A non-invasive blood glucose near-infrared spectroscopy data correction method based on spectral characteristic peaks

Abstract

The present invention mainly relates to the field of near-infrared spectroscopy. In order to solve the problems of easy deviation, weak stability and low accuracy of spectral data in practical applications of near-infrared non-invasive blood glucose detection, the present invention provides a non-invasive blood glucose near-infrared spectral data correction method based on spectral characteristic peaks, the core idea of ​​which is: first, use a near-infrared spectrometer to collect blood sample spectral data, and calculate the second-order derivative of the blood sample spectral data to obtain spectral characteristic peaks, set the spectral characteristic peak point as the reference correction point, and then rearrange the wavelength points according to the number of wavelength points and the number of reference points of the near-infrared spectrometer to obtain the reference array, write the light intensity value of the blood sample spectral data reference correction point into the reference array to calculate the array correction coefficient, and finally correct the blood sample spectral data in combination with the array correction coefficient to complete the data correction. The method described in the present invention is mainly suitable for data correction of miniaturized near-infrared spectrometers.

Description

Technical Field

[0001] The invention relates to the technical field of near infrared spectroscopy, and in particular to a non-invasive blood sugar near infrared spectroscopy data correction method based on spectral characteristic peaks. Background Art

[0002] Current blood glucose testing methods include non-invasive, minimally invasive and invasive. Among them, invasive and minimally invasive testing have higher accuracy and can meet the requirements of clinical blood glucose testing. However, both invasive and minimally invasive testing will bring pain and discomfort to diabetic patients. Therefore, the demand for non-invasive blood glucose testing has emerged.

[0003] Among many non-invasive blood glucose detection methods, the miniaturized near-infrared detection method has become the most widely used detection method due to its simple equipment, high efficiency and cost advantages. However, due to the limitations of its own hardware conditions, the spectral data stability of the miniaturized near-infrared spectrometer is weak and the accuracy is low. Especially in the actual application of miniaturized near-infrared non-invasive blood glucose detection, the spectral data acquisition process requires extremely precise, the acquisition process is not standardized, the equipment is not in uniform contact with the skin of the person being tested during acquisition, and the acquisition environment changes are very likely to cause the spectral data to shift, thereby affecting the prediction ability of the spectral model and reducing the detection accuracy of the non-invasive blood glucose detection technology. Therefore, it is urgent to find a spectral data correction method to solve the problems of unstable spectral data and low accuracy in miniaturized near-infrared non-invasive blood glucose detection technology. Summary of the invention

[0004] The technical problems to be solved by the present invention are:

[0005] A non-invasive blood glucose near-infrared spectral data correction method based on spectral characteristic peaks is provided to solve the problem of spectral data deviation caused by weak data stability and low precision of near-infrared spectrometer in traditional non-invasive blood glucose collection, which affects the prediction ability of the spectral model and has low detection accuracy.

[0006] The technical solution adopted by the present invention to solve the above technical problems is as follows:

[0007] A non-invasive blood glucose near-infrared spectroscopy data correction method based on spectral characteristic peaks comprises the following steps:

[0008] Step 1: Collect spectral data of blood samples using a near-infrared spectrometer;

[0009] Step 2: Calculate the second-order derivative of the collected blood sample spectrum data to obtain the spectrum characteristic peak, and set the spectrum characteristic peak as the reference correction point;

[0010] Step 3: Select a spectral device according to the peak wavelength point position of the characteristic peak of the blood sample spectrum, and rearrange the wavelength points of the blood sample spectrum data according to the number of wavelength points and the number of reference points of the spectral device to obtain a reference array;

[0011] Step 4: Write the light intensity value of the reference correction point of the blood sample spectrum data into the reference group number to calculate the correction coefficient;

[0012] Step 5: Correct the spectral data of the blood sample to be corrected in combination with the correction coefficient.

[0013] Furthermore, the specific method of step 1 is: using a spliced ​​multi-sensor near-infrared spectroscopic device to collect near-infrared spectral data of the blood sample, collecting spectra multiple times for each single collection point to obtain multiple spectral data, calculating the average of the multiple spectral data, and using the averaged single spectral data as the actual spectral data of the blood sample at the collection point.

[0014] Furthermore, the specific method of step 3 is: select M spectral characteristic peak points as reference correction points, insert non-reference correction points into two adjacent reference correction points on average, rearrange the blood sample spectral data, and the wavelength point array obtained by rearrangement is the reference array.

[0015] Furthermore, the specific method of step 4 is: the light intensity values ​​of the M reference correction points are P1, P2, ..., P M , write the light intensity values ​​of the M reference correction points into the reference array, and calculate the light intensity value of the non-reference wavelength point inserted between the reference correction points P1 and P2 in the first row, and obtain its light intensity value:

[0016]

[0017] Where a is the ath wavelength point in the first row of the reference array, and N is the number of wavelength points contained in the first row;

[0018] The light intensity values ​​at all wavelength points in the reference array except the reference correction points can be calculated using the light intensity values ​​of the remaining reference correction points in the same calculation method;

[0019] To calculate the correction coefficients for each wavelength point in the reference array, it is necessary to first calculate the mean light intensity P of the M reference correction points. mean =(P1+P2+……+P M ) / M, taking the N wavelength points in the first row of the reference array as an example, the corresponding array correction coefficients are:

[0020]

[0021] Among them, (α1, α2,…α N-1 , α N ) are respectively the array correction coefficients corresponding to the N wavelength points in the first row of the reference array;

[0022] The correction coefficients corresponding to all wavelength points in the reference array can be calculated using the same calculation method.

[0023] Furthermore, the specific method of step 5 is: if the light intensity values ​​of the N wavelength points in the first row of the blood spectrum to be corrected are (P1, X2, X3, ... X N-1 , P2), the corresponding correction coefficients are (α1, α2, ... α N-1 , α N ), combined with the correction coefficient, the light intensity values ​​of the N wavelength points in the first row are corrected. The corrected light intensity values ​​are (P1 / α1, X2 / α2, X3 / α3, ... X N-1 / α N-1 , P2 / α N ), correct all wavelength points in the reference array in the same way, reorganize the reference array into a single-row spectrum, and then arrange the single-row spectrum from left to right in the order of wavelength value from small to large, and the reorganized single-row spectrum data is the corrected blood sample spectrum data;

[0024] The light intensity values ​​at the M spectral reference correction points are all the same value, and when the spectral data is reorganized, only a single spectral reference correction point needs to be retained.

[0025] Beneficial effects of the present invention:

[0026] The non-invasive blood glucose near-infrared spectral data correction method based on spectral characteristic peaks described in the present invention corrects the blood sample spectral data in the near-infrared non-invasive blood glucose detection technology through multiple spectral characteristic peak reference points, thereby solving the problems of easy deviation, weak stability and low accuracy of spectral data in the actual application of miniaturized near-infrared non-invasive blood glucose detection, and effectively improving the data stability and detection accuracy of the miniaturized near-infrared non-invasive blood glucose detection technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The present invention is a schematic diagram of a non-invasive blood glucose near-infrared spectroscopy data correction method based on spectral characteristic peaks.

[0028] Figure 2 3 is a graph of second-order derivative spectral data of blood sample spectral data in the embodiment.

[0029] Figure 3 Schematic diagram of the arrangement of the reference array described in the embodiment. DETAILED DESCRIPTION

[0030] like Figure 1As shown, the non-invasive blood glucose near-infrared spectral data correction method based on spectral characteristic peaks described in the present invention first uses a near-infrared spectrometer to non-invasively collect blood sample spectral data, and performs a second-order derivative on the blood sample spectral data to obtain spectral characteristic peaks, and sets the spectral characteristic peak point as a reference correction point, and then rearranges the wavelength points according to the number of wavelength points and the number of reference points of the spectrometer to obtain a reference array, writes the light intensity value of the reference correction point of the blood sample spectral data into the reference array to calculate the array correction coefficient, and finally corrects the non-invasively collected blood sample spectral data in combination with the array correction coefficient to complete the data correction.

[0031] Example:

[0032] In this embodiment, a spliced ​​multi-sensor miniaturized near-infrared spectrometer is used to non-invasively collect near-infrared spectral data of a blood sample of a person being tested by attaching it to the skin of a human body. A total of 4 spectral data are collected for a single collection point, that is, 4 spectral data are obtained, and the 4 spectral data are averaged. The single spectral data after averaging is the actual spectral data of the blood sample at the collection point. The commonly used single-sensor miniaturized near-infrared spectrometer covers a band range of 400 nanometers. The spliced ​​multi-sensor miniaturized near-infrared spectrometer used in this embodiment includes 4 single sensors, and the band ranges are 800nm ​​to 1200nm, 1200nm to 1600nm, 1600nm to 2000nm, and 2000nm to 2400nm, respectively. The single sensor device that initially only covers a band range of 400 nanometers can be expanded to cover 1600 nanometers, which almost covers the entire near-infrared band range.

[0033] The half-width of the second-order derivative spectrum is only about 1 / 3 of the half-width of the original spectrum, which can easily distinguish the small shoulder peaks on both sides of the strong peak, and is very effective in correctly determining the peak position and shoulder peak position. The peak of the blood sample spectrum, that is, the wavelength point position of the characteristic peak of the blood sample spectrum, can be clearly distinguished through the second-order derivative. In this embodiment, the collected blood sample spectrum data is subjected to SG second-order derivative, where the half-window width is set to 5 and the highest order of the polynomial is set to 5, such as Figure 2 As shown in FIG. 1 , the most obvious spectrum peak appears at a wavelength of 980 nm, and the peak point is the spectrum characteristic peak of the blood sample in the near-infrared band. The spectrum characteristic peak is set as the reference correction point.

[0034] According to the peak wavelength point position of the characteristic peak of the blood sample spectrum, a single sensor miniaturized near-infrared spectroscopy device is selected as the actual acquisition device. According to the number of wavelength points and the number of reference points of the spectral device, the wavelength point of the blood sample spectrum data is rearranged, and the wavelength point array obtained by rearrangement is the reference array. In this embodiment, the wavelength range of the characteristic peak of the blood sample spectrum is 980nm, and a single sensor miniaturized near-infrared device with a sensor band range of 800nm ​​to 1200nm is selected as the actual acquisition device. In the actual application of miniaturized near-infrared non-invasive blood glucose detection, if the spectral data of the blood sample contains too few wavelength points, it is easy to cause the problem that the collected spectral data contains fewer available signals, large randomness, and poor stability, thereby affecting the accuracy of the miniaturized near-infrared non-invasive blood glucose analysis technology. Therefore, in actual applications, usually a single spectral data of a blood sample contains more than 40 wavelength points. By collecting blood spectral data with more wavelength points, the spectral data is smoother, ensuring that there are enough available signals in the spectral data, effectively improving the stability and accuracy of the spectral data, while too many wavelength points have a small effect on spectral analysis, affecting the efficiency of spectral analysis. Therefore, the resolution of the miniaturized near-infrared spectroscopy device used in this embodiment is set to 10nm. In addition to the spectral characteristic peak points, it contains 40 wavelength points, and the corresponding band ranges are (800nm, 810nm, ... 970nm, 990nm, ... 1200nm). For the spectral characteristic peak points, that is, the reference correction points, the number can be selected to be M, where (2≤M≤41). Similarly, too many reference correction points will affect the spectral analysis efficiency, and too few reference correction points will increase the spectral analysis error. In this embodiment, 5 spectral reference correction points are selected to correct the blood sample spectral data, such as Figure 3 As shown, 01 / 02 / 03 / 04 / 05 are 5 spectral reference correction points respectively. The spectral data of the blood sample are rearranged, and the 40 non-reference correction points are rearranged by averaging and inserting them into two adjacent reference correction points. The wavelength point array obtained by the rearrangement is the reference array.

[0035] In this embodiment, the light intensity values ​​of the five reference correction points are written into the reference array, and the light intensity values ​​of the five median points are set to be P1, P2, P3, P4, and P5, respectively, where P1 is the light intensity value of the reference correction point 01 in the reference array, P2 is the light intensity value of the reference correction point 02 in the reference array, P3 is the light intensity value of the reference correction point 03 in the reference array, P4 is the light intensity value of the reference correction point 04 in the reference array, and P5 is the light intensity value of the reference correction point 05 in the reference array. The light intensity values ​​of the wavelength points in the first row of the reference array except the reference correction points are calculated, and the light intensity values ​​are obtained as follows:

[0036]

[0037] Where a is the ath wavelength point in the 1st row of the reference array.

[0038] Similarly, the light intensity values ​​of each wavelength point in the second row of the reference array except the reference correction point are calculated, and the light intensity values ​​are obtained as follows:

[0039]

[0040] Where b is the bth wavelength point in the 2nd row of the reference array.

[0041] Similarly, the light intensity values ​​at all wavelength points in the reference array except the reference correction point can be calculated, and then these light intensity values ​​are further matched with each wavelength point one by one and written into the reference array.

[0042] When calculating the array correction coefficient, first calculate the average light intensity P of the five reference correction points. mean =(P1+P2+P3+P4+P5) / 5, and then calculate the array correction coefficient for each wavelength point in the reference array one by one. Taking the 12 wavelength points in the first row of the reference array (including the reference correction points) as an example, the corresponding array correction coefficients are:

[0043]

[0044] Among them, (α1, α2,…α 11 , α 12 ) are the array correction coefficients corresponding to the 12 wavelength points in the first row of the reference array.

[0045] Similarly, the array correction coefficients corresponding to all wavelength points in the reference array can be calculated.

[0046] In this embodiment, the 12 wavelength points in the first row of the reference array are used as an example to correct the spectrum data of the non-invasively collected blood sample. If the light intensity values ​​of the 12 wavelength points in the first row of the reference array in the spectrum data of the non-invasively collected blood sample are (P1, X2, X3, ... X 11 , P2), and the reference array coefficients corresponding to the 12 wavelength points in the first row of the reference array are (α1, α2, ... α 11 , α 12 ), combined with the array correction coefficient, the light intensity values ​​of the 12 wavelength points in the first row are corrected one by one, and the corrected light intensity values ​​of the 12 wavelength points in the first row are (P1 / α1, X2 / α2, X3 / α3, ... X 11 / α 11 , P2 / α 12 ).

[0047] Similarly, the same spectral data correction method is used to correct the light intensity values ​​at all wavelength points in the reference array one by one. After the correction of the light intensity value data of all wavelength points is completed, the corrected light intensity values ​​are matched with the wavelength points one by one. After the matching is completed, the new reference array containing the corrected light intensity values ​​is reorganized again. Because the light intensity values ​​at the 5 spectral reference correction points are the same value after the spectral data correction is completed, therefore, when reorganizing the spectral data, only a single spectral reference correction point needs to be retained, that is, the (12*4) reference array is reorganized into a single row of spectral data (1*41), and then the single row of spectral data is arranged from left to right from small to large wavelength values ​​to obtain a reorganized single row of spectral data, the single row of spectral data contains 41 wavelength points, and the wavelength point distribution method is (800nm, 810nm, ... 970nm, 980nm, 990nm, ... 1200nm), and the reorganized single row of spectral data is the non-invasively collected blood sample spectral data after the correction.

Claims

1. A non-invasive blood glucose near-infrared spectroscopy data correction method based on spectral characteristic peaks, characterized in that: The following steps are involved: Step 1: Collect spectral data of blood samples using a near-infrared spectrometer; Step 2: Calculate the second-order derivative of the collected blood sample spectrum data to obtain the spectrum characteristic peak, and set the spectrum characteristic peak as the reference correction point; Step 3: Select a spectral device according to the peak wavelength point position of the characteristic peak of the blood sample spectrum, and rearrange the wavelength points of the blood sample spectrum data according to the number of wavelength points and the number of reference points of the spectral device to obtain a reference array; The specific method of step 3 is: select M spectral characteristic peak points as reference correction points, insert the non-reference correction points into two adjacent reference correction points on average, rearrange the blood sample spectral data, and the wavelength point array obtained by rearrangement is the reference array Step 4: Write the light intensity value of the reference correction point of the blood sample spectrum data into the reference group number to calculate the correction coefficient; The specific method of step 4 is: the light intensity values ​​of the M reference correction points are P1, P2, ..., P M , write the light intensity values ​​of the M reference correction points into the reference array, and calculate the light intensity value of the non-reference wavelength point inserted between the reference correction points P1 and P2 in the first row, and obtain its light intensity value: Where a is the ath wavelength point in the first row of the reference array, and N is the number of wavelength points contained in the first row; The light intensity values ​​at all wavelength points in the reference array except the reference correction points can be calculated using the light intensity values ​​of the remaining reference correction points in the same calculation method; To calculate the correction coefficients for each wavelength point in the reference array, it is necessary to first calculate the mean light intensity P of the M reference correction points. mean =(P1+P2+……+P M ) / M, for the N wavelength points in the first row of the reference array, the corresponding array correction coefficients are: Among them, (α1, α2,…α N-1 , α N ) are respectively the array correction coefficients corresponding to the N wavelength points in the first row of the reference array; The correction coefficients corresponding to all wavelength points in the reference array are calculated using the same calculation method; Step 5: Correct the spectral data of the blood sample to be corrected in combination with the correction coefficient.

2. The method for correcting non-invasive blood glucose near-infrared spectroscopy data based on spectral characteristic peaks according to claim 1, characterized in that: The specific method of step 1 is: use a spliced ​​multi-sensor near-infrared spectroscopic device to collect near-infrared spectral data of the blood sample, collect spectra multiple times for each single collection point to obtain multiple spectral data, calculate the average of the multiple spectral data, and use the averaged single spectral data as the actual spectral data of the blood sample at the collection point.

3. The method for correcting non-invasive blood glucose near-infrared spectroscopy data based on spectral characteristic peaks according to claim 1, characterized in that: The specific method of step 5 is: if the light intensity values ​​of the N wavelength points in the first row of the blood spectrum to be corrected are (P1, X2, X3, ... X N-1 , P2), the corresponding correction coefficients are (α1, α2, ... α N-1 , α N ), combined with the correction coefficient, the light intensity values ​​of the N wavelength points in the first row are corrected. The corrected light intensity values ​​are (P1 / α1, X2 / α2, X3 / α3, ... X N-1 / α N-1 , P2 / α N ), correct all wavelength points in the reference array in the same way, reorganize the reference array into a single-row spectrum, and then arrange the single-row spectrum from left to right in ascending order according to the wavelength value, and the reorganized single-row spectrum data is the corrected blood sample spectrum data; The light intensity values ​​at the M spectral reference correction points are all the same value, and when the spectral data is reorganized, only a single spectral reference correction point needs to be retained.

Images