Bioinformatics analysis system and bioinformatics analysis program

Abstract

The biological information analysis system according to the present application includes: a heartbeat cycle calculating section that calculates a heartbeat cycle of a subject from biological information acquired from the subject; an analysis unit determining section that divides the heartbeat cycle into analysis units using a predetermined segment length; a clustering section that clusters the heartbeat cycle for each analysis unit using a probability distribution model related to the heartbeat cycle of the analysis unit; and a determination section that determines arrhythmia based on a result of the clustering.

Description

Technical Field

[0001] This invention relates to a series of bioinformatics analysis systems and bioinformatics analysis programs for determining arrhythmias based on biological information obtained from an object. Background Technology

[0002] The heart pumps blood throughout the body by repeatedly contracting and expanding. This activity (hereafter referred to as heartbeat) is kept regular by weak electrical stimulation of the heart muscle cells, but when this electrical stimulation becomes abnormal, the heart's activity also becomes abnormal. This is called arrhythmia. There are various types of arrhythmias, some of which can cause fatal cardiac arrest, and others can lead to serious illnesses such as stroke. Therefore, early detection of arrhythmias is crucial. For example, techniques are known to use bioinformatics obtained from a subject by a bioinformatics analyzer to determine arrhythmias.

[0003] Here, Non-Patent Document 1 discloses a technique related to a method for determining arrhythmias. Specifically, among various types of arrhythmias, atrial fibrillation (AF) is identified based on the magnitude of deviations in the periodic information of the heartbeat (hereinafter referred to as the heartbeat cycle). Figure 19 This is a diagram illustrating an example of the distribution of heartbeat cycles in atrial fibrillation. Additionally, Figure 19 The curve shown represents an approximate curve (fitted curve). Atrial fibrillation is a long-term, continuous arrhythmia characterized by irregular heartbeat cycles; therefore, it can be approximated based on curves such as... Figure 19 The relationship between the heartbeat cycle and the probability density is shown, and the determination is based on the magnitude of the deviation of the heartbeat cycle.

[0004] However, in the method of Non-Patent Literature 1, there is also a concern that during periods of frequent premature contractions (PVC, PAC) of arrhythmia, they may be misdiagnosed as atrial fibrillation. Figure 20A and Figure 20B This is a diagram illustrating an example of the distribution of the heartbeat cycle during premature ventricular contractions. Furthermore, Figure 20A and Figure 20B The curve shown represents an approximate curve (fitted curve). Compensatory premature contractions in premature contractions are arrhythmias where the heartbeat occurs ahead of the normal rhythm, thus increasing the deviation of the heartbeat cycle and being misdiagnosed as atrial fibrillation (see reference). Figure 20A To address this issue, Patent Document 1 assumes multiple distributions for the heartbeat cycle and, based on the number of distributions and the magnitude of their deviations, determines premature contractions in addition to atrial fibrillation (see [reference]). Figure 20B ).

[0005] Existing technical documents Non-patent literature Non-patent literature 1: B Logan, J Healey. Robust Detection of Atrial Fibrillation for a Long Term Telemonitoring System, Computers in Cardiology 2005;32:619-622. Patent documents Patent document 1: Japanese Patent Application Publication No. 2009-089883. Summary of the Invention

[0006] The problem the invention aims to solve While arrhythmias can be determined based on the distribution of the heartbeat cycle, as in Non-Patent Document 1 and Patent Document 1, the detection accuracy of the heartbeat cycle distribution in the prior art is low. For example, assuming that the distribution is divided into multiple parts in atrial fibrillation or that the distributions that should be divided into multiple parts in premature contractions are mixed, this can lead to misdiagnosis. Therefore, there is a need to seek high-precision detection of the heartbeat cycle distribution for arrhythmia determination.

[0007] To address the aforementioned problems, this invention provides a bioinformatics analysis system and a bioinformatics analysis program capable of classifying arrhythmias with high accuracy.

[0008] Solution for solving the problem The bioinformatics analysis system of the present invention, which achieves the above objectives, includes the following structure.

[0009] (1) A bioinformatics analysis system that uses the biological information of an object to determine arrhythmia, the bioinformatics analysis system comprising: The heartbeat cycle calculation unit calculates the interval of the object's heartbeat activity as the heartbeat cycle based on the biological information obtained from the object; The analysis unit determines the segment length as the analysis unit to divide the heartbeat cycle; A clustering unit that uses a probability distribution model related to the heartbeat cycles of the analysis unit to cluster the heartbeat cycles of the analysis unit; and The determination unit determines arrhythmias based on the clustering results.

[0010] (2) According to the bioinformatics analysis system described in (1) above, wherein, The probability distribution model is generated through learning related to the heartbeat cycle of the unit of analysis.

[0011] (3) According to the bioinformatics analysis system described in (1) above, wherein, The probability distribution model used is a Gaussian mixture model.

[0012] (4) According to the bioinformatics analysis system described in (1) above, wherein, The clustering department uses information content criteria to select the probability distribution model and uses the selected probability distribution model to perform clustering.

[0013] (5) The bioinformatics analysis system described in (1) above, wherein, The determination unit determines arrhythmia based on the number of clusters in the clustering results, the number of heartbeat cycles belonging to each cluster, the distribution width information of the heartbeat cycles belonging to each cluster, and / or the time information of the heartbeat cycles belonging to each cluster.

[0014] (6) The bioinformatics analysis system according to any one of (1) to (3) above, wherein, The clustering unit has: A first clustering unit uses a probability distribution model related to the heartbeat cycles of the analysis unit to cluster the heartbeat cycles for the analysis unit; and The second clustering unit performs additional clustering when the clustering result is that two clusters exist in the heartbeat cycle. The bioinformatics analysis system also includes an arrhythmia reference information generation unit that generates arrhythmia reference information based on the additional clustering results. The second clustering unit uses the heartbeat cycle of the cluster with fewer heartbeat cycles belonging to the two clusters, as well as any one of the heartbeat cycles before and after that heartbeat cycle, to calculate the merged heartbeat cycle, and performs the additional clustering of the heartbeat cycle and the merged heartbeat cycle of the analysis unit using the probability distribution model. The determination unit makes a determination of arrhythmia based on the clustering results and the arrhythmia reference information.

[0015] (7) The bioinformatics analysis system according to any one of (1) to (3) above, wherein, It also includes an arrhythmia reference information generation unit, which generates arrhythmia reference information when the clustering result shows that there are three clusters in the heartbeat cycle. The arrhythmia reference information generation unit determines that the longest cluster contains the longest heartbeat cycle time information, and the shortest cluster contains the shortest heartbeat cycle time information. It then generates arrhythmia reference information based on the similarity between the number of heartbeat cycles belonging to the longest cluster and the number of heartbeat cycles belonging to the shortest cluster. The determination unit makes a determination of arrhythmia based on the clustering results and the arrhythmia reference information.

[0016] (8) The bioinformatics analysis system according to any one of (1) to (3) above, wherein, It also includes an arrhythmia reference information generation unit, which generates arrhythmia reference information when the clustering result indicates that there are three clusters in the heartbeat cycle. The arrhythmia reference information generation unit determines the longest cluster (where the heartbeat cycle time information belongs to the longest cluster), the shortest cluster (where the heartbeat cycle time information belongs to the shortest cluster), and the remaining intermediate clusters from the three clusters. It then calculates merged time information using the heartbeat cycle time information belonging to the longest cluster and the shortest cluster, and generates arrhythmia reference information based on the similarity between the merged time information and the heartbeat cycle time information belonging to the intermediate clusters. The determination unit determines the arrhythmia based on the clustering results and the arrhythmia reference information.

[0017] (9) The bioinformatics analysis system according to any one of (1) to (8) above, wherein, The heartbeat cycle of the analysis unit is baseline-corrected, and the heartbeat cycle is clustered for the analysis unit after baseline correction.

[0018] (10) A bioinformatics analysis program that uses the biological information of an object to determine arrhythmia, the bioinformatics analysis program causing a computer to perform the following series of processes: The interval of the object's heartbeat activity is calculated based on the biological information obtained from the object to serve as the heartbeat cycle; The heartbeat cycle is divided into segments by reading a pre-set segment length from the storage unit and using that segment length as the unit of analysis. Clustering of the heartbeat cycles for the analysis unit is performed using a probability distribution model associated with the heartbeat cycles of the analysis unit; and The determination of arrhythmia is based on the results of the clustering.

[0019] The effects of the invention According to the present invention, arrhythmias can be classified with high accuracy by performing clustering based on a probability distribution model on heartbeat cycles that exhibit characteristics of arrhythmias and by utilizing the results of such clustering. Attached Figure Description

[0020] Figure 1 This is a diagram illustrating a structural example of the bioinformatics analysis system according to Embodiment 1.

[0021] Figure 2 This is a diagram showing a detailed example of an electrocardiogram signal used for analysis.

[0022] Figure 3 This is a flowchart illustrating the arrhythmia determination process involved in Implementation 1.

[0023] Figure 4 This is a flowchart illustrating the model construction process involved in Implementation 1.

[0024] Figure 5A This is a diagram (one of) used to illustrate the model building process involved in Implementation Method 1.

[0025] Figure 5B This is a diagram (second one) used to illustrate the model building process involved in Implementation Method 1.

[0026] Figure 5C This is a diagram (third one) used to illustrate the model building process involved in Implementation Method 1.

[0027] Figure 6 This is a diagram illustrating an example of how the clustering department calculates the Bayesian information content standard.

[0028] Figure 7A This is a diagram (one of) illustrating an example of an electrocardiogram signal used in the decision processing of the decision unit.

[0029] Figure 7B This is a diagram (second one) illustrating an example of an electrocardiogram signal used in the decision processing of the decision unit.

[0030] Figure 7C This is a diagram (third one) illustrating an example of the electrocardiogram signal used in the decision processing of the decision unit.

[0031] Figure 8A This is a diagram (one of) used to illustrate an example of clustering results used in the decision processing of the decision unit.

[0032] Figure 8B This is a diagram (the second one) used to illustrate an example of clustering results used in the decision processing of the decision unit.

[0033] Figure 8CThis is a diagram (third one) used to illustrate an example of clustering results used in the decision processing of the decision unit.

[0034] Figure 9 This is a diagram illustrating a structural example of the bioinformatics analysis system according to Embodiment 2.

[0035] Figure 10 This is a flowchart illustrating the arrhythmia determination process involved in Implementation 2.

[0036] Figure 11 This is a flowchart illustrating the arrhythmia reference information generation process involved in Implementation 2.

[0037] Figure 12 This is a diagram illustrating a structural example of the bioinformatics analysis system according to Embodiment 3.

[0038] Figure 13 This is a flowchart illustrating the arrhythmia reference information generation process involved in Implementation 3.

[0039] Figure 14 This is a diagram illustrating a structural example of the bioinformatics analysis system according to Embodiment 4.

[0040] Figure 15 This is a flowchart illustrating the arrhythmia determination process involved in Implementation 4.

[0041] Figure 16 This is a flowchart illustrating the second clustering process involved in Implementation 4.

[0042] Figure 17A This is a diagram (one of) used to illustrate an example of the first clustering process of the first clustering unit.

[0043] Figure 17B This is a diagram (the second one) used to illustrate an example of the first clustering process in the first clustering section.

[0044] Figure 18A This is a diagram (one of) used to illustrate an example of the second clustering process in the second clustering section.

[0045] Figure 18B This is a diagram (the second one) illustrating an example of the second clustering process in the second clustering section.

[0046] Figure 19 This is a diagram illustrating an example of the distribution of the heartbeat cycle in atrial fibrillation.

[0047] Figure 20A This is a diagram (one) illustrating an example of the distribution of the heartbeat cycle during premature contractions.

[0048] Figure 20BThis is a diagram (the second one) showing an example of the distribution of the heartbeat cycle during premature contractions. Detailed Implementation

[0049] The embodiments of the bioinformatics analysis system according to the present invention will now be described in detail with reference to the accompanying drawings. However, the present invention is not limited to these embodiments. Furthermore, the various embodiments of the present invention are not independent, but can be combined and appropriately implemented separately.

[0050] (Implementation Method 1) Figure 1 This diagram illustrates a structural example of the bioinformatics analysis system according to Embodiment 1 of the present invention. The bioinformatics analysis system includes an electrocardiogram signal measuring device 2 as a bioinformatics measuring device, an analysis system 3, a receiving terminal 4, and a learning device 5.

[0051] The bioinformatics measuring device is arbitrarily determined based on the bioinformatics to be acquired and is not limited. Furthermore, it is assumed that the bioinformatics measuring device has a connector electrically connected to the analysis system 3, a communication device having a communication unit for communication with the bioinformatics analysis system, or an input port for writing the acquired bioinformatics. In this embodiment 1, an example will be described as follows: the bioinformatics is an electrocardiogram (ECG) signal 10, the bioinformatics measuring device is an ECG signal measuring device 2, the ECG signal measuring device 2 has an input port for writing the ECG signal 10, which is the acquired bioinformatics, and the subject 1 is a human. Furthermore, in this embodiment 1, an example of acquiring the ECG signal 10 to monitor the state of the subject 1 (whether it is arrhythmia) is described; however, besides the ECG signal 10, there are no particular limitations as long as the heartbeat cycle can be estimated, and it can also be a pulse wave signal, heart sounds, etc.

[0052] according to Figure 1 The electrocardiogram (ECG) signal 10, which is the subject of analysis, is obtained from the subject 1 via the ECG signal measuring device 2. Furthermore, the subject 1 can be a human, an animal, or other similar species, without particular limitation.

[0053] Additionally, the receiving terminal 4 may be a smartphone held by the subject 1 or an observer (e.g., a surgeon, a factory health manager, or a worker in the vicinity).

[0054] As an example of an electrocardiogram (ECG) signal measuring device 2, a wearable ECG machine is given. Specifically, the ECG signal measuring device 2 includes a wearable main body worn on a subject 1, multiple electrodes, and an ECG machine 100 electrically connected to each electrode. The ECG machine 100 and the like are fixed to the wearable main body using a strap or the like.

[0055] The electrocardiograph 100 is an example of a measuring device for acquiring electrocardiogram (ECG) signals 10. The ECG 100 has the functions of continuously acquiring ECG signals 10 from a subject 1, storing the acquired ECG signals 10, and transmitting data through communication with an analysis system 3. Alternatively, it can be configured such that the ECG 100 transmits data containing ECG signals 10 to a server device, and the server device transmits the ECG signals 10 to the analysis system 3. Hereinafter, the subject 1 will be described as a person wearing an ECG signal measuring device 2.

[0056] In this embodiment 1, the electrocardiogram (ECG) signal 10 obtained from the subject 1 includes information about the subject 1, the date and time of acquisition, the location, and the acquisition result. The acquisition result is waveform data 11 obtained by plotting the waveform with time on the horizontal axis and voltage on the vertical axis (see reference). Figure 1 ).

[0057] Figure 2 This is a diagram illustrating a detailed example of the electrocardiogram (ECG) signal 10 used for analysis. The ECG signal 10 may include waveform components called P wave 11a, Q wave 11b, R wave 11c, S wave 11d, and T wave 11e, the shapes of which are used as criteria for determining arrhythmias. For example, waveform data 11 is formed by repeating patterns of these waveform components. Furthermore, the shape of the waveform data 11 of the ECG signal 10 may change depending on the subject and the method of wearing the ECG machine 100.

[0058] Back Figure 1 The analysis system 3 includes a communication unit 311, a heartbeat cycle calculation unit 312, an analysis unit decision unit 313, a clustering unit 314, a judgment unit 315, an input / output unit 316, a control unit 317, and a storage unit 318.

[0059] The communication unit 311 can communicate with the electrocardiogram signal measuring device 2 and the receiving terminal 4 via a communication network. This communication network can be, for example, an existing public landline network, a LAN (Local Area Network), or a WAN (Wide Area Network), and can be either wired or wireless. The communication unit 311 may consist of, for example, a connector electrically connected to the communication object, a communication device having a communication unit with the electrocardiogram signal measuring device 2, or an input port for a medium storing data.

[0060] The heartbeat cycle calculation unit 312 detects the heartbeat from bio-information and calculates the heartbeat cycle as its interval.

[0061] The analysis unit 313 divides the heartbeat cycle calculated by the heartbeat cycle calculation unit 312 into predetermined segments and uses them as the analysis unit.

[0062] The clustering unit 314 selects a model for clustering, performs heartbeat cycle clustering using the selected model, and obtains the number of clusters and information about each cluster (the number of heartbeat cycles belonging to each cluster, distribution width information, time information, etc.). In this embodiment 1, the model used is the model generated by the learning device 5.

[0063] The determination unit 315 determines whether it is an arrhythmia based on the clustering results, which is the state of subject 1. At this time, if it is an arrhythmia, the determination unit 315 determines atrial fibrillation / premature contraction.

[0064] The input / output unit 316 is composed of devices with input and output functions, and outputs various information under the control of the control unit 317. The input / output unit 316 includes user interfaces such as a keyboard, mouse, and microphone, and has an input function that accepts the input of the electrocardiogram signal 10 output from the electrocardiogram signal measuring device 2. Furthermore, the input / output unit 316 has output functions such as a display made of liquid crystal or organic EL (electroluminescence), and a speaker for outputting sound.

[0065] Furthermore, as for the input function of the input / output unit 316, it is assumed that it has a biological information input mechanism, such as an input port for a medium on which biological information has been written. If the input function of the input / output unit 316 overlaps with the function of the communication unit 311, either one can have that function.

[0066] In addition, for the biological information input to the analysis system 3, in addition to using the previously acquired ECG signal 10, the acquisition of the ECG signal 10 by the ECG signal measuring device 2 and its input to the analysis system 3 can be carried out in parallel. In this case, it is assumed that the ECG signal measuring device 2 always sends the values ​​acquired in a continuous manner, or summarizes and sends the values ​​acquired in a batch manner at certain intervals, etc., without any particular limitation.

[0067] The control unit 317 uniformly controls the operation of the analysis system 3. Furthermore, the control unit 317 transmits the determination result of the determination unit 315, along with wearer information, to the receiving terminal 4 via the communication unit 311. Wearer information includes personal information such as the subject's name and gender, and the management number of the wearable device or peripheral device worn by the subject, used to identify the subject. Additionally, the control unit 317 can also output the determination result of the determination unit 315 through the output function of the input / output unit 316.

[0068] The storage unit 318 stores various data, including programs used to operate the analysis system 3 and data generated by each part (e.g., segment length). The various programs also include programs for arrhythmia determination processing performed using the learned model. The storage unit 318 is constructed using ROM (Read Only Memory) pre-installed with various programs, and RAM (Random Access Memory), HDD (Hard Disk Drive), SSD (Solid State Drive), etc., which store the operation parameters and data of each process.

[0069] Various programs can also be recorded on computer-readable recording media such as HDDs, flash memory, CD-ROMs, DVD-ROMs, and Blu-ray (registered trademark) for widespread distribution. Furthermore, the Communications Department 311 can also obtain various programs via communication networks.

[0070] The analysis system 3, with the above functional structure, is a computer constructed using one or more hardware components such as a CPU (Central Processing Unit), GPU (Graphics Processing Unit), ASIC (Application Specific Integrated Circuit), or FPGA (Field Programmable Gate Array). The analysis system 3 can be constructed separately from the electrocardiogram (ECG) signal measuring device 2, or it can be integrated with the ECG signal measuring device 2.

[0071] The receiving terminal 4 includes a communication unit 41, an output unit 42, and a control unit 43.

[0072] The communication unit 41 receives information output from the analysis system 3 via a communication network. For example, the communication unit 41 receives the determination result of the determination unit 315 and the wearer information from the analysis system 3.

[0073] The output unit 42 is composed of output devices such as a display device and a printing device, and outputs various information under the control of the control unit 43. The output unit 42 may have output functions such as a display made of liquid crystal or organic EL, and a speaker.

[0074] The control unit 43 uniformly controls the operation of the receiving terminal 4. In addition, the control unit 43 outputs the determination result of the determination unit 315 through the output function of the output unit 42.

[0075] The receiving terminal 4, which has the above functional structure, is a computer constructed using one or more hardware components such as a CPU, GPU, ASIC, or FPGA. Furthermore, the receiving terminal 4 may also have a user interface.

[0076] Next, we will explain the arrhythmia determination process based on the biological information obtained from subject 1. Figure 3 This is a flowchart illustrating the arrhythmia determination process according to Embodiment 1. In the arrhythmia determination process, firstly, the communication unit 311 or the input / output unit 316 acquires the electrocardiogram signal 10 obtained from the subject 1 (step S101). The control unit 317 processes the waveform data 11 of the acquired electrocardiogram signal 10 as needed to generate data for determination processing.

[0077] In this embodiment 1, the following case will be used as an example for explanation: The input / output unit 316 has an input port for a medium on which the electrocardiogram signal 10 is written. After the acquisition of the electrocardiogram signal 10 is completed, the electrocardiogram signal 10 stored in the medium is read.

[0078] After acquiring the electrocardiogram (ECG) signal 10, the heartbeat cycle calculation unit 312 detects the heartbeats from the ECG signal 10 and estimates the heartbeat cycle as the interval (step S102). Figure 2 As explained, the electrocardiogram signal 10 may include waveform components called P wave, Q wave, R wave, S wave, and T wave, which are detected by the heart's beating. The heartbeat cycle calculation unit 312 captures the timing of the heart's beating by detecting these waveform components (especially the R wave) and calculates the estimated heartbeat cycle based on their intervals.

[0079] After estimating the heartbeat cycle, the analysis unit determination unit 313 determines the analysis unit by dividing the heartbeat cycle into segments of a predetermined length (step S103). The analysis unit determination unit 313 divides the calculated heartbeat cycle into segments of a predetermined length and sets them as analysis units. As for the unit of segments, it is assumed that the time unit is divided into segments of a predetermined length, or the heartbeat unit is divided into segments each time a certain number of heartbeats are detected, but there are no particular limitations. However, in relation to the determination of atrial fibrillation, it is desirable to set the time unit to 30 [sec] or more, and to set the heartbeat cycle calculated according to the beat unit of that time unit as the analysis unit. In addition, overlap can be freely set in the analysis units. In this embodiment 1, the following case will be described as an example: the analysis unit determination unit 313 does not set overlap, but divides the heartbeat cycle into segments of 60 [sec] and sets them as analysis units.

[0080] Then, the clustering unit 314 determines the model to be used for clustering (step S104).

[0081] Here, the model determined by the clustering unit 314 will be described. This model is constructed by the learning device 5, specifically a probability distribution model. Regarding the probability distribution model, a Gaussian Mixture Model (GMM) or a Poisson Mixture Model (PMM) is assumed, but there is no particular limitation. In this embodiment 1, the Gaussian Mixture Model (GMM) will be used as an example for explanation.

[0082] Figure 4 This is a flowchart illustrating the model construction process involved in Embodiment 1.

[0083] Before constructing the model, the learning device 5 sets the number of clusters to be classified as the hyperparameter of the Gaussian mixture distribution model (step S201). In this embodiment 1, an example is given with any number of clusters to be classified from 1 to 3, but there is no particular limitation. This means that the heartbeat cycles in the analysis unit are classified with any number of clusters from 1 to 3, and an independent model is constructed with each number of clusters. Hereinafter, the Gaussian mixture distribution model with a cluster number of 1 is described as model (1), the Gaussian mixture distribution model with a cluster number of 2 is described as model (2), and the Gaussian mixture distribution model with a cluster number of 3 is described as model (3).

[0084] After setting the number of clusters, the learning device 5 sets the initial parameters of the model (step S202). Specifically, the learning device 5 sets the initial values ​​of the parameters (mean, covariance matrix, and mixing coefficients) of the Gaussian distribution of each cluster number. The method for determining the initial values ​​of the parameters is not particularly limited; it is assumed that a predetermined value is used, a random value is used, or a value obtained by a clustering method such as k-means is used.

[0085] After initial setup, the learning device 5 optimizes the model using the EM (Expectation-Maximization) algorithm (step S203). The EM algorithm is a method for maximum likelihood estimation related to the parameters of a probability distribution model. It consists of an expectation step (E-Step) to calculate the probability (likelihood) of each data point within a Gaussian distribution in each cluster, and a maximization step (M-Step) to re-estimate the parameters of the Gaussian distribution in each cluster based on the calculated likelihood. In the EM algorithm, the E-step and M-step iterate until a convergence criterion is met. Examples of convergence criterion include the change in parameters or likelihood within each iteration being less than a threshold, or the number of iterations being greater than a threshold, but there are no specific limitations.

[0086] Figures 5A-5C This is a diagram used to illustrate the model construction process involved in Embodiment 1. Figure 5A The Gaussian distribution in model (1) is shown. Figure 5B The Gaussian distribution in model (2) is shown. Figure 5C The Gaussian distribution in model (3) is shown. From Figures 5A-5C As can be seen, the Gaussian distribution is optimized based on the number of clusters.

[0087] According to the process described above, the model construction of learning device 5 is completed.

[0088] Back Figure 3 The clustering unit 314 determines the model to be used for clustering from the models constructed by the learning device 5. Specifically, in this embodiment 1, three models (1) to (3) are prepared, and the clustering unit 314 selects the model to be used for clustering from these models. As a method for selecting the model, it is assumed that in addition to human selection, selection is based on information criteria such as Bayesian Information Criterion and Akaike's Information Criterion, but there are no particular limitations. In this embodiment 1, the following case will be used as an example: the model is selected based on the condition of minimizing the Bayesian Information Criterion (hereinafter referred to as BIC), which is one of the information criteria. BIC can evaluate the applicability to the data through the likelihood function. By selecting the model with the minimum BIC, the optimal model can be selected in many cases.

[0089] Figure 6 This is a diagram used to illustrate an example of BIC calculation performed by the clustering department. Figure 6 It shows Figures 5A-5C Examples of BIC calculations in each model are shown. Figure 6 In model (3), the BIC is smaller than that of other models, so it can be judged as the optimal model and selected as the model for clustering, thus determining the number of clusters to be classified.

[0090] Based on the above, the model decision processing performed by the clustering unit 314 is completed.

[0091] Back Figure 3 The clustering unit 314 performs clustering processing of heartbeat cycles using the determined model (step S105). As a result of the clustering processing, the clustering unit 314 obtains the number of clusters and information about each cluster (the number of heartbeat cycles belonging to each cluster, distribution width information, time information, etc.).

[0092] After clustering, the determination unit 315 performs a determination process based on the clustering results to determine whether it is an arrhythmia, which is the state of subject 1 (step S106). More specifically, the determination unit 315 determines whether it is atrial fibrillation, premature contractions, or something else.

[0093] Figures 7A-7C This is a diagram illustrating an example of the electrocardiogram signal 10 used in the determination process of the determination unit. Figures 8A-8C This is a diagram illustrating an example of clustering results used in the decision processing of the decision unit.

[0094] Figure 7A The ECG signal of 10A without arrhythmia is shown. Additionally, Figure 8A An example of the clustering results of heartbeat cycles in an ECG signal 10A is shown. In the absence of arrhythmia in the ECG signal 10A, the heartbeat cycle is constant, and therefore, as a result of clustering, it becomes a cluster with small bias.

[0095] Figure 7B The ECG signal 10B shows an atrial fibrillation. Additionally, Figure 8B This illustrates an example of clustering results for heartbeat cycles in ECG signal 10B. In the case of ECG signal 10B with atrial fibrillation, the heartbeat cycles become irregular, resulting in a cluster with high bias as a clustering outcome.

[0096] Figure 7C The ECG signal 10C shows a ventricular extrasystoles. Additionally, Figure 8C This example illustrates the clustering results of cardiac cycles in ECG signal 10C. In the case of ECG signal 10C with ventricular premature contractions, the cardiac cycle before the ventricular premature contraction is short, and the cardiac cycle after the ventricular premature contraction is long. Therefore, as a result of clustering, it forms three clusters with small bias.

[0097] like Figures 7A-7C , Figures 8A-8C As shown, the number of clusters and the information of each cluster (the number of heartbeat cycles belonging to each cluster, distribution width information, time information, etc.) exhibit characteristics in atrial fibrillation and premature contractions, thus enabling the determination of arrhythmias based on these characteristics.

[0098] In this embodiment 1, for example, if the number of clusters is any one of 1 to 3, and the standard deviation of the cluster with the most heartbeat cycles is 0.064 [sec] or more, the determination unit 315 determines it as "atrial fibrillation". Furthermore, if the number of clusters is 2 or 3, and the standard deviation of each cluster is less than 0.064 [sec], the determination unit 315 determines it as "premature contractions". In this case, Figure 7A Among the ECG signals without arrhythmia shown in the diagram, 10A is classified as "other". Figure 7B In the ECG signal 10B shown, which indicates atrial fibrillation, it is identified as "atrial fibrillation". Figure 7C The ECG signal 10C shown, which indicates an early ventricular contraction, is identified as an "early contraction".

[0099] Then, the control unit 317 sends information related to the determination result of the determination unit 315 and the wearer's information to the receiving terminal 4 (step S107). The receiving terminal 4 displays information about the state of the subject 1 (whether it is arrhythmia). For example, the observer holding the receiving terminal 4 confirms the determination result.

[0100] According to the above-described Embodiment 1, it is possible to perform clustering based on a probability distribution model on heartbeat cycles that exhibit arrhythmia characteristics, and to classify arrhythmias with high accuracy using the results of this clustering.

[0101] According to this embodiment 1, the number of clusters and the information of each cluster (the number of heartbeat cycles belonging to each cluster, distribution width information, time information, etc.) can be obtained as the result of clustering based on the probability distribution model, and therefore they can be used to classify arrhythmias with high accuracy.

[0102] Furthermore, according to Embodiment 1, high-precision clustering based on a probability distribution model can be performed, thus enabling high-precision classification of extra-term contractions in which the heartbeat cycle is divided into multiple clusters.

[0103] In this embodiment 1, a learning device 5 separate from the analysis system 3 is provided to build the model. However, the analysis system 3 can also be configured to have the function of the learning device 5. Alternatively, the model building process in the arrhythmia determination process can be skipped by directly reading the learned model.

[0104] Furthermore, in this embodiment 1, the heartbeat cycles estimated from the electrocardiogram signal 10 are clustered as is, but data processing can also be performed through other methods. One example of data processing is baseline correction. In baseline correction, for example, the clustering unit 314 removes the baseline in the heartbeat cycles using methods such as moving averages or cutoff in the frequency space. This removes heartbeat variations caused by exercise or mental excitement, thus enabling highly accurate classification of arrhythmias. Furthermore, assuming the above data processing is performed by the clustering unit 314, it can also be performed by the analysis unit decision unit 313, the control unit 317, etc., without limitation.

[0105] (Implementation Method 2) Next, Embodiment 2 of the present invention will be described. Embodiment 2 includes an analysis system 3A for acquiring location information, replacing the analysis system 3 described in Embodiment 1. Hereinafter, the same reference numerals will be used to denote the same components as in Embodiment 1, and the different parts will be described.

[0106] Figure 9 This diagram illustrates a structural example of the bioinformatics analysis system according to Embodiment 2. The bioinformatics analysis system according to Embodiment 2 includes an electrocardiogram signal measuring device 2, an analysis system 3A, a receiving terminal 4, and a learning device 5.

[0107] The analysis system 3A includes a communication unit 311, a heart rate cycle calculation unit 312, an analysis unit decision unit 313, a clustering unit 314, a judgment unit 315, a first arrhythmia reference information generation unit 319, an input / output unit 316, a control unit 317, and a storage unit 318.

[0108] The first arrhythmia reference information generation unit 319 performs a determination process based on the number of heartbeat cycles belonging to each cluster, and generates arrhythmia reference information as a parameter indicating whether an extra-term contraction can be determined.

[0109] Next, the arrhythmia determination process in this embodiment 2 will be explained. Figure 10 This is a flowchart illustrating the arrhythmia determination process according to Embodiment 2. In the arrhythmia determination process according to Embodiment 2, steps S101 to S105 of Embodiment 1 described above (see...) Figure 3 Similarly, after estimating the heartbeat cycle and determining the analysis unit based on the obtained electrocardiogram signal 10, the heartbeat cycle is clustered using the model (steps S301~S305).

[0110] In this embodiment 2, after clustering, the control unit 317 determines whether the result of the clustering process is that the number of clusters is three (step S306). If the control unit 317 determines that the number of clusters is not three (step S306: No), it proceeds to step S308. Conversely, if the control unit 317 determines that the number of clusters is three (step S306: Yes), it proceeds to step S307.

[0111] In step S307, the first arrhythmia reference information generation unit 319 generates arrhythmia reference information. When the number of clusters in the clustering result is three, the first arrhythmia reference information generation unit 319 generates arrhythmia reference information as a parameter indicating whether an extra-term contraction can be determined, based on the number of heartbeat cycles belonging to each cluster.

[0112] Figure 11This is a flowchart illustrating the arrhythmia reference information generation process involved in Embodiment 2.

[0113] First, the first arrhythmia reference information generation unit 319 determines the longest cluster (hereinafter referred to as "cluster Long") and the shortest cluster (hereinafter referred to as "cluster Short") among the three clusters obtained by clustering, which have the longest time information of the heartbeat cycle. (Step S401)

[0114] In this case, within the analysis unit, in the case of compensatory premature contraction during premature contraction, the shortened heart rate cycle before the premature contraction is considered to belong to the cluster Short, and the lengthened heart rate cycle after the premature contraction is considered to belong to the cluster Long.

[0115] The time information mentioned here could be, for example, the average heartbeat cycle of each cluster.

[0116] Next, the first arrhythmia reference information generation unit 319 acquires the number of heartbeat cycles belonging to the Long and Short clusters respectively, and evaluates the similarity of the number of heartbeat cycles belonging to the two clusters (step S402). In this embodiment 2, it is assumed that relative error is used as a method for evaluating the similarity of the number of heartbeat cycles belonging to the two clusters, but there is no particular limitation. The relative error can be obtained, for example, by dividing the absolute value of the difference between two values ​​by the larger of the two values. In this embodiment 2, it is assumed that the relative error is obtained by dividing the absolute value of the difference between two values ​​by the larger of the two values.

[0117] Then, the first arrhythmia reference information generation unit 319 generates arrhythmia reference information based on the similarity of the number of heartbeat cycles belonging to the two clusters obtained from the evaluation (step S403). Specifically, the arrhythmia reference information is a parameter indicating whether an analysis unit is a candidate for premature contraction. In this embodiment 2, it is assumed that a value with 2 is set to "1" if it can be determined to be a candidate for premature contraction, and set to "0" if it cannot be determined to be a candidate for premature contraction, but there is no particular limitation.

[0118] In this embodiment 2, for example, the first arrhythmia reference information generation unit 319 determines the number of heartbeat cycles belonging to the Long cluster to be 0 (i.e., "not a candidate for premature contraction") if the relative error between the number of heartbeat cycles belonging to the Short cluster is 0.1 or more, and determines the number of heartbeat cycles belonging to the Short cluster to be 1 (i.e., "a candidate for premature contraction") if the error is less than 0.1. This is because compensatory premature contractions are arrhythmias that occur prematurely outside the normal rhythm. Therefore, assuming that the analysis unit is premature contractions, it can be determined that the heartbeat cycles belonging to the Long cluster are calculated based on the waveform of the premature contraction and the subsequent waveform, and the heartbeat cycles belonging to the Short cluster are calculated based on the waveform of the premature contraction and the previous waveform. Furthermore, it can be determined that the number of heartbeat cycles belonging to these clusters is similar to the number of premature contractions occurring in the analysis unit.

[0119] Furthermore, if the control unit 317 determines that the number of clusters is not three (step S306: No), the first arrhythmia reference information generation unit 319 can also generate a predetermined value as arrhythmia reference information. For example, in this embodiment 2, if the number of clusters in the clustering result is not three, the first arrhythmia reference information generation unit 319 generates a "missing value" as arrhythmia reference information.

[0120] Based on the above, the arrhythmia reference information generation and processing of the first arrhythmia reference information generation unit 319 is completed.

[0121] Back Figure 10 In step S308, after generating arrhythmia reference information, the determination unit 315 performs a determination process based on the clustering results to determine whether it is an arrhythmia, which is the state of the subject 1. Specifically, the determination unit 315 determines whether it is atrial fibrillation, premature contractions, or something else.

[0122] Then, the control unit 317 sends information related to the determination result of the determination unit 315 and the wearer's information to the receiving terminal 4 (step S309). The receiving terminal 4 displays information about the state of the subject 1 (whether it is arrhythmia). For example, the observer holding the receiving terminal 4 confirms the determination result.

[0123] According to the above description of Embodiment 2, similar to Embodiment 1, it is possible to classify arrhythmias with high accuracy by performing clustering based on a probability distribution model on heartbeat cycles that exhibit characteristics of arrhythmias and using the results of this clustering.

[0124] Furthermore, in this embodiment 2, a bioinformatics analysis system can be provided in which, when the number of clusters in the clustering results is three, the first arrhythmia reference information generation unit 319 generates arrhythmia reference information based on the number of heartbeat cycles belonging to each cluster, thereby enabling high-precision classification of arrhythmias.

[0125] Furthermore, in this embodiment 2, the cluster information of cluster Long and cluster Short is used as is, but data processing can also be implemented by other methods with reference to embodiment 1.

[0126] As one example of other data processing, intermittent peaks are excluded.

[0127] In premature contractions, in addition to compensatory peaks where heartbeats occur ahead of the normal rhythm, there are also intermittent peaks where heartbeats occur intermittently with the normal rhythm. When intermittent peaks occur, the number of heartbeat cycles belonging to the Long cluster does not increase, only the number of heartbeat cycles belonging to the Short cluster increases. Therefore, the number of heartbeat cycles belonging to both the Long and Short clusters is no longer similar to premature contractions occurring in the analysis unit. Thus, in the arrhythmia determination processing according to Embodiment 2, premature contractions are not identified, which is a concern as it reduces the classification accuracy of arrhythmias. As a method for excluding intermittent peaks, the following situations are assumed, but not particularly limited: if there are two or more consecutive heartbeat cycles belonging to the Short cluster in the time series of the analysis unit, they are considered as intermittent peaks and excluded from the determination objects in advance, and the determination processing of the first arrhythmia reference information generation unit 319 is performed. Thus, in the first arrhythmia reference information generation unit 319, only compensatory peaks can be targeted, thus enabling high-precision classification of arrhythmias.

[0128] (Implementation Method 3) Next, Embodiment 3 of the present invention will be described. Embodiment 3 includes an analysis system 3B for acquiring location information, replacing the analysis system 3 described in Embodiment 1. Hereinafter, the same reference numerals will be used to denote the same components as in Embodiment 1, and the different parts will be described.

[0129] Figure 12 This diagram illustrates a structural example of the bioinformatics analysis system according to Embodiment 3. The bioinformatics analysis system according to Embodiment 3 includes an electrocardiogram signal measuring device 2, an analysis system 3B, a receiving terminal 4, and a learning device 5.

[0130] The analysis system 3B includes a communication unit 311, a heart rate cycle calculation unit 312, an analysis unit decision unit 313, a clustering unit 314, a judgment unit 315, a second arrhythmia reference information generation unit 320, an input / output unit 316, a control unit 317, and a storage unit 318.

[0131] When the number of clusters in the clustering result is three, the second arrhythmia reference information generation unit 320 performs a determination process based on the time information of the heartbeat cycle belonging to each cluster, and generates arrhythmia reference information as a parameter indicating whether an extra-term contraction can be determined.

[0132] according to Figure 10 The flowchart shown executes the arrhythmia determination process in Embodiment 3. Here, in Embodiment 3, in step S307, the second arrhythmia reference information generation unit 320... Figure 13 The flowchart shown generates reference information for cardiac arrhythmias.

[0133] Figure 13 This is a flowchart illustrating the arrhythmia reference information generation process involved in Embodiment 3.

[0134] First, the second arrhythmia reference information generation unit 320 determines the longest cluster (hereinafter referred to as cluster Long) among the three clusters, the shortest cluster (hereinafter referred to as cluster Short) whose time information of the heartbeat cycle is the longest, and the remaining cluster (hereinafter referred to as cluster Center) (step S501).

[0135] The time information referred to here is, for example, the mean of the heartbeat cycles belonging to each cluster. In this context, within the unit of analysis, compensatory premature contractions are considered to occur within premature contractions; the shortened heartbeat cycle before a premature contraction belongs to the cluster Short; the longer heartbeat cycle after a premature contraction belongs to the cluster Long; and the normal heartbeat cycle unrelated to premature contractions belongs to the cluster Center.

[0136] Next, the second arrhythmia reference information generation unit 320 calculates the merged time information based on the cluster Long and the cluster Short (step S502). Regarding the merged time information, a value representing the relationship between the two clusters is calculated based on the time information of the heartbeat cycles belonging to the cluster Long and the time information of the heartbeat cycles belonging to the cluster Short. In this embodiment 3, it is assumed that the average of the time information of the heartbeat cycles belonging to the cluster Long and the heartbeat cycles belonging to the cluster Short is used as the merged time information, but there is no particular limitation.

[0137] Next, the second arrhythmia reference information generation unit 320 evaluates the similarity between the merged time information and the time information of the heartbeat cycles belonging to the cluster Center (step S503). In this embodiment 3, as a method for evaluating similarity, it is assumed that a relative error is used, but there are no particular limitations. The relative error can be obtained, for example, by dividing the absolute value of the difference between two values ​​by the larger of the two values. In this embodiment 3, it is assumed that the relative error is obtained by dividing the absolute value of the difference between two values ​​by the larger of the two values.

[0138] Then, the second arrhythmia reference information generation unit 320 generates arrhythmia reference information as a parameter indicating whether an extra-term contraction can be determined based on the similarity between the evaluated merged time information and the time information of the heartbeat cycle belonging to the cluster Center (step S504). Specifically, the arrhythmia reference information is a parameter indicating whether an extra-term contraction candidate is an analytical unit. In this embodiment 3, it is assumed that a value with 2 is set to "1" if an extra-term contraction candidate can be determined, and set to "0" if an extra-term contraction candidate cannot be determined, but there is no particular limitation.

[0139] In this embodiment 3, for example, the second arrhythmia reference information generation unit 320 determines "0" (i.e., "not a candidate for premature contraction") when the relative error between the merged time information and the time information of the heartbeat cycle belonging to the cluster Center is 0.2 or more, and determines "1" (i.e., "a candidate for premature contraction") when the error is less than 0.2. This is because compensatory premature contractions are arrhythmias that occur prematurely outside the normal rhythm. Therefore, assuming that the analysis object is premature contraction, it can be determined that the average heartbeat cycle of the cluster with the longest heartbeat cycle time information among the three clusters is similar to the heartbeat cycle of the cluster with the shortest heartbeat cycle time information among the three clusters, and is also similar to the heartbeat cycle of the remaining clusters among the three clusters, i.e., the clusters representing normal rhythms.

[0140] Furthermore, when the control unit 317 determines that the number of clusters is not three (step S306: No), the second arrhythmia reference information generation unit 320 can also generate a predetermined value as arrhythmia reference information. For example, in this embodiment 3, when the number of clusters in the clustering result is not three, the second arrhythmia reference information generation unit 320 generates a "missing value" as arrhythmia reference information.

[0141] Based on the above, the arrhythmia reference information generation and processing of the second arrhythmia reference information generation unit 320 is completed.

[0142] The determination unit 315 performs a determination process based on the clustering results to determine whether it is an arrhythmia, which is the state of the subject 1. In this embodiment 3, for example, if the number of clusters is any one of 1 to 3 and the standard deviation of the cluster with the most heartbeat cycles is 0.064 [sec] or more, the determination unit 315 determines it as "atrial fibrillation". If the number of clusters is 2 and the standard deviation of each cluster is less than 0.064 [sec], or if the number of clusters is 3 and the standard deviation of each cluster is less than 0.064 [sec] and the arrhythmia reference information is "1", the determination unit 315 determines it as "premature contraction".

[0143] According to the above description of Embodiment 3, similar to Embodiment 1, it is possible to perform clustering based on a probability distribution model on the heartbeat cycles that exhibit characteristics of arrhythmia, and use the results of this clustering to classify arrhythmias with high accuracy.

[0144] Furthermore, in this embodiment 3, a bioinformatics analysis system can be provided in which, when the number of clusters in the clustering results is three, the second arrhythmia reference information generation unit 320 generates arrhythmia reference information based on the time information of the heartbeat cycle belonging to each cluster, and arrhythmias can be classified with high accuracy.

[0145] Furthermore, in this embodiment 3, the information of each cluster is used as is, but data processing can also be implemented through other processes, referring to embodiments 1 and 2.

[0146] (Implementation Method 4) Next, Embodiment 4 of the present invention will be described. Embodiment 4 includes an analysis system 3C for acquiring location information, replacing the analysis system 3 described in Embodiment 1. Hereinafter, the same reference numerals will be used to denote the same constituent elements as in Embodiment 1, and the different parts will be described.

[0147] Figure 12 This diagram illustrates a structural example of the bioinformatics analysis system according to Embodiment 4. The bioinformatics analysis system according to Embodiment 4 includes an electrocardiogram signal measuring device 2, an analysis system 3C, a receiving terminal 4, and a learning device 5.

[0148] The analysis system 3C includes a communication unit 311, a heart rate cycle calculation unit 312, an analysis unit decision unit 313, a clustering unit 314A, a judgment unit 315, a third arrhythmia reference information generation unit 321, an input / output unit 316, a control unit 317, and a storage unit 318.

[0149] The clustering unit 314A has a first clustering unit 314a and a second clustering unit 314b.

[0150] The first clustering unit 314a, like the clustering unit 314, selects a model for clustering and performs heartbeat cycle clustering using the selected model to obtain the number of clusters and information about each cluster.

[0151] If the number of clusters in the clustering result of the first clustering unit 314a is two, the second clustering unit 314b performs additional clustering processing using the heartbeat cycle in the analysis unit.

[0152] The third arrhythmia reference information generation unit 321 generates arrhythmia reference information based on the additional clustering results of the second clustering unit 314b.

[0153] Next, the arrhythmia determination process in this embodiment 4 will be explained. Figure 15 This is a flowchart illustrating the arrhythmia determination process according to Embodiment 4. In the arrhythmia determination process according to Embodiment 4, steps S101 to S105 of Embodiment 1 described above (see...) Figure 3 Similarly, after estimating the heartbeat cycle and determining the analysis unit based on the acquired electrocardiogram signal 10, the first clustering unit 314a performs clustering processing of the heartbeat cycle (first clustering processing) through the model (steps S601 to S605).

[0154] In this embodiment 4, after the clustering process, the control unit 317 determines whether the result of the first clustering process is that the number of clusters is two (step S606). If the control unit 317 determines that the number of clusters is not two (step S606: No), the process proceeds to step S609. Conversely, if the control unit 317 determines that the number of clusters is two (step S606: Yes), the process proceeds to step S607.

[0155] In step S607, the second clustering unit 314b performs additional clustering processing (second clustering processing). Figure 16 This is a flowchart illustrating the second clustering process involved in Embodiment 4.

[0156] First, the second clustering unit 314b detects the heartbeat cycles of the clusters with fewer heartbeat cycles belonging to the two clusters obtained by the first clustering unit 314a (step S701).

[0157] Next, the second clustering unit 314b replaces the detected heartbeat cycle with a merged heartbeat cycle in the time series of the analysis unit, for both the detected heartbeat cycle and any heartbeat cycle before and after the heartbeat cycle (step S702).

[0158] At this point, decide which heartbeat cycle to use, either the one before or after the detected heartbeat cycle, as shown below.

[0159] First, if the time information of the heartbeat cycle belonging to a cluster with fewer heartbeat cycles is shorter than that of another cluster, then the heartbeat cycle after the detected heartbeat cycle is used.

[0160] Conversely, if it is longer than another cluster, the heartbeat cycle preceding the detected heartbeat cycle is used.

[0161] The time information mentioned here could be, for example, the average heartbeat cycle of each cluster.

[0162] Furthermore, in this embodiment 4, it is assumed that the heartbeat cycle is merged using the average of the detected heartbeat cycle and any heartbeat cycle before and after the detected heartbeat cycle, but there is no particular limitation.

[0163] The two heartbeat cycles used for calculation are removed from the analysis unit, and in their place, the calculated combined heartbeat cycle is added to the analysis unit, thus replacing the heartbeat cycle.

[0164] Perform the above operation on all heartbeat cycles belonging to the cluster with the fewest detected heartbeat cycles.

[0165] Next, the second clustering unit 314b sets the cluster number to 1 for the heartbeat cycle in the replaced analysis unit to construct the model (step S703). Regarding step S703, in the model construction process of this embodiment 4, specifically, in step S201 of the above embodiment 1 (refer to...) Figure 4 In S202~S203 (refer to...), the number of clusters to be classified is set to 1. Figure 4 Similarly, the model is constructed and processed.

[0166] Then, the second clustering unit 314b performs clustering of heartbeat cycles using the constructed model, and obtains cluster information (number of heartbeat cycles belonging to each cluster, distribution width information, time information, etc.) as the second clustering result (step S704). The distribution width information mentioned here is, for example, the standard deviation in the distribution of each cluster, and the time information is, for example, the mean of the heartbeat cycles belonging to each cluster.

[0167] Based on the above, the second clustering process of the second clustering unit 314b is completed.

[0168] Back Figure 10After the second clustering process, the third arrhythmia reference information generation unit 321 generates arrhythmia reference information based on the result of the second clustering process (step S608). Specifically, the arrhythmia reference information is a parameter indicating whether an analysis unit is a candidate for extra-term contraction. In this embodiment 4, it is assumed that a value with a value of 2 is set to "1" if it can be determined to be a candidate for extra-term contraction, and set to "0" if it cannot be determined to be a candidate for extra-term contraction, but there is no particular limitation. In this embodiment 4, for example, the third arrhythmia reference information generation unit 321 determines that if the standard deviation of the cluster is 0.064 [sec] or more, that is, "not a candidate for extra-term contraction", and determines that if it is less than 0.064 [sec], that is, "is a candidate for extra-term contraction".

[0169] Reference Figures 17A-18B Explain why this judgment can be made. Figure 17A and Figure 17B This is a diagram illustrating an example of the first clustering process of the first clustering section 314a. Figure 18A and Figure 18B This is a diagram illustrating an example of the second clustering process in the second clustering section.

[0170] Figure 17A The results of clustering (first clustering processing) of the heartbeat cycles performed by the first clustering unit 314a in the electrocardiogram signal 10D (not shown) in which atrial fibrillation occurred are shown. Figure 18A The results of the second clustering process of the heartbeat cycle performed by the second clustering unit 314b are shown in the electrocardiogram signal 10D in which atrial fibrillation has occurred.

[0171] Figure 17B The results of clustering (first clustering processing) of the heartbeat cycle performed by the first clustering unit 314a in the electrocardiogram signal 10E (not shown) in which ventricular extrasystoles occurred are shown. Figure 18B The results of the second clustering process of the cardiac cycle performed by the second clustering unit 314b are shown in the electrocardiogram signal 10E in which ventricular extrasystole occurred.

[0172] exist Figure 17BIn the ECG signal 10E showing a ventricular premature contraction, the clustering result of the heartbeat cycle performed by the first clustering unit 314a should generally be 3. However, the cluster Long, belonging to the prolonged heartbeat cycle after the premature contraction, overlaps with the cluster Center, belonging to the normal heartbeat cycle unrelated to the premature contraction, forming a single cluster (hereinafter referred to as "cluster LC"), thus the cluster number is 2. As such a result of the clustering of heartbeat cycles performed by the first clustering unit 314a, the situation of clusters overlapping in the premature contraction forming a single cluster is anticipated to occur when the learning device 5 constructs a model, or when the first clustering unit 314a does not select a sufficient model, etc., which could lead to misdiagnosis as atrial fibrillation, etc.

[0173] Here, as Figure 18B As shown, when the second clustering unit 314b processes the ECG signal 10E in which ventricular extrasystoles occur, the deviation of the clusters replaced by the combined heartbeat cycles becomes smaller. On the other hand, as Figure 18A As shown, when the second clustering unit 314b processes the ECG signal 10D in which atrial fibrillation has occurred, the bias of the clusters that replace the combined heartbeat cycles increases. This is because the compensatory premature contraction in premature contractions is an arrhythmia that occurs prematurely outside the normal rhythm. Therefore, in the time series of the analysis unit, the shortened heartbeat cycle before the premature contraction is adjacent to the longer heartbeat cycle after the premature contraction. Thus, the elements of cluster L hidden in cluster LC can be determined based on the cluster Short to which the shortened heartbeat cycle before the premature contraction belongs, and its mean is obtained based on cluster C, which can be judged to be similar to the normal heartbeat cycle that is unrelated to the premature contraction. For example, in the results of the additional clustering of heartbeat cycles in the second clustering unit 314b, if the standard deviation of the cluster is less than 0.064 [sec], it can be judged as a "candidate premature contraction".

[0174] Furthermore, if the control unit 317 determines that the number of clusters is not two (step S606: No), the third arrhythmia reference information generation unit 321 can also generate a predetermined value as arrhythmia reference information. In this embodiment 4, if the number of clusters in the clustering result is not two, the third arrhythmia reference information generation unit 321 generates a "missing value" as arrhythmia reference information.

[0175] Based on the above, the arrhythmia reference information generation and processing of the third arrhythmia reference information generation unit 321 is completed.

[0176] In step S609, the determination unit 315 performs a determination process based on the clustering results to determine whether it is an arrhythmia, which is the state of the subject 1. Specifically, the determination unit 315 determines whether it is atrial fibrillation, premature contractions, or something else. In this embodiment 4, for example, if the number of clusters is 1 or 3 and the standard deviation of the cluster with the most heartbeat cycles is 0.064 [sec] or more, or if the number of clusters is 2 and the standard deviation of the cluster with the most heartbeat cycles is 0.064 [sec] or more, and the arrhythmia reference information is 0, the determination unit 315 determines it as "atrial fibrillation". If the number of clusters is 2 and the standard deviation of each cluster is less than 0.064 [sec] and the arrhythmia reference information is 1, or if the number of clusters is 3 and the standard deviation of each cluster is less than 0.064 [sec], the determination unit 315 determines it as "premature contractions".

[0177] Then, the control unit 317 sends information related to the determination result of the determination unit 315 and the wearer information to the receiving terminal 4 (step S610). The receiving terminal 4 displays information about the status of the subject 1 (whether it is arrhythmia). For example, the observer holding the receiving terminal 4 confirms the determination result.

[0178] According to the above description of Embodiment 4, similar to Embodiment 1, it is possible to perform clustering based on a probability distribution model on the heartbeat cycles that exhibit characteristics of arrhythmia, and use the results of this clustering to classify arrhythmias with high accuracy.

[0179] Furthermore, according to this embodiment 4, a bioinformatics analysis system can be provided in which the second clustering unit 314b and the third arrhythmia reference information generation unit 321 can generate arrhythmia reference information based on the information of each cluster, even when the clustering accuracy of the first clustering unit 314a is insufficient, thereby enabling high-precision classification of arrhythmias.

[0180] Furthermore, in this embodiment 4, the information of each cluster is used as is, but data processing can also be implemented through other processes, referring to embodiments 1 and 2.

[0181] (Other implementation methods) The embodiments for carrying out the present invention have been described up to this point, but the present invention should not be limited to the embodiments described above. For example, the analysis system 3, 3A to 3C and the electrocardiogram signal measuring device 2 have been described separately, but the analysis system 3, 3A to 3C may also be integrated with the electrocardiogram signal measuring device 2.

[0182] Symbol Explanation 1: Subject; 2: ECG signal measuring device; 3, 3A~3C: Analysis system; 4: Receiving terminal; 5: Learning device; 10, 10A~10E: ECG signal; 11: Waveform data; 41, 311: Communication unit; 42: Output unit; 43, 317: Control unit; 312: Heart rate cycle calculation unit; 313: Analysis unit determination unit; 314, 314A: Clustering unit; 314a: First clustering unit; 314b: Second clustering unit; 315: Judgment unit; 316: Input / output unit; 318: Storage unit; 319: First arrhythmia reference information generation unit; 320: Second arrhythmia reference information generation unit; 321: Third arrhythmia reference information generation unit.

Claims

1. A bioinformatics analysis system that uses the biological information of an object to determine cardiac arrhythmia, said bioinformatics analysis system comprising: The heartbeat cycle calculation unit calculates the interval of the object's heartbeat activity as the heartbeat cycle based on the biological information obtained from the object; The analysis unit determines the segment length as the analysis unit to divide the heartbeat cycle; The clustering unit uses a probability distribution model related to the heartbeat cycle of the analysis unit to cluster the heartbeat cycles for the analysis unit. as well as The determination unit determines arrhythmias based on the clustering results.

2. The bioinformatics analysis system according to claim 1, wherein, The probability distribution model is generated through learning related to the heartbeat cycle of the unit of analysis.

3. The bioinformatics analysis system according to claim 1, wherein, The probability distribution model used is a Gaussian mixture model.

4. The bioinformatics analysis system according to claim 1, wherein, The clustering department uses information content criteria to select the probability distribution model and uses the selected probability distribution model to perform clustering.

5. The bioinformatics analysis system according to claim 1, wherein, The determination unit determines arrhythmia based on the number of clusters in the clustering results, the number of heartbeat cycles belonging to each cluster, the distribution width information of the heartbeat cycles belonging to each cluster, and / or the time information of the heartbeat cycles belonging to each cluster.

6. The bioinformatics analysis system according to any one of claims 1 to 3, wherein, The clustering unit has: The first clustering unit uses a probability distribution model related to the heartbeat cycle of the analysis unit to cluster the heartbeat cycles for the analysis unit. as well as The second clustering unit performs additional clustering when the clustering result is that two clusters exist in the heartbeat cycle. The bioinformatics analysis system also includes an arrhythmia reference information generation unit that generates arrhythmia reference information based on the additional clustering results. The second clustering unit uses the heartbeat cycles of the cluster with fewer heartbeat cycles belonging to the two clusters, as well as any one of the heartbeat cycles before and after the heartbeat cycle, to calculate the merged heartbeat cycle, and performs the additional clustering of the heartbeat cycles and the merged heartbeat cycles of the analysis unit using the probability distribution model. The determination unit makes a determination of arrhythmia based on the clustering results and the arrhythmia reference information.

7. The bioinformatics analysis system according to any one of claims 1 to 3, wherein, It also includes an arrhythmia reference information generation unit, which generates arrhythmia reference information when the clustering result shows that there are three clusters in the heartbeat cycle. The arrhythmia reference information generation unit determines that the longest cluster contains the longest heartbeat cycle time information, and the shortest cluster contains the shortest heartbeat cycle time information. It then generates arrhythmia reference information based on the similarity between the number of heartbeat cycles belonging to the longest cluster and the number of heartbeat cycles belonging to the shortest cluster. The determination unit makes a determination of arrhythmia based on the clustering results and the arrhythmia reference information.

8. The bioinformatics analysis system according to any one of claims 1 to 3, wherein, It also includes an arrhythmia reference information generation unit, which generates arrhythmia reference information when the clustering result indicates that there are three clusters in the heartbeat cycle. The arrhythmia reference information generation unit determines, from the three clusters, the longest cluster (containing the longest heartbeat cycle time information), the shortest cluster (containing the shortest heartbeat cycle time information), and the remaining intermediate clusters. It then calculates merged time information using the heartbeat cycle time information belonging to the longest cluster and the shortest cluster, and generates arrhythmia reference information based on the similarity between the merged time information and the heartbeat cycle time information belonging to the intermediate clusters. The determination unit determines the arrhythmia based on the clustering results and the arrhythmia reference information.

9. The bioinformatics analysis system according to claim 1, wherein, The heartbeat cycle of the analysis unit is baseline-corrected, and the heartbeat cycle is clustered for the analysis unit after baseline correction.

10. A bioinformatics analysis program that uses the biological information of an object to determine cardiac arrhythmia, the bioinformatics analysis program causing a computer to perform the following series of processes: The interval of the object's heartbeat activity is calculated based on the biological information obtained from the object to serve as the heartbeat cycle; The heartbeat cycle is divided into segments by reading a pre-set segment length from the storage unit and using that segment length as the unit of analysis. Clustering of the heartbeat cycles for the analysis unit is performed using a probability distribution model associated with the heartbeat cycle of the analysis unit. as well as The determination of arrhythmia is based on the results of the clustering.

Images