changed over time (Figure 4A).

Example 2
Patient B, 73 years old. Diabetes has not been diagnosed. No associated chronic diseases. Examination results indicate good general medical condition.

Electrocardiogram (ECG) monitoring was performed. ECGs were taken 5 to 20 times per week from February 2016 to August 2019 (Figure 5A).

A series of recorded electrocardiograms (ECGs) of at least lead I , marked with a patient identifier, was stored in a database. A series of ECGs was generated by adding each newly recorded patient ECG to a series of previously stored ECGs throughout the observation period.

A Fourier transform was performed on each of the recorded ECGs to obtain a Fourier amplitude spectrum for each of the recorded ECGs.

The Fourier amplitude spectrum of each of the recorded ECGs was discretized to obtain a discrete form (DECG) of the original Fourier amplitude spectrum of each of the recorded ECGs (examples in Figures 1 and 2).

Finally, all ECGs in the series were divided into two classes: those with signs of CMHD and those without signs of CMHD (Figure 5A).

A computational procedure was performed to analyze the DECGs to detect the presence or absence of cardiac manifestations of carbohydrate metabolism disorders (CHMD) by determining the similarities/differences between the DECGs of patients with carbohydrate metabolism disorders (CHMD) from the set of annotated ECGs and the DECGs of healthy controls from the set of annotated ECGs.

The degree of difference between the analyzed DECG and that of healthy subjects was calculated and used as an integrated index of carbohydrate metabolism disorders.

Based on the temporal variation of the integrated index of carbohydrate metabolic disorder, the temporal variation of the level of carbohydrate metabolic disorder in patients was evaluated.

Based on the "follow-up history" including the recorded temporal variations in previous values of the integrated index of carbohydrate metabolic disorders, the level of carbohydrate metabolic disorders in patient B was predicted.

Over the entire observation period, the total proportion of ECGs with cardiac signs of CMHD was 47% and 53% without signs (Figure 5B).

The evolution of the increase or decrease in the incidence of patients with ECGs with signs of CMHD was monitored based on observation of the entire array of previously accumulated patient ECGs by comparing the incidence of ECGs with signs of CMHD for previous and subsequent observation periods of days, weeks, months and years.

The apparent trend in patient B, who did not have diagnosed diabetes (dashed line), shows a decreased incidence of CMHD symptoms, which may indicate a positive transition toward normalization of carbohydrate metabolism.

Therefore, this is an example of follow-up of patients with a high incidence of CHMD manifestations.

Example 3
Patient C, 44 years old. Diabetes has not been diagnosed. No associated chronic diseases. Examination results indicate good general condition.

Electrocardiogram (ECG) monitoring was performed. ECGs were taken 5 to 20 times per week from April 2017 to August 2019 (Figure 6A).

A series of recorded electrocardiograms (ECGs) of at least lead I , marked with a patient identifier, was stored in a database. A series of ECGs was generated by adding each newly recorded patient ECG to a series of previously stored ECGs throughout the observation period.

A Fourier transform was performed on each of the recorded ECGs to obtain a Fourier amplitude spectrum for each of the recorded ECGs.

The Fourier amplitude spectrum of each of the recorded ECG sets was discretized to obtain a discrete form (DECG) of the original Fourier amplitude spectrum of each of the recorded ECG sets (examples in Figures 1 and 2).

Finally, all ECGs in the series were divided into two classes: those with signs of CMHD and those without signs of CMHD (Figure 6A).

A computational procedure was performed to analyze the DECGs to detect the presence or absence of cardiac manifestations of carbohydrate metabolism disorders (CHMD) by determining the similarities/differences between the DECGs of patients with carbohydrate metabolism disorders (CHMD) from the set of annotated ECGs and the DECGs of healthy controls from the set of annotated ECGs.

The degree of difference between the analyzed DECG and that of healthy subjects was calculated and used as an integrated index of carbohydrate metabolism disorders.

Based on the temporal variation of the integrated index of carbohydrate metabolic disorder, the temporal variation of the level of carbohydrate metabolic disorder in patients was evaluated.

Based on the "follow-up history" including the recorded temporal variations in previous values of the integrated index of carbohydrate metabolic disorders, the level of carbohydrate metabolic disorders in patient C was predicted.

Over the entire observation period, the total proportion of ECGs with cardiac signs of CMHD was 29% and 71% without signs (Figure 6B).

The evolution of the increase or decrease in the incidence of ECGs with signs of CMHD in patients was monitored based on observation of the entire array of previously accumulated patient ECGs by comparing the incidence of ECGs with signs of CMHD for previous and subsequent observation periods of days, weeks, months and years.

The apparent trend (dashed line) for patient C, who did not have diagnosed diabetes, shows a decreased incidence of CMHD symptoms (Figure 4A), which may indicate a positive transition toward normalization of carbohydrate metabolism.

Therefore, this is an example of patient follow-up with a reduced incidence of CHMD manifestations.

Example 4
Patient D, 34 years old. Undiagnosed diabetes. No associated chronic diseases. Examination results indicate good general condition.

Electrocardiogram (ECG) monitoring was performed. ECGs were taken 5 to 20 times per week from April 2017 to August 2019 (Figure 7A).

A series of recorded electrocardiograms (ECGs) of at least lead I , marked with a patient identifier, was stored in a database. A series of ECGs was generated by adding each newly recorded patient ECG to a series of previously stored ECGs throughout the observation period.

A Fourier transform was performed on each of the recorded ECGs to obtain a Fourier amplitude spectrum for each of the recorded ECGs.

The Fourier amplitude spectrum of each of the recorded ECGs was discretized to obtain a discrete form (DECG) of the original Fourier amplitude spectrum of each of the recorded ECGs (examples in Figures 1 and 2).

Finally, all ECGs in the series were divided into two classes: those with signs of CMHD and those without signs of CMHD (Figure 7A).

A computational procedure was performed to analyze the DECGs to detect the presence or absence of cardiac manifestations of carbohydrate metabolism disorders (CHMD) by determining the similarities/differences between the DECGs of patients with carbohydrate metabolism disorders (CHMD) from the set of annotated ECGs and the DECGs of healthy controls from the set of annotated ECGs.

The degree of difference between the analyzed DECG and that of healthy subjects was calculated and used as an integrated index of carbohydrate metabolism disorders.

Based on the temporal variation of the integrated index of carbohydrate metabolic disorder, the temporal variation of the level of carbohydrate metabolic disorder in patients was evaluated.

Based on the "follow-up history" including the recorded temporal variations in previous values of the integrated index of carbohydrate metabolic disorders, the level of carbohydrate metabolic disorders in patient D was predicted.

Over the entire observation period, the total proportion of ECGs with cardiac signs of CMHD was 9% and ECGs without signs was 91% (Figure 7B).

The evolution of the increase or decrease in the incidence of ECGs with signs of CMHD in patients was monitored based on observation of the entire array of previously collected ECGs of patients by comparing the incidence of ECGs with signs of CMHD for the previous and subsequent observation periods of days, weeks, months and years.

The apparent trend (dashed line) for patient D, who did not have diagnosed diabetes, shows a decreased incidence of CMHD symptoms, which may indicate a positive transition toward normalization of carbohydrate metabolism.

Therefore, this is an example of patient follow-up with a reduced incidence of CHMD manifestations.

Claims (7)

A computer-assisted method for non-invasive detection of carbohydrate metabolism disorders from an electrocardiogram (ECG) of the heart, considered as an open nonlinear self-oscillating system characterized by an autoregression of Fermi-Pasta-Ulam (FPU), said ECG being a representation of the autoregression of said FPU,
taking at least one electrocardiogram (ECG) of at least lead I ;
performing a Fourier transform on the at least one ECG to obtain a Fourier magnitude spectrum of the at least one ECG;
discretizing the Fourier magnitude spectrum of the at least one ECG to obtain a discrete form of the original Fourier magnitude spectrum (DECG) of the at least one ECG;
Using the obtained DECG for computer processing, the computer processing comprising:
detecting the presence or absence of cardiac manifestations of carbohydrate metabolism disorder (CHMD) by analyzing the DECG to determine similarities/differences between DECGs of patients with carbohydrate metabolism disorder (CHMD) from the set of annotated ECGs and DECGs of healthy subjects from the set of annotated ECGs, wherein similarity between the analyzed DECGs and DECGs of CHMD patients and lack of similarity between the analyzed DECGs and DECGs of healthy subjects is an indication of carbohydrate metabolism disorder;
A computer-assisted method comprising: 2. The computer-assisted method of claim 1, wherein the discretizing comprises constructing an envelope of at least one of the Fourier transformed ECGs of at least Lead I in an averaged form. 2. The computer-assisted method of claim 1, wherein the discretizing comprises constructing a harmonic peak envelope of at least one of the Fourier transformed ECGs of at least Lead I. 1. A computer-assisted method for screening a population to identify individuals with indications of carbohydrate metabolism disorders from a cardiac electrocardiogram (ECG), viewed as an open nonlinear self-oscillating system characterized by an autoregression of Fermi-Pasta-Ulam (FPU), wherein the ECG is a representation of the autoregression of the FPU,
specifying target sensitivity and specificity values for screening;
preparing a screening step, said preparing including determining from a table the minimum required number of electrocardiograms (ECGs) to be taken per subject, taking into account the specified target sensitivity and specificity values, and a threshold number of ECGs with signs of CHMD for each age group study group, above which a patient in said age group study group is recognized to have CHMD;
and screening the population to identify individuals having an indication of a carbohydrate metabolism disorder, said screening comprising:
taking said minimum required number of electrocardiograms (ECGs) of at least lead I for one subject as determined from said table;
performing a Fourier transform on each taken ECG from said determined number of ECGs to obtain a Fourier magnitude spectrum of each taken ECG;
discretising the Fourier magnitude spectrum of each of the taken ECGs to obtain a discrete form of the original Fourier magnitude spectrum (DECG) of each of the taken ECGs;
using the DECG for computer processing, the computer processing comprising:
detecting the presence or absence of cardiac signs of carbohydrate metabolism disorder (CHMD) by analyzing the DECGs to determine similarities/differences between DECGs of patients with carbohydrate metabolism disorder (CHMD) from the set of annotated ECGs and DECGs of healthy subjects from the set of annotated ECGs, wherein similarity between the analyzed DECGs and DECGs of CHMD patients and lack of similarity between the analyzed DECGs and DECGs of healthy subjects is an indication of carbohydrate metabolism disorder;
performing a statistical analysis of the results of the screening, the statistical analysis including: determining the number of ECGs having signs of CHMD from a minimum number of ECGs to be taken per subject; and comparing the number of ECGs having signs of CHMD with a predefined threshold; and determining that the subject has a carbohydrate metabolism disorder based on the statistical analysis of the results of the screening ;
A computer-assisted method comprising: The computer-assisted method of claim 4, further comprising determining a severity of carbohydrate metabolic disorder based on the number of ECGs with indicia of CHMD in a subject that exceed the threshold, and making a prediagnosis of diabetes based on the severity. 1. A computer-assisted method for monitoring the level of carbohydrate metabolic disorders in a patient based on an electrocardiogram (ECG) of the heart, considered as an open nonlinear self-oscillating system characterized by an autoregression of Fermi-Pasta-Ulam (FPU), said ECG being a representation of the autoregression of said FPU;
collecting in a database, for each patient from a plurality of monitored patients, a series of recorded electrocardiograms (ECGs) of at least lead I of said patient, marked with a patient identifier, said series of ECGs being generated by adding each newly recorded patient ECG marked with said identifier to a series of ECGs of a plurality of patients previously collected throughout an observation period;
performing a Fourier transform on each of the recorded ECGs to obtain a Fourier magnitude spectrum for each of the recorded ECGs;
discretizing the Fourier magnitude spectrum of each of the recorded ECGs to obtain a discrete form of the original Fourier magnitude spectrum (DECG) of each of the recorded ECGs;
Using the obtained DECG for computer processing, the computer processing comprising:
detecting the presence or absence of cardiac manifestations of carbohydrate metabolism disorder (CHMD) by analyzing the DECG to determine similarities/differences between DECGs of patients with carbohydrate metabolism disorder (CHMD) from the set of annotated ECGs and DECGs of healthy subjects from the set of annotated ECGs, wherein similarity between the analyzed DECGs and DECGs of CHMD patients and lack of similarity between the analyzed DECGs and DECGs of healthy subjects is an indication of carbohydrate metabolism disorder;
Calculating the degree of difference between the analyzed DECG and the DECG of healthy subjects, and using the degree as an integrated index of carbohydrate metabolism disorder;
assessing the temporal variation of the level of carbohydrate metabolic disorder in the monitored patient based on the temporal variation of the integrated indicator of carbohydrate metabolic disorder;
predicting said level of carbohydrate metabolism disorder in said monitored patient based on a "follow-up history" including recorded temporal variations in previous values of said integrated index of carbohydrate metabolism disorder;
The computer-aided method includes: 1. A computer-assisted method for monitoring the level of carbohydrate metabolic disorders in a patient based on an electrocardiogram (ECG) of the heart, considered as an open nonlinear self-oscillating system characterized by an autoregression of Fermi-Pasta-Ulam (FPU), said ECG being a representation of the autoregression of said FPU;
collecting in a database, for each patient from a plurality of monitored patients, a series of recorded electrocardiograms (ECGs) of at least lead I of said patient, marked with a patient identifier, said series of ECGs being generated by adding each newly recorded patient ECG marked with said identifier to a series of ECGs of a plurality of patients previously collected throughout an observation period;
performing a Fourier transform on each of the recorded ECGs to obtain a Fourier magnitude spectrum for each of the recorded ECGs;
discretizing the Fourier magnitude spectrum of each of the recorded ECGs to obtain a discrete form of the original Fourier magnitude spectrum (DECG) of each of the recorded ECGs;
Using the obtained DECG for computer processing, the computer processing comprising:
detecting the presence or absence of cardiac manifestations of carbohydrate metabolism disorder (CHMD) by analyzing the DECG to determine similarities/differences between DECGs of patients with carbohydrate metabolism disorder (CHMD) from the set of annotated ECGs and DECGs of healthy subjects from the set of annotated ECGs, wherein similarity between the analyzed DECGs and DECGs of CHMD patients and lack of similarity between the analyzed DECGs and DECGs of healthy subjects is an indication of carbohydrate metabolism disorder;
assessing the temporal variation of the level of carbohydrate metabolic disorder in the monitored patient based on the incidence of electrocardiograms with signs of CHMD;
predicting the level of carbohydrate metabolic disorder in the monitored patient based on a "follow-up history" including recorded temporal variations in previous values of the incidence of electrocardiograms with signs of CHMD;
A computer-assisted method comprising:

Images