Non-invasive blood glucose monitoring device and method based on ballistocardiogram

Abstract

The application relates to a non-invasive blood glucose monitoring device and method based on a ballistocardiogram, which comprises a photoelectric transceiver module, a sensing module, an MCU and a terminal; the photoelectric transceiver module comprises a light source and a photoelectric detector; light emitted by the light source is incident on one port of the sensing module through an optical fiber or an optical fiber connector; light output by the sensing module reaches the photoelectric detector from another port through an optical fiber or an optical fiber connector; the photoelectric detector transmits data to the MCU; and the MCU is connected with the terminal through a wireless module. The application collects a ballistocardiogram (BCG) signal through an optical fiber sensor, analyzes the signal in combination with a deep learning algorithm, obtains continuous blood glucose values of a human body, and realizes non-invasive, non-sensing and continuous blood glucose level monitoring.

Description

Technical Field

[0001] This invention relates to the field of blood glucose monitors, and more specifically to a non-invasive blood glucose monitoring device and method based on cardiac impaction. Background Technology

[0002] An electrocardiogram (ECG) records the electrical potential changes of the heart with each heartbeat using electrical signals. The magnitude of these electrical changes varies from person to person, thus the characteristics of an ECG can reflect an individual's heart health. According to a published paper (Tobore I, Li J, Kandwal A, et al. Statistical and spectral analysis of ECG signal towards achieving non-invasive blood glucose monitoring[J]. BMCMed Inform Decis Mak,19, 266, 2019.), ECG signals can also reflect information about blood glucose levels. For example, in cases of hypoglycemia, the ECG signal shows a prolonged QT interval, and an increased heart rate can also be observed. However, acquiring ECG signals requires patching multiple parts of the body, which involves direct contact with the skin and requires operation by professional medical personnel, making it unsuitable for patients to monitor their blood glucose levels in a non-invasive, real-time manner.

[0003] BCG (Body Impact Gauge) is a method of describing the forces exerted by the heart on the body surface during a heartbeat. As the heart functions normally, the contraction and relaxation of blood vessels exert pressure on them. With each heartbeat, the flow of blood causes changes in the body's center of mass, resulting in minute movements on the body surface. Changes in blood sugar levels can affect these movements. For example... Figure 1 As shown, this BCG signal contains H, I, J, K, L, M, and N waves. A standard BCG signal has a shape largely consistent with an ECG signal; for example, the J wave corresponds to the R wave in an ECG signal, but there are some differences in specific waveform details. A stable BCG signal contains a large number of heart health signals and other vital signs, including blood glucose information.

[0004] Currently, ECG signal detection devices are bulky and complex to operate; wearable devices, although simple to operate, cannot achieve a truly undetectable experience; and the application of BCG signal-based devices is limited to the acquisition and monitoring of heartbeat and respiratory signals, and the characteristics of BCG waves recorded entirely by fiber optic sensors and their related applications to blood glucose have not yet been explored. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a non-invasive and non-contact blood glucose monitoring device and method based on cardiac impaction, which acquires BCG signals through fiber optic sensors and analyzes them with deep learning algorithms to obtain continuous blood glucose values ​​of the human body, thereby achieving non-invasive, non-contact and continuous blood glucose level monitoring.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A non-invasive blood glucose monitoring device based on cardiac impact mapping includes a photoelectric transceiver module, a sensing module, an MCU, and a terminal. The photoelectric transceiver module includes a light source and a photodetector. Light emitted by the light source is incident on one port of the sensing module through an optical fiber or an optical fiber connector. Light output by the sensing module reaches the photodetector from the other port through an optical fiber or an optical fiber connector. The photodetector transmits data to the MCU. The MCU is connected to the terminal via a wireless module.

[0008] Furthermore, the sensing module includes two ports and a mode interference fiber optic assembly.

[0009] Furthermore, the light source is a laser light source or a light-emitting diode.

[0010] Furthermore, the sensing module employs a fiber optic interferometer with near-period filtered spectral characteristics or a spectral characteristic device generated by a bent component.

[0011] Furthermore, the photodetector is a PIN photodiode detector, an APD photodetector, or a single-photon photodetector.

[0012] A blood glucose monitoring method based on a non-invasive blood glucose monitoring device using cardiac impaction includes the following steps:

[0013] Step S1: The person being tested lies down or leans against the sensing module. The heartbeat will exert pressure on the sensing module, causing the energy of the light to change during transmission.

[0014] Step S2: The photodetector receives a light signal containing patient characteristic information and transmits it to the MCU;

[0015] Step S3: In the MCU, the photocurrent is amplified, filtered, converted from analog to digital, and processed. After calculation, it is transmitted to the terminal through the wireless module to obtain the cardiac impact signal.

[0016] Step S4: Combine deep learning algorithms to perform modeling and analysis to obtain the corresponding blood glucose value.

[0017] Furthermore, step S4 specifically includes:

[0018] Step S41: Perform preprocessing on the cardiac impaction signal to remove baseline drift and power frequency interference;

[0019] Step S42: Perform segmented processing on the preprocessed signal to obtain the time-domain image features of each segment of the preprocessed signal;

[0020] Step S43: Based on the time-domain image features, the blood glucose values ​​corresponding to different waveforms are obtained by modeling using deep learning algorithms, and the patient's blood glucose signal is extracted from the cardiac impact diagram.

[0021] Step S44: Display the patient's specific blood glucose value on the terminal software, and classify it into three categories based on the magnitude of the blood glucose value: hypoglycemia, normal or hyperglycemia.

[0022] Furthermore, the time-domain image features consist of 19 features, which are composed of the reference distances and slopes of 9 directly connected reference points, namely, HI band length, HI band slope, HJ band length, HJ band slope, HK band length, HK band slope, HL band length, HL band slope, IJ band length, IJ band slope, IK band length, IK band slope, IL band length, IL band slope, JK band length, JK band slope, JL band length, and JL band slope; the distance between two adjacent JJ ​​bands is used as another judgment feature.

[0023] Deep learning algorithms are used to model and analyze the target data, and then the patient's blood glucose information is obtained from the preprocessed cardiac impulse signal, as shown in the following formula:

[0024] BS = X1*W1 + X2*W2 + ... + X19*W19

[0025] X1 to X19 are the 19 feature values ​​of the cardiac impact map, W1 to W19 are the parameters that need to be adjusted during the learning process, and BS is the output blood glucose value.

[0026] Compared with the prior art, the present invention has the following advantages:

[0027] This invention acquires BCG signals through fiber optic sensors and analyzes them using deep learning algorithms to obtain continuous blood glucose values, achieving non-invasive, unobtrusive, and continuous blood glucose level monitoring. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the cardiac impulse map of demodulation in the linear operating region according to an embodiment of the present invention;

[0029] Figure 2 This is a schematic diagram of the device structure of the present invention;

[0030] Figure 3This is a schematic diagram of the optical fiber assembly of the present invention;

[0031] Figure 4 This is a cardiac impact diagram according to one embodiment of the present invention;

[0032] Figure 5 This is a schematic diagram of cardiac impaction feature extraction in one embodiment of the present invention;

[0033] Figure 6 This is a schematic diagram of the data processing flow in one embodiment of the present invention;

[0034] Figure 7 This is a schematic diagram of the method flow of the present invention;

[0035] Figure 8 This is a schematic diagram showing the comparison results between the results of a hospital glucose tolerance test and the results of an embodiment of the present invention. Detailed Implementation

[0036] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0037] Please refer to Figure 1-8 This invention provides a non-invasive blood glucose monitoring device based on cardiac impact mapping, comprising a photoelectric transceiver module, a sensing module, an MCU, and a terminal. The photoelectric transceiver module includes a light source and a photodetector. Light emitted by the light source LD is incident on port 1 of the sensing module 5 through an optical fiber or optical fiber connector 11. The light incident on port 1 of the sensing module 5 passes through sensing loops 13, 34, and 24 and is then output from port 2. The output light passes through port 2 or optical fiber connector 22 and reaches the photodetector PD. The sensing module 5 has two ports, 1 and 2, and a mode interference fiber assembly, which forms an interference filtering spectrum through mode interference or other methods. When the light beam passes through the photodetector, it becomes a photocurrent. In the MCU, the photocurrent undergoes a series of processing steps such as amplification, filtering, analog-to-digital conversion, and calculation before being transmitted wirelessly to the terminal via methods such as WiFi or Bluetooth.

[0038] Preferably, the sensing module includes two ports and a mode interference fiber optic assembly. The fiber optic cable in the assembly is wound around a cylinder. The cylinder can be metal or plastic.

[0039] Preferably, the light source is a laser light source or a light-emitting diode.

[0040] Preferably, the sensing module employs a fiber optic interferometer with near-period filtered spectral characteristics or a spectral characteristic device generated by a bent component.

[0041] Preferably, the photodetector is a PIN photodiode detector, an APD photodetector, or a single-photon photodetector.

[0042] Preferably, the MCU (Microcontroller Unit) is used to perform a series of processes on the PIN photodetector (PD), including electrical amplification, filtering, analog-to-digital conversion, and calculations, which are then transmitted to the host computer via Bluetooth. The host computer software then performs various processing, analysis, display, and alarm functions on the received results.

[0043] In this embodiment, the implementation steps of the device are as follows: Figure 6 As shown, it includes:

[0044] (1) Collect the BCG signal to be monitored; when the patient lies down or leans against the sensor module 5, the heartbeat will exert pressure on the sensor module, causing the energy of the light to change during transmission, so that the photodetector can receive the light signal with patient characteristic information.

[0045] (2) Filter the BCG signal to be monitored to remove interference. Obtain the target data from the cardiac impaction signal, and first perform a preprocessing step, namely, baseline drift removal and power frequency interference removal. Then, segment the preprocessed signal and obtain the time-domain image features of each segment of the preprocessed signal, such as... Figure 3 As shown in the figure, the BCG signal exhibits H, I, J, K, L, M, and N waves.

[0046] (3) Data processing procedure as follows Figure 4-5 As shown, target data can be obtained after screening from cardiac impact maps. This target data is then segmented, and the temporal image features of each segment's preprocessed signal are obtained, such as HI band length, HI band slope, HJ band length, HJ band slope, HK band length, HK band slope, HL band length, HL band slope, IJ band length, IJ band slope, IK band length, IK band slope, IL band length, IL band slope, JK band length, JK band slope, JL band length, JL band slope, and the JJ interval between adjacent cardiac impact maps. Then, a deep learning algorithm is used to model and analyze the target data, thereby obtaining the patient's blood glucose information from the preprocessed cardiac impact signal, as shown in the following formula:

[0047] BS = X1*W1 + X2*W2 + ... + X19*W19

[0048] X1 to X19 are the 19 feature values ​​of the cardiac impact map, W1 to W19 are the parameters that need to be adjusted during the learning process, and BS is the output blood glucose value.

[0049] (4) Input all the extracted signal features into the pre-trained classification model; the terminal software displays the blood glucose value and classifies the blood glucose level according to the blood glucose value.

[0050] Using the above method, the predicted values ​​were compared with the hospital's glucose tolerance test results. The error between the predicted values ​​using this method and the hospital's test results was within the allowable standard. Figure 7 As shown.

[0051] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made within the scope of the claims of the present invention should be included in the scope of the present invention.

Claims

1. A blood glucose monitoring method using a non-invasive blood glucose monitoring device based on cardiac impact mapping, characterized in that, The non-invasive blood glucose monitoring device includes a photoelectric transceiver module, a sensing module, an MCU, and a terminal; the photoelectric transceiver module includes a light source and a photodetector. The light emitted by the light source is incident on one port of the sensing module through an optical fiber or an optical fiber connector; the light output by the sensing module reaches the photodetector from the other port through an optical fiber or an optical fiber connector. The photodetector transmits data to the MCU; the MCU connects to the terminal via a wireless module. The blood glucose monitoring method includes the following steps: Step S1: The person being tested lies down or leans against the sensing module. The heartbeat will exert pressure on the sensing module, causing the energy of the light to change during transmission. Step S2: The photodetector receives a light signal containing patient characteristic information and transmits it to the MCU; Step S3: In the MCU, the photocurrent is amplified, filtered, converted from analog to digital, and processed. After calculation, it is transmitted to the terminal through the wireless module to obtain the cardiac impact signal. Step S4: Use deep learning algorithms to perform modeling and analysis to obtain the corresponding blood glucose values. Step S41: Perform preprocessing on the cardiac impaction signal to remove baseline drift and power frequency interference; Step S42: Perform segmented processing on the preprocessed signal to obtain the time-domain image features of each segment of the preprocessed signal; Step S43: Based on the time-domain image features, the blood glucose values ​​corresponding to different waveforms are obtained by modeling using deep learning algorithms, and the patient's blood glucose signal is extracted from the cardiac impact diagram. Step S44: Display the patient's specific blood glucose value on the terminal software, and classify it into three categories based on the magnitude of the blood glucose value: hypoglycemia, normal or hyperglycemia; The time-domain image features consist of 19 features, which are composed of the reference distances and slopes of 9 directly connected reference points. These features are: HI band length, HI band slope, HJ band length, HJ band slope, HK band length, HK band slope, HL band length, HL band slope, IJ band length, IJ band slope, IK band length, IK band slope, IL band length, IL band slope, JK band length, JK band slope, JL band length, and JL band slope. The distance between two adjacent JJ ​​bands is used as another feature for judgment. Deep learning algorithms are used to model and analyze the target data, and then the patient's blood glucose information is obtained from the preprocessed cardiac impulse signal, as shown in the following formula: BS = X1*W1 + X2*W2 + ... + X19*W19 X1 to X19 are the 19 feature values ​​of the cardiac impact map, W1 to W19 are the parameters that need to be adjusted during the learning process, and BS is the output blood glucose value.

2. The blood glucose monitoring method of the non-invasive blood glucose monitoring device based on cardiac impact mapping according to claim 1, characterized in that, The sensing module includes two ports and a mode interference fiber optic assembly.

3. The blood glucose monitoring method of the non-invasive blood glucose monitoring device based on cardiac impact mapping according to claim 1, characterized in that, The light source is either a laser or a light-emitting diode.

4. The blood glucose monitoring method of the non-invasive blood glucose monitoring device based on cardiac impact mapping according to claim 1, characterized in that, The sensing module employs a fiber optic interferometer or a spectral device with near-period filtered spectral characteristics, or a device with spectral characteristics generated by a bent component.

5. The blood glucose monitoring method of the non-invasive blood glucose monitoring device based on cardiac impact mapping according to claim 1, characterized in that, The photodetector is a PIN photodiode detector, an APD photodetector, or a single-photon photodetector.

Images