Physiological value sensing device and sensing method thereof

Abstract

This invention relates to a physiological value sensing device and its sensing method. The physiological value sensing device includes an input module and a processing module. The input module provides multi-band light to illuminate a subject, generating multiple optical signals, including multiple interference signals and a physiological signal to be measured. The processing module establishes a normal physiological model, which has multiple sample physiological values ​​corresponding to the multi-band light. The processing module converts the interference signals and the physiological signal to be measured into the normal physiological model for fitting, thereby eliminating interference signals based on the sample physiological values ​​and generating a physiological signal corresponding to the physiological value to be measured.

Description

Physiological value sensing device and sensing method thereof The present invention relates to a physiological value sensing device and a sensing method thereof, and more particularly to a non-invasive blood glucose sensing device and a sensing method thereof. Non-invasive blood glucose testing technology is a technology that measures blood glucose levels without puncturing the skin or extracting blood samples. The most obvious advantage of this technology is that it does not require the use of needles for puncture, which is more comfortable than traditional blood glucose testing methods and reduces pain and discomfort for patients. In addition, non-invasive testing methods also reduce the risk of infection for the subjects and more actively promote public health. Moreover, such devices are easy to operate and convenient for users to use at home or in their daily lives, without the need for special training or the assistance of medical professionals. Most importantly, non-invasive testing can provide continuous monitoring functions and provide more detailed data all the time, which helps patients manage related diseases more easily. However, non-invasive testing technologies still have some shortcomings. For example, the accuracy of current non-invasive blood glucose testing technologies is relatively low, affected by a variety of factors, including poor optical performance of the testing equipment, individual differences in subjects, and interference from various physiological values ​​in the subjects' bodies. This often leads to doubts about the accuracy of non-invasive blood glucose testing technologies. To overcome these issues, the industry urgently needs an innovative non-invasive physiological value sensing device and sensing method to improve these issues. The main purpose of the present invention is to provide an innovative physiological value sensing device and sensing method thereof to improve the problems of poor optical performance, individual differences between subjects, and various physiological value interferences in the subjects' bodies that affect sensing accuracy in conventional technologies. To achieve the above objectives, the present invention provides a physiological value sensing device comprising an input module and a computing module. The input module is configured to provide a multi-band light beam to illuminate a subject to be measured, thereby generating a plurality of optical signals, including a plurality of interference signals and a physiological signal to be measured. The computing module is configured to establish a physiological normality model having a plurality of sample physiological values ​​corresponding to the multi-band light beam. The computing module converts the interference signals and the physiological signal to be measured into the physiological normality model for fitting. The computation module then eliminates the interference signals based on the sample physiological values ​​and generates a physiological value to be measured corresponding to the physiological signal to be measured. In one embodiment of the physiological value sensing device of the present invention, the input module includes a plurality of light-emitting units and at least one light-detecting unit. The light-emitting units are used to provide the multi-band light and can adjust the light intensity of the multi-band light. The light-detecting unit is used to receive the reflected light of the multi-band light after diffuse reflection by the subject and generate an optical signal based on the reflected light. In one embodiment of the physiological value sensing device of the present invention, the light emitting unit can provide the multi-band light with a wavelength range of approximately 800 nm to 1700 nm and a wavelength interval of not less than 100 nm. In one embodiment of the physiological value sensing device of the present invention, the input module further includes a microstructure disposed adjacent to the light detection unit to prevent the multi-band light provided by the light emitting unit from being received by the light detection unit before being reflected by the subject. In one embodiment of the physiological value sensing device of the present invention, the computing module includes a processing submodule, a fitting submodule, and an evaluation submodule. The processing submodule is used to process and convert interference signals and physiological signals to be measured into a physiological normal model. The fitting submodule is used to fit the converted interference signals and physiological signals to be measured with the physiological normal model to generate a fitting result, and after eliminating the interference signal based on the sample physiological value, generates a physiological value to be measured corresponding to the physiological signal to be measured. The evaluation submodule is used to output a feature importance evaluation based on the fitting result. In one embodiment of the physiological value sensing device of the present invention, a storage module is further included to store the optical signal generated by the input module, the converted interference signal, the physiological signal to be measured, the physiological value to be measured, and the feature importance evaluation. In one embodiment of the physiological value sensing device of the present invention, a post-processing module is further included to create a long-term trend report based on the optical signal. In one embodiment of the physiological value sensing device of the present invention, an output module is further included to output the physiological value to be measured, the feature importance evaluation and the long-term trend report. To achieve the above-mentioned objectives, the present invention provides a physiological value sensing method, comprising: establishing a physiological normality model, wherein the physiological normality model has a plurality of sample physiological values ​​corresponding to a multi-band light; providing the multi-band light to illuminate a subject to be measured to generate a plurality of optical signals, wherein the optical signals include a plurality of interference signals and a physiological signal to be measured; converting the interference signals and the physiological signal to be measured into the physiological normality model; eliminating the interference signals based on the sample physiological value; and generating a physiological value to be measured corresponding to the physiological signal to be measured. In one embodiment of the physiological value sensing method of the present invention, providing multi-band light is providing multi-band light with a wavelength range of approximately 800 nm to 1700 nm and a wavelength interval of not less than 100 nm. In one embodiment of the physiological value sensing method of the present invention, an artificial intelligence model is further provided to fit the converted interference signal and the physiological signal to be measured with the sample physiological value of the physiological normal model to generate a fitting result. After eliminating the interference signal based on the fitting result, the physiological value to be measured corresponding to the physiological signal to be measured is generated. After referring to the drawings and the following description of the embodiments, those skilled in the art will be able to understand other objectives of the present invention, as well as the technical means and implementation aspects of the present invention. The following examples illustrate the present invention. These examples are not intended to limit the present invention to any specific environment, application, or method. Therefore, the description of these examples is intended solely for the purpose of illustrating the present invention and is not intended to limit the present invention. It should be noted that in the following examples and figures, devices not directly related to the present invention are omitted from illustration, and the dimensional relationships between the devices in the figures are for ease of understanding only and are not intended to limit the actual scale. As shown in Figure 1, based on the theory that infrared light in the wavelength range of 600 nanometers (nm) to 2600 nanometers (nm) can be absorbed and scattered by components such as water, glucose, fat, hemoglobin, and protein in human tissue, resulting in changes in absorbance, the present invention discloses a physiological value sensing device and sensing method thereof, particularly a blood glucose concentration sensing device with multi-band optical signals and a sensing calculation method thereof. Specifically, the physiological value sensing device of the present invention is a non-invasive continuous glucose monitoring (NICGM) device that provides real-time, continuous, and long-term blood glucose concentration monitoring for the subject. Please refer to FIG2 , which shows a physiological value sensing device 1 according to the present invention. The device includes an input module 10, a computing module 20, a storage module 30, a post-processing module 40, and an output module 50. Please also refer to FIG3 , which shows a top view of several possible implementations of the input modules 10 in the physiological value sensing device 1 according to the present invention. The input modules 10 include several light-emitting units 12, at least one light-detecting unit 14, and a microstructure 16. In a preferred embodiment, the light-emitting units 12 include three or more light-emitting diodes (LEDs) that provide infrared light in different wavelength bands ranging from approximately 800 nm to 1700 nm, with wavelength intervals of at least 100 nm. Each LED can adjust the light intensity between 0% and 100% of the light output, and the spacing between the LEDs is preferably maintained between 1 mm and 2 mm. In addition, the photodetector unit 14 has one or more photodiodes that correspond to the multiple light-emitting units 12 to receive the multi-band infrared light emitted by the light-emitting units 12. When in use, the photodetector unit 14 receives the reflected light of the infrared light after diffuse reflection from the human body of the subject, and generates a plurality of optical signals based on the reflected light. When the physiological value sensing device 1 of the present invention has a plurality of photodetector units 14, the spacing between each photodetector unit 14 must be maintained fixed, and the spacing range is preferably maintained between 1 mm and 2 mm. It should be noted that after the corresponding light-emitting unit 12 irradiates the subject with infrared light, the optical signals generated by the photodetector unit 14 include various substances in the subject's body, including water, glucose, fat, hemoglobin, and protein, which are generated by the corresponding infrared light reflection. Taking the application of the present invention to a non-invasive multi-band blood glucose concentration sensing device as an example, the above-mentioned multiple optical signals include several interference signals such as water, fat, hemoglobin and protein, and a physiological signal to be measured represented by glucose concentration (i.e., blood glucose concentration). Secondly, the microstructure 16 is used to separate the light-emitting unit 12 from the light-detecting unit 14. Depending on the actual embodiment, the microstructure 16 has an adjustable appearance to maintain the spacing between the microstructure 16 and the light-emitting unit 12 within a range of 1 mm to 3 mm. The input module 10 can have multiple spacing ranges to accommodate multiple light-emitting units 12. In a preferred embodiment, the microstructure 16 is positioned to prevent light emitted by the light-emitting unit 12 from being directly received by the light-detecting unit 14 without being reflected by the patient's body. Therefore, the spatial relationship between the microstructure 16 and the light-emitting unit 12 is preferably such that the effective light emission angle of the light-emitting unit 12 is limited to within 30 degrees, substantially blocking most light rays greater than 30 degrees from being substantially blocked by the microstructure 16. Furthermore, the microstructure 16, disposed adjacent to the light-detecting unit 14, maintains a fixed spacing between the microstructure 16 and the light-detecting unit 14, for example, preferably within a range of 1 mm to 2 mm. To improve the accuracy of physiological value sensing, one feature of the physiological value sensing device of the present invention is the establishment of a physiological normality model. This physiological normality model maps the optical signals obtained when human tissue is illuminated by multi-band light to the physiological values ​​of multiple samples of the tissue at the time. Specifically, when establishing the physiological normality model, the values ​​of various physiological characteristics (e.g., water, glucose, lipids, fat, hemoglobin, and protein) in the human tissue are first obtained through testing by a biochemical analyzer. These values ​​are then combined with the optical signals generated by the input module 10 during optical testing. After calculation by the computing module 20, a normality model is established, which aligns the sample physiological values ​​with the optical signals. This physiological normality model can be used for fitting and prediction during future testing of subjects. More specifically, in a preferred embodiment, the light-emitting unit 12 of the present invention includes multiple light-emitting diodes that provide light of different wavelengths and intensities. Furthermore, the multiple light-emitting diodes are spaced at different distances from the light-detecting unit 14. When establishing the physiological normality model, by controlling the three variables of wavelength, distance, and light intensity, a complete physiological normality model can be established to calculate physiological values ​​close to those in the real state, thereby improving sensing accuracy. As shown in Figure 2, the computational module 20 includes a processing submodule 22, a fitting submodule 24, and an evaluation submodule 26. The processing submodule 22 processes the optical signal generated by the optical detection unit 14 in the input module 10. Specifically, it converts the interference signals (e.g., characteristic signals of water, lipids, fat, hemoglobin, and protein) and the physiological signal to be measured (e.g., characteristic signal of glucose) contained in the optical signal into the physiological normal model. It should be noted that converting the optical signal from the input module into the physiological normal model here means that the optical signal obtained from the measured subject must be in the same data format as the physiological normal model to meet the requirements of proper conversion. For example, when establishing a physiological normality model, if the data format is established using three wavelength bands of light (A nanometer (nm), B nanometer (nm), and C nanometer (nm), distances (P millimeters (mm), Q millimeters (mm), and R millimeters (mm), and light intensities of 100%, 90%, ..., and 10%), then when measuring the optical signal of the subject under test, the input module must also collect relevant signals corresponding to the same data format mentioned above to meet the requirements of normal conversion. Next, the fitting submodule 24 is used to fit the converted optical signal to the physiological normal model. Specifically, it fits the multiple interference signals in the optical signal and the physiological signal to be measured to the physiological normal model to produce a fitting result. Based on the known physiological values ​​of the sample in the physiological normal model, it eliminates interference signals other than glucose and generates a physiological value to be measured (e.g., blood glucose concentration) corresponding to the physiological signal to be measured. Please refer to Figure 4, which shows a schematic diagram of the fitting submodule performing model fitting in one embodiment of the present invention. In this embodiment, the processing submodule 22 receives multiple photoelectric signals corresponding to blood glucose, albumin / protein, lipids / fats, etc. collected by the input module. Each photoelectric signal contains 18 sets of photoelectric signals corresponding to different wavelength bands, distances, and light intensity characteristic distributions. These signals are then converted into multiple photoelectric characteristic values ​​based on the data format of the physiological normal model. The fitting submodule of the present invention utilizes an artificial intelligence model or a machine learning model to fit the converted interference signals (including photoelectric characteristic values ​​of albumin / protein, lipid / fats, etc.) and the physiological signal to be measured (i.e., the photoelectric characteristic value of blood glucose) with known sample physiological values ​​from a physiological normality model to generate a fitting result. The fitting submodule then uses the known sample physiological values ​​to correct the interference signals (e.g., corresponding to the photoelectric characteristic values ​​of albumin / protein, lipid / fats, etc.) in the optical signal, and generates the physiological value to be measured corresponding to the physiological signal to be measured (i.e., the photoelectric characteristic value of blood glucose). This physiological value to be measured is the physiological value of the subject to be measured, such as blood glucose concentration. In other words, the physiological value estimated by the fitting submodule of the present invention eliminates the influence of interfering substances in the human body on the physiological value. Therefore, compared with the prior art, it can estimate the physiological value of the subject closer to the true state of the subject. The following will use an embodiment to illustrate how the present invention applies an artificial intelligence model to perform model fitting on the converted values ​​using an artificial intelligence / machine learning (AI / ML) model and estimate the physiological value to be measured. As previously described, the present invention establishes a physiological normality model using wavelength, distance, and light intensity as characteristic parameters. These three types are further subdivided into three wavelength bands: A nm, B nm, and C nm; three distances between the light-emitting unit and the light-detecting unit: P mm, Q mm, and R mm; and ten light intensities: 100% (L00), 90% (L90), ..., and 10% (L10), for a total of 90 combinations of data. As previously mentioned, when the subject's physiological values ​​are subsequently measured, the functions of various substances in the subject's body must be established using these 90 combinations of data. For a single substance, its absorption rate varies at a fixed location and in different wavelength bands, resulting in different absorption thresholds for light intensities in different wavelength bands. Furthermore, different locations affect the depth of penetration of light of different wavelength bands into biological tissue. The following is an example formula, where function f is an artificial intelligence model and the corresponding substance concentration value can be calculated through an optimization process: Glucose value = f (PAL001, PAL901, …, PAL101, PBL001, …, RCL101) Protein value = f (PAL002, PAL902, …, PAL102, PBL002, …, RCL102) After actual measurement, the glucose value can be organized into a matrix form as follows: In the data matrix, the value of each feature column is converted using the above physiological normal model to establish the norm formula: ,in is the norm mean, The function f is then optimized, which includes optimizing the error function and generating the decision process glucose value. = For each function f(PAL001, PAL901, …, PAL101, PBL001, …, RCL101), there is a possibility of optimizing (minimizing) the error. Taking the mean square error as an error function, there is in is the number of samples, is the actual value / observed value of the function, If the decision tree type is used as the decision process for the function model calculation value, there are two spatial divisions, region 1 and region 2, for each j variables The value s & And solve For the first j variables The best value can be found s An example of the decision tree's decision-making process can be seen in Figure 5. It should be noted that Figure 5 is for illustration only and does not represent a real-world scenario. For example, if the characteristic value of PAL001 (glucose characteristic item: distance between the light-emitting unit and the light-detecting unit is P, light-emitting wavelength is A, and light intensity is 100%) calculated by the algorithm described above is less than or equal to 0.256, the decision tree will shift to the left. Conversely, if the PAL001 characteristic value is greater than 0.256, the decision tree will shift to the right. The next characteristic value will then be evaluated. If the characteristic value reaches PBL001, which is less than or equal to 1.123, the decision tree will estimate the blood glucose concentration to be 89.56 mg / dL. Secondly, another feature of the present invention is that the computing module 20 further includes an evaluation submodule 26 for performing model analysis based on the fitting results of the fitting submodule 24 and outputting a feature importance assessment, such as, but not limited to, the SHAP (SHapley Additive exPlanations, abbreviated as SHAP) feature importance assessment. This assessment considers the contribution of each feature to the prediction result and assigns a SHAP value to each feature. These values ​​describe the degree of influence of each feature on the model prediction, with positive values ​​indicating an increase in the predicted value and negative values ​​indicating a decrease in the predicted value. SHAP values ​​can be used to explain how the artificial intelligence model uses various optical features to predict blood glucose concentration. These optical features may include physiological values, environmental factors, etc. Using SHAP values, the impact of each feature on the output of the artificial intelligence model can be evaluated, thereby better understanding the prediction process of the artificial intelligence model. Based on the above, referring to FIG6 , the physiological value sensing method disclosed in the present invention includes the following steps: In step 601, a physiological normal model is first established, wherein the physiological normal model has a plurality of sample physiological values ​​corresponding to a multi-band light. In step 602, the multi-band light is provided to illuminate a subject to be measured to generate a plurality of optical signals, wherein the optical signals include a plurality of interference signals and a physiological signal to be measured. In step 603, the interference signals and the physiological signal to be measured are converted to the physiological normal model. In step 604, the interference signals are eliminated based on the sample physiological values. In step 605, a physiological value to be measured is generated corresponding to the physiological signal to be measured. Specifically, taking the measurement of blood glucose concentration as an example, the blood glucose sensing method of the present invention can correct the effects of other interfering substances in the body on blood glucose concentration measurement. The system primarily utilizes an input module to measure the subject's blood glucose level and generate multiple optical signals for evaluating blood glucose and eliminating interfering substances. Based on these optical signals, a computational module then obtains a set of uncorrected characteristic values ​​related to the subject's blood glucose level, including blood glucose concentration, protein concentration, lipid concentration, water contribution, and other substance contribution values. Finally, calculations based on a previously established physiological signal normalization model generate a corrected blood glucose value, representing the subject's true blood glucose level. In addition, as shown in FIG2 , the storage module 30 of the physiological value sensing device 1 of the present invention is used to store the optical signals generated by the input module 10, the interference signals converted by the computing module 20, the physiological signals to be measured, the physiological values ​​to be measured, the intermediate signals such as the feature importance assessment, and the final results. Furthermore, the post-processing module 40 of the physiological value sensing device 1 is used to calculate the values ​​stored in the storage module 30, including the optical signals, to create a long-term trend report after the system has been operating for a period of time. The output module 50 of the physiological value sensing device 1 of the present invention is used to output the physiological values ​​to be measured, the feature importance assessment, and the long-term trend report, and can output the values ​​of the computing module and the graphical file so that the results can be presented on a display. The physiological value sensing device and sensing method of the present invention have the advantages of being non-invasive, using multi-band light, capable of continuous long-term monitoring, and capable of real-time offline detection of physiological values. Compared to conventional technologies, the present invention can solve (1) the pain caused to the subject by invasive detection devices or methods; (2) the potential inflammation of the skin invasion site due to long-term wear if physiological values ​​need to be continuously monitored, causing psychological concerns and additional burdens for the subject, caregiver, or medical system; and (3) the numerous interference factors in human tissue that affect the physiological values ​​to be measured. The sensing device of the present invention and its associated sensing algorithm can eliminate specific interference factors and calibrate the relevant values, thereby providing physiological values ​​that are close to the true value. The above embodiments are intended only to illustrate the present invention and its technical features, and are not intended to limit the scope of protection of the present invention. Any modifications or equivalent arrangements readily accomplished by one skilled in the art fall within the scope of the present invention. The scope of protection of the present invention shall be determined by the scope of the patent application. 1: Physiological value sensing device 10: Input module 12: Light emitting unit 14: Light detection unit 16: Microstructure 20: Computation module 22: Processing submodule 24: Fitting submodule 26: Evaluation submodule 30: Storage module 40: Post-processing module 50: Output module Figure 1 is a schematic diagram illustrating changes in the relative infrared light absorption rates of various substances in human tissue; Figure 2 is a schematic diagram illustrating a physiological value sensing device according to the present invention; Figure 3 is a top-down schematic diagram illustrating several implementations of an input module in the physiological value sensing device according to the present invention; Figure 4 is a schematic diagram illustrating model fitting performed by a fitting submodule according to one embodiment of the present invention; Figure 5 is a schematic diagram illustrating the decision-making process of a decision tree according to one implementation of the present invention; and Figure 6 is a schematic diagram illustrating a flow chart of a physiological value sensing method according to the present invention. 1: Physiological value sensing device 10: Input module 20: Computing module 22: Processing submodules 24: Fitting submodule 26:Evaluation submodule 30: Storage module 40: Post-processing module 50: Output module

Claims

1. A physiological numerical sensing device, comprising: an input module for providing multi-band light to illuminate a subject to generate a plurality of optical signals, wherein the optical signals include a plurality of interference signals and a physiological signal to be measured, and the interference signals are formed by a plurality of interfering substances coexisting in human tissue, the interfering substances including water, lipids, fats, heme, and proteins; wherein the input module includes a plurality of light-emitting units and at least one light-detecting unit, the light-emitting units being used to provide the multi-band light and adjust the light intensity of the multi-band light, the at least one light-detecting unit being used to receive reflected light from the multi-band light after reflection by the subject, and to generate the optical signals based on the reflected light; and a processing module for, A normal physiological model is established, which has multiple sample physiological values ​​corresponding to the multi-band light. The normal physiological model is established by controlling three variables: the band of the multi-band light, the distance between the light-emitting units and the at least one light-detecting unit, and the light intensity of the multi-band light. The interference signals and the physiological signal to be measured are converted into the normal physiological model. The interference signals are eliminated based on the sample physiological values, and a physiological value to be measured corresponding to the physiological signal to be measured is generated.

2. The physiological value sensing device as claimed in claim 1, wherein the light-emitting units can provide multi-band light with a wavelength range of about 800 nanometers (nm) to 1700 nanometers (nm) and a wavelength interval of not less than 100 nanometers (nm).

3. The physiological value sensing device as claimed in claim 1, wherein the input module further includes a microstructure disposed adjacent to the at least one photodetector to prevent the multi-band light provided by the light-emitting units from being received by the at least one photodetector before being reflected by the subject.

4. The physiological value sensing device as described in claim 1, wherein the computing module includes a processing submodule, a fitting submodule, and an evaluation submodule. The processing submodule is used to process and convert the interference signals and the physiological signal to be measured to the normal physiological model. The fitting submodule is used to fit the converted interference signals and the physiological signal to be measured to the normal physiological model to generate a fitting result, and to generate the physiological value to be measured corresponding to the physiological signal to be measured after excluding the interference signals based on the physiological values ​​of the samples. The evaluation submodule is used to output a feature importance assessment based on the fitting result.

5. The physiological value sensing device as described in claim 4 further includes a storage module for storing the optical signals generated by the input module, the converted interference signals, the physiological signal to be measured, the physiological value to be measured, and the feature importance assessment.

6. The physiological value sensing device as described in claim 5 further includes a post-processing module for generating a long-term trend report based on the optical signals.

7. The physiological value sensing device as described in claim 6 further includes an output module for outputting the physiological value to be measured, the feature importance assessment, and the long-term trend report.

8. A method for sensing physiological values, comprising: providing an input module comprising a plurality of light-emitting units and at least one light-detecting unit, wherein the light-emitting units are configured to provide multi-band light to illuminate a subject and adjust the light intensity of the multi-band light, and the at least one light-detecting unit is configured to receive reflected light from the multi-band light after it has been reflected by the subject, and generate a plurality of optical signals based on the reflected light, wherein the optical signals include a plurality of interference signals and a physiological signal to be measured, and the interference signals are formed by a plurality of interfering substances coexisting in human tissue, the interfering substances including water, lipids, fats, heme, and proteins; establishing a normal physiological model, wherein the normal physiological model is established by controlling three variables: the wavelength of the multi-band light, the distance between the light-emitting units and the at least one light-detecting unit, and the light intensity of the multi-band light, and has a plurality of sample physiological values ​​corresponding to the multi-band light; The interference signals and the physiological signal to be measured are converted to the normal physiological model; the interference signals are eliminated based on the physiological values ​​of the samples; and a physiological value to be measured is generated corresponding to the physiological signal to be measured.

9. The physiological numerical sensing method as described in claim 8, wherein the multi-band light is provided by providing the multi-band light with a wavelength range of about 800 nanometers (nm) to 1700 nanometers (nm) and a wavelength interval of not less than 100 nanometers (nm).

10. The physiological value sensing method as described in claim 8 further includes providing an artificial intelligence model to fit the converted interference signals and the physiological signal to be measured with the sample physiological values ​​of the normal physiological model to generate a fitting result, and generating the physiological value to be measured corresponding to the physiological signal to be measured after eliminating the interference signals based on the fitting result.

Images