A non-invasive blood glucose estimation method, system and device based on multi-index quality score gating and a storage medium
By employing a multi-index quality scoring gating method, the instability problem caused by low-quality signals in non-invasive blood glucose estimation is resolved, thereby improving the stability and reliability of the model and making it suitable for deployment on edge devices.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- YISHAN MEDICAL IND MANAGEMENT GRP CO LTD
- Filing Date
- 2026-05-14
- Publication Date
- 2026-07-14
AI Technical Summary
In existing non-invasive blood glucose estimation techniques, the direct involvement of low-quality pulse wave signals in model inference leads to unstable results and large errors. Furthermore, the lack of a multi-indicator quality fusion evaluation mechanism affects the stability and reliability of the model.
A multi-index quality scoring gating method is adopted. After acquiring the photoplethysmography (PPG) signal and performing preprocessing, indicators such as amplitude, period stability, noise, and dual-wavelength consistency are extracted to generate a quality score. Based on the score, a hierarchical gating strategy is executed to control the inference process of the blood glucose estimation model.
It improves the accuracy of signal availability assessment, enhances the stability and engineering fault tolerance of model output, reduces the computational power consumption of invalid inference, and strengthens user interaction capabilities, making it suitable for deployment on edge devices.
Smart Images
Figure CN122376094A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of non-invasive physiological testing and intelligent health analysis technology, and in particular to a non-invasive blood glucose estimation method, system, device and storage medium based on multi-index quality scoring gating. Background Technology
[0002] Blood glucose monitoring plays a crucial role in screening for abnormal glucose metabolism, identifying diabetes risk, and managing long-term health. Current blood glucose testing methods mostly rely on blood sampling, which presents challenges such as being invasive, inconvenient for high-frequency monitoring, and having low user compliance. Non-invasive blood glucose estimation technology based on photoplethysmography (PPG) signals has attracted widespread attention due to its potential for non-invasiveness, portability, and continuous monitoring.
[0003] However, pulse wave signals are easily affected by factors such as finger placement, local perfusion level, ambient light interference, motion artifacts, changes in contact pressure, and device noise. When a poor-quality pulse wave signal is directly fed into a blood glucose estimation model, it can easily lead to unstable model output, increased errors, and even misleading results.
[0004] Current technologies typically address signal quality issues through simple filtering, amplitude thresholding, or manual waveform observation. They lack a quality gating mechanism that can integrate and evaluate multiple quality factors and further serve as a basis for model admission. Therefore, there is an urgent need to provide a non-invasive blood glucose estimation method based on multi-index quality scoring gating to improve the stability, reliability, and engineering usability of non-invasive blood glucose estimation. Summary of the Invention
[0005] The purpose of this invention is to provide a non-invasive blood glucose estimation method, system, device, and storage medium based on multi-index quality scoring gating, so as to solve the problems of unstable results, large errors, and insufficient practicality in existing non-invasive blood glucose estimation due to the direct participation of low-quality pulse wave signals in model inference.
[0006] To achieve the above objectives, this invention provides a non-invasive blood glucose estimation method based on multi-index quality score gating, comprising the following steps: acquiring photoplethysmography (PPG) signals from the target site of the subject; preprocessing the PPG signals to obtain preprocessed pulse signals; extracting at least three types of signal quality evaluation indicators based on the preprocessed pulse signals, and generating quality scores based on the signal quality evaluation indicators; and performing hierarchical gating based on the quality scores to control the blood glucose estimation model to execute in one of the following modes: normal reasoning, restricted reasoning, or prohibited reasoning.
[0007] Preferably, the photoplethysmography (PPG) signal is a dual-wavelength PPG signal, comprising red light and infrared light, with the center wavelength of the red light being approximately 660 nm and the center wavelength of the infrared light being approximately 880 nm.
[0008] Preferably, the signal quality evaluation indicators include at least three categories: amplitude or injection indicators, period stability indicators, and noise or artifact indicators; more preferably, it also includes dual-wavelength channel consistency indicators.
[0009] Preferably, the method for generating the quality score includes: standardizing or normalizing each quality evaluation indicator; fusing them according to preset weights; and outputting the quality score.
[0010] Preferably, the graded gating includes a three-level strategy: high-quality signals allow direct entry into the blood glucose estimation model; medium-quality signals enter the restricted inference process; and low-quality signals prohibit entry into the blood glucose estimation model and prompt for remeasurement.
[0011] Compared with existing technologies, this invention has at least the following advantages: It generates quality scores through multi-index fusion rather than relying on a single threshold, thus improving the accuracy of signal availability evaluation; it uses the quality score as the model admission criterion, avoiding the direct participation of low-quality pulse wave signals in blood glucose estimation and improving model output stability; it employs three types of gating strategies—allowing, restricting, and prohibiting—to give the system better engineering fault tolerance and user interaction capabilities; it is suitable for deployment on edge devices, which helps improve real-time performance and reduce computational consumption caused by invalid inference; and it is compatible with dual-wavelength PPG non-invasive blood glucose estimation systems, facilitating the formation of a complete patent portfolio. Attached Figure Description
[0012] Figure 1 This is a schematic diagram of the method flow of the present invention.
[0013] Figure 2 This is a schematic diagram of the system structure of the present invention.
[0014] Figure 3 This is a schematic diagram of the hierarchical gating logic of the present invention.
[0015] Figure 4 This is a schematic diagram illustrating the generation of the quality score in this invention. Detailed Implementation
[0016] The present invention will be further described below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited thereto.
[0017] In one embodiment, a finger-clip optical detection device is used to acquire dual-wavelength photoplethysmography (PPG) signals from the finger area of the subject. The dual wavelengths include red light and infrared light, with the center wavelength of the red light being approximately 660 nm and the center wavelength of the infrared light being approximately 880 nm. The sensor outputs dual-channel PPG signals, and an on-device processor samples and buffers these signals.
[0018] Subsequently, the acquired pulse wave signals are preprocessed. Preprocessing may include one or more of the following: DC drift removal, bandpass filtering, median filtering, outlier removal, and baseline correction, to reduce the impact of ambient light, low-frequency drift, and high-frequency noise on subsequent analysis.
[0019] Based on the preprocessed pulse wave signal, multiple signal quality evaluation indicators are extracted. For example, amplitude or perfusion indicators are used to evaluate whether the current pulse wave has sufficient effective fluctuation amplitude; period stability indicators are used to characterize the periodic changes between multiple consecutive pulse cycles; noise or artifact indicators are used to measure the degree of high-frequency noise or motion artifact interference; and dual-wavelength channel consistency indicators are used to evaluate the synchronization and consistency between the red light channel and the infrared light channel.
[0020] In a preferred embodiment, after normalizing each quality evaluation index, a quality score (quality_score) is generated using a weighted fusion method. For example, it can be represented as: quality_score = Σ(wi×Qi). Here, Qi represents the standardized value corresponding to different quality evaluation indices, and wi represents the preset weight.
[0021] After the quality score is generated, hierarchical gating logic is executed. For example: when quality_score ≥ T1, the current signal quality is determined to be high, and blood glucose estimation model inference is allowed; when T2 ≤ quality_score < T1, the current signal quality is determined to be medium, and at least one of the following can be executed: extending the sampling duration, increasing the number of sampling segments, downgrading inference, or outputting a low confidence prompt; when quality_score < T2, the current signal quality is determined to be low, entry into the blood glucose estimation model is prohibited, and a remeasurement prompt is output.
[0022] When inference is allowed, time-domain, frequency-domain, and morphological features for blood glucose estimation are further extracted from the pulse wave signal and input into the blood glucose estimation model to obtain the estimated blood glucose value or blood glucose risk indicator. The blood glucose estimation model can be a linear regression model, a random forest model, a lightweight neural network model, or other models suitable for edge deployment.
[0023] In one embodiment, to improve user experience and interpretability of results, the system not only outputs the blood glucose estimate, but also simultaneously outputs the signal quality level, measurement confidence level, or remeasurement suggestion, thereby avoiding user misjudgment of the output results under low-quality data conditions.
[0024] Combination Figure 111 indicates PPG signal acquisition, 12 indicates signal preprocessing, 13 indicates quality index extraction, 14 indicates quality score generation, 15 indicates quality gating judgment, 16 indicates normal inference, 17 indicates restricted inference, 18 indicates prohibited inference, 19 indicates blood glucose estimation result, 20 indicates low confidence result, and 21 indicates remeasurement prompt.
[0025] Combination Figure 2 1 represents the signal acquisition module, 2 represents the preprocessing module, 3 represents the quality evaluation module, 4 represents the quality scoring module, 5 represents the gating module, 6 represents the inference module, and 7 represents the feedback module.
[0026] Combination Figure 3 31 indicates a quality score, 32 indicates high quality, 33 indicates medium quality, 34 indicates low quality, 35 indicates normal reasoning and output, 36 indicates restricted reasoning and output, and 37 indicates prohibited reasoning and prompt for retesting.
[0027] Combination Figure 4 41 represents dual-wavelength PPG signal, 42 represents amplitude or injection index, 43 represents period stability index, 44 represents noise or artifact index, 45 represents dual-wavelength consistency index, 46 represents weighted fusion, and 47 represents quality score.
[0028] The present invention can also be embedded in a terminal device in software form, or stored as a program in a computer-readable storage medium, and executed by a processor to complete the above method.
Claims
1. A non-invasive blood glucose estimation method based on multi-index quality scoring gating, characterized in that, The process includes the following steps: S1, acquiring the photoplethysmography (PPG) signal of the target area of the object to be tested; S2, preprocessing the PPG signal to obtain a preprocessed pulse wave signal; S3, extracting multiple signal quality evaluation indicators from the preprocessed pulse wave signal, wherein the multiple signal quality evaluation indicators include at least amplitude or perfusion indicators, period stability indicators, and noise or artifact indicators; S4, standardizing and weighting the multiple signal quality evaluation indicators to generate a quality score characterizing the availability of the current pulse wave signal; S5, implementing hierarchical gating based on the quality score to control the execution strategy of the blood glucose estimation model; S6, when the quality score reaches a first preset condition, performing normal inference and outputting a blood glucose estimation value or a blood glucose risk indicator. S7. When the quality score is within the second preset condition, perform restricted inference and output a blood glucose estimation result with confidence prompts; S8. When the quality score does not meet the minimum preset condition, prohibit the execution of the blood glucose estimation model and output a remeasurement prompt.
2. The method according to claim 1, characterized in that, The photoplethysmography (PPG) signal is a dual-wavelength PPG signal, which includes a red light channel and an infrared light channel. The center wavelength of the red light channel is approximately 660 nm, and the center wavelength of the infrared light channel is approximately 880 nm.
3. The method according to claim 1, characterized in that, The multiple signal quality evaluation indicators include at least three of the following: amplitude or perfusion indicators, period stability indicators, noise or artifact indicators, and dual-wavelength channel consistency indicators. The amplitude or perfusion indicators include at least one of pulse wave AC component amplitude, AC to DC component ratio, and perfusion intensity. The period stability indicators include at least one of peak-valley detection success rate, difference between adjacent pulse cycles, heart rate stability, and waveform repeatability. The noise or artifact indicators include at least one of high-frequency noise energy, number of abnormal abrupt changes, motion artifact intensity, and baseline drift. The dual-wavelength channel consistency indicator is used to characterize the consistency between the red light channel and the infrared light channel in pulse cycle position, waveform change trend, or characteristic response.
4. The method according to claim 1, characterized in that, The quality score in step S4 is generated as follows: each signal quality evaluation index is normalized to its corresponding standardized value; corresponding weights are assigned according to the importance of each signal quality evaluation index; and the standardized values are weighted and summed or weighted and combined to obtain the quality score.
5. The method according to claim 1, characterized in that, The graded gating in step S5 includes: allowing the blood glucose estimation model to execute in normal mode when the quality score is greater than or equal to the first threshold; performing restricted inference when the quality score is less than the first threshold but greater than or equal to the second threshold; and prohibiting the blood glucose estimation model from executing when the quality score is less than the second threshold. The restricted inference includes at least one of the following: extending the sampling time before inference, increasing the number of sampling segments before inference, calling a simplified model or a downgraded model for inference, and attaching a low confidence level identifier when outputting the results.
6. The method according to claim 1, characterized in that, The blood glucose estimation model is a machine learning model built based on pulse wave features. The pulse wave features include at least one of time-domain features, frequency-domain features, and morphological features. The blood glucose estimation model includes at least one of linear regression model, random forest model, lightweight neural network model, or other models suitable for edge deployment.
7. The method according to claim 1, characterized in that, The output of the blood glucose estimation result with confidence prompts includes simultaneously outputting the blood glucose estimation value and at least one of the following: a confidence level corresponding to the quality score, a risk warning, or a measurement suggestion.
8. A non-invasive blood glucose estimation system based on multi-index quality scoring gating, characterized in that, include: The signal acquisition module is used to acquire the photoplethysmography (PPG) signal of the target part of the object under test. The preprocessing module is used to preprocess the photoplethysmography (PPG) signal. The quality evaluation module is used to extract various signal quality evaluation indicators. The quality scoring module is used to standardize and weightedly fuse the multiple signal quality evaluation indicators to generate a quality score; the gating module is used to control the execution strategy of the blood glucose estimation model based on the quality score; and the inference module is used to output the blood glucose estimation value or blood glucose risk indicator when the gating is enabled. The feedback module is used to output low confidence prompts, remeasurement prompts, or limiting results when gating is restricted or prohibited; wherein, the quality evaluation module, quality scoring module, gating module, and inference module are deployed in the edge processor, which is any one of a microcontroller, an embedded processor, or a system-on-a-chip.
9. A non-invasive blood glucose estimation device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program that, when executed by the processor, implements the method according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1 to 7.