A non-invasive biochemical content detection method and system based on raman spectrum

Abstract

The present application relates to the technical field of biomedical detection, and discloses a noninvasive biochemical content detection method and system based on Raman spectrum, which obtains higher noninvasive biochemical measurement accuracy and smaller measurement error through four models. Specifically, model A is used to obtain the Raman peak after removing the water peak of the pure biochemical aqueous solution with a known gradient concentration as a candidate characteristic peak array; model B is used to obtain a characteristic peak parameter set of the biochemical gradient concentration according to the candidate characteristic peak array; model C is used to perform supervised learning and closed-loop training through the first skin Raman spectrum and the actual biochemical content obtained through blood testing, so as to obtain a characteristic peak parameter set of the skin biochemical; and model D is used to perform unsupervised learning and open-loop training through the second skin Raman spectrum and the characteristic peak parameter set, so as to obtain the biochemical content under the skin. The present application can solve the problem of characteristic peak superposition in the quantitative measurement of mixed substances in the prior art, and can achieve the purpose of accurately selecting dynamic characteristic peaks.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biomedical detection, and in particular to a non-invasive biochemical content detection method and system based on Raman spectrum. BACKGROUND

[0002] The biochemical gradient solution includes glucose, uric acid and progesterone, and the gradient concentration refers to a series of reagents with the concentration of the biochemical aqueous solution from low to high. The low concentration and the high concentration refer to the low or high content of the biochemical in the human body. For example, for glucose in blood, the content is usually 2.1 mmol / L to 30.0 mmol / L, and the gradient interval can be set according to the blood glucose measurement error, for example, 0.10 mmol / L. Blood test refers to the test of the venous blood collected by volunteers in the hospital laboratory using a biochemical analyzer. Non-invasive Raman (inspection) refers to the non-invasive Raman spectrum detection of the skin of the volunteers.

[0003] Raman spectrum is a scattered spectrum. Raman spectrum analysis is based on the Raman scattering effect discovered by Indian scientist C.V. Raman, which analyzes the scattered spectrum with different frequencies of incident light to obtain information about molecular vibration and rotation, and is applied to the analysis method of molecular structure research. Non-elastic scattering produced by the interaction of laser with molecular vibration, optical phonon in solid, etc. is called Raman scattering.

[0004] Currently, when using Raman spectrum to detect the content of biochemicals, there is a problem of overlapping of characteristic peaks in quantitative measurement of mixed substances, which makes it difficult to support accurate selection of dynamic characteristic peaks. SUMMARY

[0005] The present application provides a non-invasive biochemical content detection method and system based on Raman spectrum, to solve the problem of overlapping of characteristic peaks in quantitative measurement of mixed substances when using Raman spectrum to detect the content of biochemicals in the prior art, which makes it difficult to support accurate selection of dynamic characteristic peaks.

[0006] In order to achieve the above-mentioned purpose, the present application realizes the technical scheme as follows:

[0007] In a first aspect, the present application provides a non-invasive biochemical content detection method based on Raman spectrum, comprising:

[0008] S100: A set of known concentrations of biochemicals aqueous solutions with the same gradient of single pure biochemicals is configured, a Raman acquisition module is used to acquire each known concentration of biochemicals aqueous solution, more than one original Raman spectrum of pure biochemicals is acquired, pure water is acquired to acquire an original Raman spectrum of pure water, and a dehydrated pure biochemical candidate characteristic peak array of all concentrations is obtained based on the original Raman spectrum of all concentrations of biochemicals aqueous solutions, pure water, and a dehydrated pure biochemical algorithm;

[0009] S200: A characteristic peak parameter set corresponding to the dehydrated pure biochemical candidate characteristic peak array of all concentrations is obtained based on the dehydrated pure biochemical candidate characteristic peak array of all concentrations and a characteristic peak positioning algorithm;

[0010] S300: A first skin Raman spectrum of a volunteer's skin is acquired by using a Raman acquisition module, the first skin Raman spectrum is taken as an original Raman spectrum, the corresponding biochemical content in the blood test result of the volunteer is recorded, and a skin biochemical characteristic peak parameter set is obtained based on the characteristic peak parameter set, the biochemical content, and a skin biochemical content algorithm;

[0011] S400: A second skin Raman spectrum on the skin of a patient is acquired by using a Raman acquisition module, and the subcutaneous biochemical content and the measurement probability of the patient are obtained based on the skin biochemical characteristic peak parameter set, the second skin Raman spectrum, and a non-invasive biochemical content algorithm.

[0012] In a second aspect, the present application provides a non-invasive biochemical content detection system based on Raman spectrum, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the method of the first aspect when executing the computer program.

[0013] Advantages:

[0014] The non-invasive biochemical content detection method based on Raman spectrum provided by the present application constructs a characteristic peak parameter set corresponding to a dehydrated pure biochemical candidate characteristic peak array, and through dehydration processing, the detection accuracy can be improved. The Raman acquisition module is used to respectively acquire biochemical gradient solution, blood test, and non-invasive acquisition, and through a series of processing and analysis of the acquired Raman spectrum combined with the characteristic peak parameter set, the problem of characteristic peak superposition in the quantitative measurement of mixed substances in the prior art can be solved, and the purpose of supporting accurate selection of dynamic characteristic peaks can be achieved. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 One of the flowcharts of a non-invasive biochemical content detection method based on Raman spectrum according to a preferred embodiment of the present application;

[0016] Figure 2Figure 2 is a flow chart of a preferred embodiment of the present application, which is a non-invasive biochemical content detection method based on Raman spectrum;

[0017] Figure 3 Figure 3 is a simplified schematic diagram of a Raman peak group after baseline correction and noise elimination for a preferred embodiment of the present application;

[0018] Figure 4(a) is a Raman peak group of a biochemical solution, Figure 4(b) is a Raman peak group of pure water, and Figure 4(c) is a Raman peak group of pure biochemical after water removal;

[0019] Figure 5 Figure 5 is a schematic diagram of a Raman acquisition module for a preferred embodiment of the present application. DETAILED DESCRIPTION

[0020] The technical solutions of the present application will be described in detail below. Obviously, the described embodiments are only some of the embodiments of the present application, but not all. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0021] Unless otherwise defined, the technical terms or scientific terms used in the present application shall have the usual meanings understood by those skilled in the art. The terms "first", "second", and similar terms used in the present application do not represent any order, number, or importance, but are only used to distinguish different objects. Similarly, the terms "one" or "a" and similar terms do not represent a quantity limitation, but represent the existence of at least one. The terms "connected" or "connected" and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms "up", "down", "left", "right", and the like are only used to represent relative positional relationships, and when the absolute positions of the described objects change, the relative positional relationships also change accordingly.

[0022] It should be understood that when using Raman spectrum to detect biochemical content, there is a problem of overlapping characteristic peaks in quantitative measurement of mixed substances, which makes it difficult to support accurate selection of dynamic characteristic peaks. Based on this, the present application provides a non-invasive biochemical content detection method based on Raman spectrum, which can improve detection accuracy through dehydration processing, solve the problem of overlapping characteristic peaks in quantitative measurement of mixed substances in the prior art, and achieve the purpose of supporting accurate selection of dynamic characteristic peaks.

[0023] Please refer to Figures 1-2 The non-invasive biochemical content detection method based on Raman spectrum provided by the present application comprises:

[0024] S100: configuring a set of pure biochemicals with a plurality of known concentrations of biochemicals aqueous solutions in the same gradient, collecting each known concentration of biochemicals aqueous solution by using a Raman collection module, collecting more than one original Raman spectrum of pure biochemicals, collecting an original Raman spectrum of pure water, and obtaining a candidate characteristic peak array of pure biochemicals after water removal based on the original Raman spectrum of all concentrations of biochemicals aqueous solutions, the original Raman spectrum of pure water, and a pure biochemicals after water removal algorithm.

[0025] As shown in Figure 1 S100 can be regarded as constructing a model A, S200 can be regarded as constructing a model B, S300 can be regarded as constructing a model C, and S400 can be regarded as constructing a model D. In other words, the method of the present application specifically includes: proposing to obtain higher non-invasive biochemical measurement accuracy and smaller measurement error by using four models, wherein model A is used to obtain the Raman peak after water removal of the known gradient concentration pure biochemicals aqueous solution as a candidate characteristic peak array, model B is used to obtain a characteristic peak parameter set of biochemical gradient concentration according to the candidate characteristic peak array, model C is used to obtain a characteristic peak parameter set of skin biochemicals by supervised learning and closed-loop training based on the actual biochemical content obtained by the first skin Raman spectrum and blood test, and model D is used to obtain the biochemical content under the skin by unsupervised learning and open-loop training based on the second skin Raman spectrum and the characteristic peak parameter set.

[0026] In this step, the gradient interval in the same gradient can be set according to the blood glucose measurement error, for example, 0.10 mmol / L. Herein, only an example is given, and the gradient interval can also be adjusted within a certain range in other feasible embodiments.

[0027] In addition, it is worth explaining that since a plurality of concentrations of biochemicals aqueous solutions are collected, the number of original Raman spectra of pure biochemicals obtained is more than one.

[0028] S200: obtaining a characteristic peak parameter set corresponding to the candidate characteristic peak array of pure biochemicals after water removal based on all concentrations according to a characteristic peak positioning algorithm.

[0029] S300: collecting a first skin Raman spectrum of a volunteer's skin by using a Raman collection module, taking the first skin Raman spectrum as an original Raman spectrum, recording the biochemical content in the test result of the corresponding volunteer blood test, and obtaining a skin biochemical characteristic peak parameter set based on a characteristic peak parameter set, a test biochemical content, and a skin biochemical content algorithm.

[0030] In this step, the first skin Raman spectrum can also be referred to as skin Raman spectrum 1, and only an example is given herein without limitation.

[0031] S400: Collecting a second skin Raman spectrum on the patient's skin by using the Raman collection module, and obtaining the subcutaneous biochemical content and measurement probability of the patient based on the biochemical feature peak parameter set, the second skin Raman spectrum and the non-invasive biochemical content algorithm.

[0032] In this step, the second skin Raman spectrum can also be referred to as skin Raman spectrum 2, which is only an example and is not limited.

[0033] In this embodiment, the candidate feature peak array, the feature peak parameter set, the skin biochemical feature peak parameter set and the subcutaneous biochemical content are the corresponding working results of each step. It should be noted that, since the method provided in the present application can be mainly applied to an artificial intelligence working system, the candidate feature peak array, the feature peak parameter set, the skin biochemical feature peak parameter set and the subcutaneous biochemical content are dynamically and iteratively changed, rather than in a fixed mode.

[0034] It is worth emphasizing that in the present application, the input data at the input end of the Raman collection module includes but is not limited to: biochemical gradient solution, blood test and non-invasive Raman comparative collection, and non-invasive Raman collection.

[0035] The biochemical gradient solution refers to a biochemical substance that needs to be detected by the system of the present application, such as but not limited to glucose, uric acid and progesterone. The gradient concentration refers to a series of reagents from low to high in the concentration of the aqueous solution of the biochemical substance. It is emphasized here that the low concentration and the high concentration are respectively low to high in the content of the biochemical substance in the human body. For example, for glucose in the blood, the concentration is usually 2.1 mmol / L to 30.0 mmol / L. The gradient interval can be set with reference to the blood glucose measurement error, for example, 0.10 mmol / L.

[0036] The blood test in the blood test and non-invasive Raman comparative collection includes but is not limited to the test of the venous blood collected from the volunteer by the biochemical instrument in the hospital laboratory. The non-invasive Raman (inspection) refers to the non-invasive Raman spectrum detection of the skin of the volunteer. The present application works based on the artificial intelligence deep learning mode, so a large amount of blood test and non-invasive Raman comparative collection and experiments need to be done. It should be noted here that, since the human blood and tissue fluid are circulating and changing, like the blood pressure of a person, which is dynamic, it is emphasized in the present application that the blood test and the non-invasive Raman comparative collection must be performed simultaneously, in other words, while the volunteer is being drawn blood, the non-invasive Raman spectrum collection is being performed synchronously.

[0037] After the above work is completed, the system obtains the result of the artificial intelligence deep learning, and further, the non-invasive Raman spectrum detection of the patient can analyze and infer the biochemical content and its probability of the patient.

[0038] In other words, the non-invasive biochemical content detection method based on Raman spectrum provided in the application, when implemented, first obtains a gradient concentration dehydrated candidate characteristic peak array of a plurality of biochemicals. Further, a characteristic peak parameter set of a plurality of biochemicals is obtained. Still further, a first skin Raman spectrum of a large number of volunteers is obtained, and under supervised learning, a skin biochemical characteristic peak parameter set of a plurality of biochemicals is trained. Finally, after detecting a second skin Raman spectrum of a patient, the content of a biochemical can be calculated by an unsupervised learning method.

[0039] In the above steps, the dehydrated pure biochemical algorithm, the skin biochemical content algorithm and the skin biochemical content algorithm belong to supervised learning, and the non-invasive biochemical content algorithm belongs to unsupervised learning algorithm.

[0040] It should be pointed out here that in the detected non-invasive Raman spectrum, a large number of potential biochemical contents are included. In order to improve the accuracy of biochemical detection, in the application, each biochemical and its concentration is learned and trained one by one. In addition, as for the volunteers and patients mentioned in the application, the two are not particularly distinguished, but the former mainly refers to the collection object during supervised learning training, and the latter mainly refers to the collection object during unsupervised learning and non-invasive detection application.

[0041] The above-mentioned non-invasive biochemical content detection method based on Raman spectrum constructs a characteristic peak parameter set corresponding to the dehydrated pure biochemical candidate characteristic peak array. Through the dehydration processing, the detection accuracy can be improved. The Raman acquisition module is used for biochemical gradient solution acquisition, blood test and non-invasive acquisition respectively, and through a series of processing and analysis of the acquired Raman spectrum combined with the characteristic peak parameter set, the problem of characteristic peak superposition in quantitative measurement of mixed substances in the prior art can be solved, and the purpose of supporting accurate selection of dynamic characteristic peaks can be achieved.

[0042] Optionally, in a preferred embodiment, the S100 comprises:

[0043] S110: determining the gradient value of the gradient concentration according to the measured concentration range value and the measurement accuracy of the biochemical content required during the test;

[0044] S120: configuring a set of biochemical aqueous solutions of the same gradient of known concentrations of a single pure biochemical according to the gradient value, and obtaining the original Raman spectrum;

[0045] S130: for the original Raman spectrum, using a test algorithm to eliminate abnormal original Raman spectrum;

[0046] S140: The Raman spectrum after removing the abnormal value is subjected to baseline correction with the baseline strength of 0 by using the asymmetric least square method;

[0047] S150: The Raman spectrum after baseline correction is subjected to noise elimination by using the discrete wavelet transform method DWT, wherein the wavelet base function is db4-db6, the decomposition layer number is 7, and the soft threshold adjustment factor SA is 0.5-2.0;

[0048] S160: The Raman peak position RS with the Raman intensity RI>0 is selected as a Raman peak, and one Raman peak data includes the Raman peak position and the Raman intensity at the peak position; wherein one Raman spectrum corresponds to one Raman peak group, and one Raman peak group includes all the Raman peaks in the Raman spectrum, and all the Raman peak groups constitute a Raman peak array;

[0049] S170: The Raman peak array of the biochemical solution with all concentrations is sequentially subtracted by the Raman peak group of pure water to obtain a candidate characteristic peak array, and the candidate characteristic peak array includes the concentration, the detection number of the same concentration, and the result of the Raman peak group.

[0050] In this embodiment, the original Raman spectrum is a one-dimensional Raman spectrum line in a superimposed state, and the original Raman spectrum refers to the Raman spectrum obtained without correction, noise elimination and other processing. In the graph corresponding to the original Raman spectrum, it is difficult to distinguish the peak value and other information because it has not been processed.

[0051] Because there are various errors such as system error, gross error, environmental interference error during measurement, it is necessary to remove these abnormal original Raman spectra in multiple measurements. In this application, the number of times of measurement is considered, and tests including but not limited to T test and Z test are used to remove abnormal values. In measurement applications, these test methods are commonly known in the prior art, and will not be described here.

[0052] In an example, a simplified schematic diagram of the Raman peak group after baseline correction and noise elimination is shown in Figure 3 . Wherein the baseline is the horizontal axis, and the Raman intensity reference is 0, Figure 3 the wave in the graph is a Raman peak, the position of the horizontal axis is a Raman peak position, and the projection of the vertical axis on the Raman peak position is a Raman intensity. One Raman peak group includes all the Raman peaks obtained after processing one Raman spectrum.

[0053] It is worth pointing out that in this implementation, the specific values of the wavelet base function, the decomposition layer number and the soft threshold adjustment factor are only examples in this application and are not limited. In other feasible embodiments, they can be adjusted to a certain extent as long as the purpose of this application is achieved, but any transformation is within the scope of protection of this application.

[0054] In an example, Fig. 4(a) is a Raman peak group of a biochemical aqueous solution, Fig. 4(b) is a Raman peak group of pure water, and Fig. 4(c) is a Raman peak group of pure biochemical after water removal. In an ideal case, the probability of the Raman peak of the biochemical overlapping with the Raman peak of pure water in the Raman peak group of the single biochemical aqueous solution is small, so the water peak can be directly subtracted from the Raman peak group. In addition, in some biochemicals, due to the superposition of the Raman peak and the Raman peak of pure water, or the superposition of the Raman peaks between biochemicals, an overlay peak with complex waveforms is generated, in which case further analysis and processing are required.

[0055] In this embodiment, the detection accuracy can be improved by the water removal process.

[0056] Optionally, in a preferred embodiment, the S200 comprises:

[0057] S210: According to the candidate feature peak array, the first average value and the first standard deviation of the Raman peak position are calculated at the same concentration, the first standard deviation is sorted in descending order as a list RSEL, the second average value and the second standard deviation of the Raman intensity are calculated, and the second standard deviation is sorted in descending order as a list RIEL;

[0058] S220: Calculate the full Raman peak width, sort the peak width in descending order as a list WPL, and calculate the intensity of all Raman peaks, sort the intensity in descending order as a list RIL;

[0059] S230: Determine the feature peak by using the peak difference screening algorithm DBSCAN, and output the feature peak parameter set, which is specifically as follows:

[0060] The preset wavenumber tolerance δ is 1 to 10, and δ-RSEL, δ-RIEL, δ-WPL and δ-RIL are calculated after clustering according to RSEL, RIEL, WPL and RIL, respectively; wherein, δ-RSEL represents the probability of obtaining RSEL after clustering, δ-RIEL represents the probability of obtaining RIEL after clustering, δ-WPL represents the probability of obtaining WPL after clustering, and δ-RIL represents the probability of obtaining RIEL after clustering;

[0061] Select the Raman peak that is optimal at the same time of δ-RSEL, δ-RIEL, δ-WPL and δ-RIL as the feature peak, and output the feature peak parameter set, which includes the biochemical name, the biochemical concentration, the feature peak position, the feature peak intensity, the feature peak width, the feature peak number, the average value, and the standard deviation.

[0062] In this embodiment, DBSCAN is a density-based clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, DBSCAN) that can automatically find dense point clusters and identify noise points (i.e. unstable, isolated false peaks).

[0063] It is worth explaining that the purpose of screening characteristic peaks is to determine the type of biochemicals and calculate the content of biochemicals. For learning of artificial intelligence using a large amount of sample data, artificial intelligence can find a more optimized screening method according to the purpose of screening characteristic peaks. The method obtained according to this idea should fall within the protection scope of the present application.

[0064] In a specific implementation, an average peak difference algorithm can be used to determine characteristic peaks and output a characteristic peak parameter set, wherein:

[0065] Using the average peak difference algorithm, in more than one Raman peak array of the same concentration, the average value of Raman intensity is calculated at a Raman peak position between 1 and 10, and the difference between each Raman intensity and the average value is calculated. For a group of Raman peak groups, the difference values are arranged in ascending order and numbered, and the number is associated with the Raman peak position. The Raman peak positions in the top 0% to 50% of the arrangement are selected as candidate characteristic peaks. For the array of candidate Raman peaks of all concentrations, the step of arranging the difference values in ascending order and numbering is performed, and the Raman peak array containing the number is output. The Raman peaks with numbers between 2 and 20 are selected as characteristic peaks, and a characteristic peak parameter set including the biochemical name, biochemical concentration, characteristic peak position, characteristic peak intensity, characteristic peak width, characteristic peak number, average value, and standard deviation is output.

[0066] In this implementation, the range of specific values is only an example and is not limited. In other feasible implementations, the range of candidate characteristic peaks can be adjusted appropriately, the number range of selected characteristic peaks can be adjusted appropriately, and other ways of numbering and marking can be used to distinguish, but any transformation is within the protection scope of the present application.

[0067] Optionally, in a preferred embodiment, the S300 includes:

[0068] S310: According to the skin type, recruit more than one volunteer for each skin type;

[0069] S320: For all volunteers, while blood tests are taken, a Raman acquisition module is used to synchronously acquire more than one first skin Raman spectrum of the volunteers. The blood test items include the content of specified biochemicals, and the specified biochemical content in the blood test results is recorded.

[0070] S330: taking the first skin Raman spectrum of the volunteers as the original Raman spectrum, performing the steps S130-S160, and outputting the first skin Raman peak array;

[0071] S340: taking the first skin Raman peak array to perform the step S170, and obtaining the candidate characteristic peak array of the dehydrated skin of the volunteers;

[0072] S350: taking the candidate characteristic peak array of the same volunteer, and converting the multi-dimensional candidate characteristic peak array into a one-dimensional candidate characteristic peak array by using the average algorithm;

[0073] S360: according to the characteristic peak parameter set of the biochemical substance name to be detected and the content of the biochemical substance, setting the peak position weight function according to each characteristic peak number, and training and generating the skin biochemical characteristic peak parameter set by using the DBSCAN screening algorithm, wherein the skin biochemical characteristic peak parameter set includes the skin type, the biochemical substance name, the peak position weight function value, and the biochemical content concentration.

[0074] The skin type referred to in the present application includes but is not limited to different skin colors, outdoor workers, indoor workers, and the like, which aims to pre-classify the Raman spectrum collected on the skin. Due to the diversity of human skin, such pre-classification of the skin is beneficial to eliminate the skin noise.

[0075] The number of the volunteers collected is greater than a set threshold value, and the set threshold value can be an integer greater than 50.

[0076] Further, after obtaining the first skin Raman spectrum of the volunteers, the first skin Raman spectrum is taken as the original spectrum, the first skin Raman peak array corresponding to the volunteers is obtained by using the same manner as the steps S130-S160, and the first skin Raman peak array of all different concentrations is sequentially subtracted from the Raman peak group of pure water to obtain the candidate characteristic peak array, wherein the candidate characteristic peak array includes the concentration, the detection number of the same concentration, and the result of the Raman peak group. In other words, the processing manner of the first skin Raman spectrum of the volunteers in the present application is consistent with the processing manner of the original Raman spectrum.

[0077] It is worth explaining that the DBSCAN screening algorithm is a general screening algorithm, which can output the required result, such as the biochemical content concentration, only by giving it appropriate input information and debugging appropriate parameters. It is worth emphasizing that the volunteers in the present application mainly refer to the collection objects in the supervised learning training, in other words, the main purpose of the steps in the embodiment is to obtain appropriate parameters based on the information of the volunteers.

[0078] Optionally, in a preferred embodiment, the S400 comprises:

[0079] S410: taking the second skin Raman spectrum as the original Raman spectrum, performing the steps S130-S160, and outputting a second skin Raman peak array;

[0080] S420: performing the step S170 using the second skin Raman peak array to obtain a candidate characteristic peak array of the dehydrated skin of the patient;

[0081] S430: searching for a matching target array in the skin biochemical characteristic peak parameter set according to the candidate characteristic peak array of the dehydrated skin of the patient, obtaining a target array closest to the skin biochemical characteristic peak parameter set, and outputting the skin type, biochemical name, and biochemical content concentration of the target array.

[0082] Specifically, the S430 includes:

[0083] According to the candidate characteristic peak array of the dehydrated skin of the patient, searching for a matching target array in the skin biochemical characteristic peak parameter set, using a DBSCAN screening algorithm to find a target array closest to the skin biochemical characteristic peak parameter set, and outputting the skin type, biochemical name, and biochemical content concentration of the target array.

[0084] Further, after obtaining the second skin Raman spectrum of the patient, taking it as the original spectrum, obtaining the second skin Raman peak array corresponding to the patient in a manner consistent with the foregoing steps S130-S160, and further subtracting the Raman peak group of pure water from all different concentrations of the second skin Raman peak array to obtain a candidate characteristic peak array, wherein the candidate characteristic peak array includes concentration, same concentration detection number, and Raman peak group result. In other words, the processing manner of the second skin Raman spectrum of the patient in the present application is consistent with the processing manner of the foregoing original Raman spectrum.

[0085] It is worth emphasizing that the patient in the present application mainly refers to the collection object in the unsupervised learning and non-invasive detection application, in other words, the main purpose of the steps in this embodiment is to identify the information of the patient by using the appropriate parameters obtained by training based on the foregoing information of the volunteers.

[0086] In an example, as shown in Figure 5 The Raman collection module specifically includes a primary light path and a secondary light path, wherein the primary light path includes a laser, a narrow-band filter, a primary condenser lens, a primary optical drill, a primary optical drill head, a half-mirror, and a collection lens connected in sequence;

[0087] The secondary light path shares the collection lens and the half-mirror with the primary light path, and the secondary light path includes a secondary condenser lens arranged between the primary condenser lens and a tertiary condenser lens, a low-pass filter arranged between the primary condenser lens and the secondary condenser lens, a secondary optical drill, and a spectrometer.

[0088] Specifically, the parameters of the Raman collection module described above include:

[0089] The output laser wavelength of the laser adopts an infrared band, including 500nm to 1200nm, and the light emitting power of the laser is between 50mW and 2000mW;

[0090] The pass wavelength of the narrowband filter and the low-pass filter matches the output wavelength of the laser, and the cutoff intensity of the filter is greater than or equal to OD6;

[0091] The spectral displacement resolution of the spectrometer is less than or equal to 10 wave numbers, the lateral resolution of the spectroscope and the photoelectric sensor is greater than 1000, the photoelectric sensor works in a refrigeration mode, the refrigeration temperature is lower than -10℃, and the integration time is between 0.1s and 3600s.

[0092] In an example, the working steps of the Raman collection module are as follows:

[0093] The laser emitted by the laser passes through a primary light path to form Raman excitation light, is focused by the collection lens to the human skin at the irradiation point, and then is detected by the collection lens and the half-mirror in the secondary light path to send the Raman scattered light around the human skin at the irradiation point to the spectrometer, and the photoelectric IO interface of the spectrometer receives the Raman scattered light, decomposes and converts it into a digital signal output of the line array or the surface array Raman spectrum, wherein the array number of the line array is equal to the lateral resolution of the photoelectric sensor.

[0094] The application also provides a non-invasive bio-chemical content detection system based on Raman spectrum, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the non-invasive bio-chemical content detection method based on Raman spectrum when executing the computer program. The non-invasive bio-chemical content detection system based on Raman spectrum can realize each embodiment of the non-invasive bio-chemical content detection system based on Raman spectrum described above, and achieve the same beneficial effects, which will not be described here.

[0095] The preferred embodiments of the application are described in detail above. It should be understood that those skilled in the art can make many modifications and changes without creative labor based on the concept of the application. Therefore, any technical solution obtained by logical analysis, reasoning or limited experiment based on the prior art according to the concept of the application shall be within the protection scope determined by the claims.

Claims

1. A non-invasive method for detecting biochemical content based on Raman spectroscopy, characterized in that, include: S100: Configure a group of known concentrations of a single pure biochemical solution with the same gradient. Use a Raman acquisition module to acquire the original Raman spectrum of each known concentration of the biochemical solution and acquire the original Raman spectrum of the pure biochemical solution. Then, acquire the original Raman spectrum of the pure water. Based on the total concentrations of the biochemical solutions, the original Raman spectrum of the pure water, and the dehydrated pure biochemical algorithm, obtain the candidate characteristic peak array of the dehydrated pure biochemical solution of all concentrations. S200: Based on the candidate characteristic peak array of dehydrated pure biochemicals of all concentrations and the characteristic peak localization algorithm, the characteristic peak parameter set corresponding to the candidate characteristic peak array of dehydrated pure biochemicals of all concentrations is obtained; S300: The Raman acquisition module is used to acquire the first skin Raman spectrum of the volunteer's skin. The first skin Raman spectrum is used as the original Raman spectrum. The content of biochemical substances in the corresponding blood test results of the volunteer is recorded. The characteristic peak parameter set of skin biochemical substances is obtained based on the characteristic peak parameter set, the content of biochemical substances in the test and the skin biochemical content algorithm. S400: The Raman acquisition module is used to acquire the second skin Raman spectrum on the patient's skin. Based on the skin biochemical characteristic peak parameter set, the second skin Raman spectrum and the non-invasive biochemical content algorithm, the patient's subcutaneous biochemical content and measurement probability are obtained.

2. The method according to claim 1, characterized in that, S100 includes: S110: Determine the gradient value of the gradient concentration based on the required concentration range and measurement accuracy of the biochemical content during the test; S120: Prepare a set of known concentrations of a single pure biochemical substance with the same gradient according to the gradient value, and obtain the original Raman spectrum; S130: For the raw Raman spectra, a verification algorithm is used to remove abnormal raw Raman spectra; S140: The asymmetric least squares method is used to perform baseline correction with a baseline intensity of 0 on the Raman spectrum after anomaly removal; S150: The discrete wavelet transform (DWT) method is used to reduce noise in the baseline-corrected Raman spectrum. The wavelet basis functions are db4-db6, the number of decomposition layers is 7, and the soft threshold adjustment factor SA is 0.5-2.

0. S160: Select the Raman peak position RS with Raman intensity RI > 0 as Raman peak. A Raman peak data includes the Raman peak position and the Raman intensity at the peak position. Among them, one Raman spectrum corresponds to one Raman peak group, and one Raman peak group includes all Raman peaks in the Raman spectrum. All Raman peak groups constitute a Raman peak array. S170: Subtract the Raman peak group of pure water from the Raman peak array of all concentrations of biochemical solutions to obtain a candidate characteristic peak array. The candidate characteristic peak array includes the concentration, the detection number of the same concentration, and the result of the Raman peak group.

3. The method according to claim 1, characterized in that, S200 includes: S210: Based on the candidate characteristic peak array, at the same concentration, calculate the first average value and the first standard deviation of the Raman peak positions, sort the first standard deviations from smallest to largest as a list RSEL, calculate the second average value and the second standard deviation of the Raman intensity, and sort the second standard deviations from smallest to largest as a list RIEL; S220: Calculate the peak width of all Raman peaks, sort the peak widths from smallest to largest into a list WPL, calculate the intensity of all Raman peaks, and sort the intensity from largest to smallest into a list RIL; S230: The DBSCAN peak difference filtering algorithm is used to determine the characteristic peaks and output the characteristic peak parameter set, as follows: With a pre-defined wavenumber tolerance δ of 1 to 10, δ-RSEL, δ-RIEL, δ-WPL, and δ-RIL are calculated based on RSEL, RIEL, WPL, and RIL, respectively, after peak clustering. Wherein, δ-RSEL represents the probability of obtaining RSEL after clustering, δ-RIEL represents the probability of obtaining RIEL after clustering, δ-WPL represents the probability of obtaining WPL after clustering, and δ-RIL represents the probability of obtaining RIEL after clustering. The Raman peaks that are simultaneously optimal for δ-RSEL, δ-RIEL, δ-WPL, and δ-RIL are selected as characteristic peaks, and a set of characteristic peak parameters is output. The set of characteristic peak parameters includes biochemical name, biochemical concentration, characteristic peak position, characteristic peak intensity, characteristic peak width, characteristic peak number, average value, and standard deviation.

4. The method according to claim 2, characterized in that, The S300 includes: S310: Based on skin type, recruit at least one volunteer for each skin type; S320: For all volunteers, while each volunteer has a blood test, a Raman acquisition module is used to simultaneously acquire one or more sets of first skin Raman spectra from the volunteer's skin. The blood test items include the content of specified biochemicals, and the content of specified biochemicals in the blood test results is recorded. S330: Take one or more sets of first skin Raman spectra as the original Raman spectra, execute steps S130 to S160, and output the first skin Raman peak array; S340: Perform step S170 using the first skin Raman peak array to obtain the candidate feature peak array of the dehydrated skin of the volunteer; S350: For the candidate feature peak array of the same volunteer, the averaging algorithm is used to transform the multidimensional candidate feature peak array into a one-dimensional candidate feature peak array. S360: For the characteristic peak parameter set of the biochemical name to be detected and the biochemical content to be tested, according to the peak number of each characteristic peak, a peak weight function is set, and the DBSCAN screening algorithm is used to train and generate the skin biochemical characteristic peak parameter set. The skin biochemical characteristic peak parameter set includes skin type, biochemical name, peak weight function value, and biochemical content concentration.

5. The method according to claim 1 or 4, characterized in that, The S400 includes: S410: Using the second skin Raman spectrum as the original Raman spectrum, execute steps S130 to S160 to output the second skin Raman peak array; S420: Perform step S170 using the second skin Raman peak array to obtain the candidate feature peak array of the patient's dehydrated skin; S430: Based on the candidate feature peak array of the patient's dehydrated skin, search for the matching target array in the skin biochemical feature peak parameter set, obtain the target array that is closest to the skin biochemical feature peak parameter set, and output the skin type, biochemical name, and biochemical concentration of the target array.

6. The method according to claim 5, characterized in that, S430 includes: Based on the candidate feature peak array of the patient's dehydrated skin, a matching target array is retrieved from the skin biochemical feature peak parameter set. The DBSCAN filtering algorithm is used to find the target array that is closest to the skin biochemical feature peak parameter set, and the skin type, biochemical name, and biochemical concentration of the target array are output.

7. The method according to claim 1, characterized in that, The Raman acquisition module specifically includes a primary optical path and a secondary optical path. The primary optical path includes a laser, a narrowband filter, a primary condenser lens, an optical fiber, a primary optical fiber head, a semi-reflective mirror, and an acquisition lens connected in sequence. The secondary optical path shares the acquisition lens and the semi-reflective half-lens with the primary optical path, and the secondary optical path includes a secondary condenser lens disposed between the primary condenser lens and the tertiary condenser lens, a low-pass filter disposed between the primary condenser lens and the secondary condenser lens, a secondary optical fiber, and a spectrometer.

8. The method according to claim 7, characterized in that, The parameters of the Raman acquisition module include: The laser outputs an infrared wavelength ranging from 500nm to 1200nm, and its luminous power is between 50mW and 2000mW. The release wavelengths of narrowband and low-pass filters are matched with the output wavelength of the laser, and the cutoff intensity of the filters is greater than or equal to OD6. The spectrometer has a spectral shift resolution of less than or equal to 10 wavenumbers, a lateral resolution of greater than 1000 for the spectroscope and photoelectric sensor, and the photoelectric sensor operates under cooling conditions with a cooling temperature below -10℃. The integration time is between 0.1s and 3600s.

9. A non-invasive biochemical content detection system based on Raman spectroscopy, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

Images