Spinal cord injury degree detection method and system based on terahertz spectrum

Abstract

The invention discloses a method and a system for detecting spinal cord injury degree based on terahertz spectrum, which specifically carry out analysis based on terahertz frequency domain spectrum, reasonably design a data analysis method to strengthen spectral characteristics, calculate refractive index, and utilize the relationship between the refractive index and absorption coefficient and blood fat content to realize effective detection of blood fat concentration of a sample. Finally, the targets of effective detection and long-term monitoring of spinal cord injury are achieved. Compared with traditional detection technologies such as biochemical detection, imaging examination and electrophysiological examination, the method has the advantages that the detection time is short, the cost is low, harmful ionization reaction cannot be caused to biological tissues, and the safety and the detection efficiency are higher.

Description

Technical Field

[0001] The present invention belongs to the technical field of terahertz spectrum detection, and in particular relates to a method and system for detecting the degree of spinal cord injury based on terahertz spectrum. Background Art

[0002] Spinal cord injury (SCI) is a neurological disease with clinical manifestations of persistent neurological deficits and severe impairment of motor function, which has attracted global health attention. Existing methods for detecting and evaluating spinal cord injury can be divided into three categories.

[0003] The first category is neurological examination, which involves evaluating the patient's motor and sensory functions through clinical observation. This method is relatively simple and non-invasive, but due to its subjective nature, its objectivity and specificity are low.

[0004] The second category includes imaging examinations, such as X-rays, computed tomography (CT), magnetic resonance imaging (MRI), etc. Although these methods can provide more specific information about spinal cord injury, they lack analysis of the molecular level and physiological indicators in patients and need to be combined with other biochemical screening methods.

[0005] The third category is electrophysiological examination, which evaluates the integrity of the nervous system by recording the electrophysiological characteristics of the body of patients with spinal cord injury. For different patients with spinal cord injury, electrophysiological measurements lack consistent measurement standards, resulting in diagnostic differences.

[0006] Generally speaking, traditional spinal cord injury detection technology has many shortcomings, such as low detection efficiency, long time consumption, high medical costs, and potential radiation damage to the human body.

[0007] Terahertz spectroscopy has now developed into a mature and reliable technology. Since terahertz waves (electromagnetic waves with a frequency of 0.1THz-10THz) cover the rotational and vibrational energy levels of many biological molecules and are particularly sensitive to subtle changes in molecular types and structures, it is possible to study the structural information and constituent elements of substances by analyzing terahertz spectral information, refractive index, absorptivity and other parameters. In addition, terahertz waves are transient, and the energy of terahertz photons is only on the order of millielectronvolts, which will not cause harmful ionization reactions to biological tissues. Therefore, terahertz spectroscopy detection is also fast and non-destructive.

[0008] Glycerol phospholipids (GP) and triglycerides (TG) have been proven to be effective biomarkers for monitoring spinal cord injury. Medical research shows that as the degree of spinal cord injury increases, the permeability of the blood-spinal cord barrier (BSCB) increases, and lipid molecules such as GP and TG can penetrate the blood-spinal cord barrier and increase their concentration in the blood. Clinical medicine has also proven that the concentration of GP and TG molecules is positively correlated with the degree and duration of spinal cord injury.

[0009] Therefore, there is an urgent need for a new method to detect changes in biomarker concentrations using terahertz spectroscopy for continuous monitoring of spinal cord injury pathology. Summary of the invention

[0010] In view of the above, the purpose of the present invention is to provide a method for detecting the degree of spinal cord injury based on terahertz spectroscopy. By detecting the content of lipid molecules TG and GP, it is possible to achieve rapid, non-destructive, and lower-cost effective detection of the degree of spinal cord injury, thereby achieving effective detection and long-term monitoring of spinal cord injury.

[0011] To achieve the above-mentioned purpose of the invention, an embodiment provides a spinal cord injury degree detection device based on terahertz spectroscopy, comprising the following steps:

[0012] Preparing a serum sample, and scanning the serum sample using an attenuated total reflection terahertz frequency domain spectrometer to obtain a terahertz frequency domain spectrum of the serum sample, and calculating the refractive index spectrum of the serum sample after preprocessing the terahertz frequency domain spectrum, and constructing a refractive index spectrum data set;

[0013] The blood lipid parameter concentration of the in vitro serum sample is measured by a mass spectrometer, and the blood lipid parameter concentration is matched with the refractive index spectrum in the refractive index data set, so as to construct a database that ensures the mapping relationship between the refractive index spectrum and the blood lipid parameter concentration;

[0014] Performing principal component analysis on the refractive index spectrum in the refractive index data set to obtain a principal component analysis result, and performing data fitting based on a first principal component included in the principal component analysis result and a blood lipid parameter concentration corresponding to the first principal component determined according to a database to obtain a detection model between the blood lipid parameter concentration and the first principal component;

[0015] The first principal component of the refractive index of the terahertz spectrum corresponding to the blood sample to be tested with unknown spinal cord injury degree is calculated, and the blood lipid parameter concentration is predicted based on the first principal component based on the constructed detection model. The spinal cord injury degree is evaluated based on the blood lipid parameter concentration, thereby realizing the spinal cord injury degree detection based on terahertz spectroscopy.

[0016] Preferably, the refractive index spectrum of the serum sample is calculated after preprocessing the terahertz frequency domain spectrum, including:

[0017] The terahertz frequency domain spectrum obtained when no serum sample to be tested is placed is used as a reference spectrum, and the terahertz frequency domain spectrum when the serum sample to be tested is placed is used as a sample spectrum, and the amplitude attenuation and phase shift caused by the serum sample to be tested are obtained by using amplitude division and phase subtraction;

[0018] The refractive index spectrum of the serum sample to be tested is calculated based on the amplitude attenuation and phase shift using the Fresnel attenuation reflection law.

[0019] Preferably, principal component analysis is performed on the refractive index spectrum in the refractive index data set, comprising:

[0020] The refractive index spectrum data is subjected to zero-mean matrix construction, covariance matrix calculation, eigenvalue and eigenvector calculation, and then the eigenvectors are arranged from large to small according to the eigenvalues, and the proportion of the cumulative variance explained by the eigenvectors is calculated. Then, each principal component is selected according to the proportion to obtain the principal component analysis results.

[0021] Preferably, the blood lipid parameters include glycerophospholipids GP and triglycerides TG;

[0022] For glycerophospholipid GP, a linear equation was used for fitting the data between glycerophospholipid GP and the first principal component.

[0023] Preferably, for triglyceride TG, when performing data fitting between triglyceride TG and the first principal component, an exponential equation is used for fitting.

[0024] Preferably, the least square method is used to solve the coefficients of the corresponding equations during equation fitting, and a detection model consisting of the determination coefficients and the equations is obtained.

[0025] Preferably, the method further comprises: based on the principal component analysis result, using a confidence ellipse to evaluate the principal component analysis result to distinguish the damage degree effect corresponding to the serum sample.

[0026] To achieve the above-mentioned purpose of the invention, the embodiment of the present invention further provides a terahertz spectroscopy spinal cord injury degree detection system, comprising an attenuated total reflection terahertz frequency domain spectrometer, a mass spectrometer, and a computing device;

[0027] The attenuated total reflection terahertz frequency domain spectrometer is used to scan the serum sample to obtain the terahertz frequency domain spectrum of the serum sample;

[0028] The mass spectrometer is used to measure the concentration of blood lipid parameters of the isolated serum sample;

[0029] The computing device is used to calculate the refractive index spectrum of the serum sample after preprocessing the terahertz frequency domain spectrum, and construct a refractive index spectrum data set; it is also used to correspond the blood lipid parameter concentration with the refractive index spectrum in the refractive index data set, and construct a database that ensures the mapping relationship between the refractive index spectrum and the blood lipid parameter concentration; it is also used to perform principal component analysis on the refractive index spectrum in the refractive index data set to obtain the principal component analysis result, and based on the first principal component included in the principal component analysis result and the blood lipid parameter concentration corresponding to the first principal component determined according to the database, perform data fitting to obtain a detection model between the blood lipid parameter concentration and the first principal component; it is also used to calculate the first principal component of the refractive index of the terahertz spectrum corresponding to the blood sample to be tested with unknown spinal cord injury degree, and predict the blood lipid parameter concentration based on the first principal component based on the constructed detection model, evaluate the spinal cord injury degree based on the blood lipid parameter concentration, and realize spinal cord injury degree detection based on terahertz spectroscopy.

[0030] Preferably, the computing device is also used to evaluate the principal component analysis results using a confidence ellipse based on the principal component analysis results to distinguish the damage degree effects corresponding to the serum samples.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] The present invention solves the shortcomings of the prior art in detecting spinal cord injuries in patients, such as long time consumption, high cost, and potential radiation damage to the human body. Specifically, the present invention conducts analysis based on terahertz frequency domain spectroscopy, rationally designs data analysis methods to enhance spectral characteristics, calculates the refractive index, and uses the relationship between the refractive index and the absorption coefficient and the blood lipid content to achieve effective detection of the blood lipid concentration of the sample, ultimately achieving the goal of effective detection and long-term monitoring of spinal cord injuries. Compared with traditional biochemical testing, imaging examinations, electrophysiological examinations and other detection technologies, the present invention has a short detection time, low cost, and will not cause harmful ionization reactions to biological tissues, and has higher safety and detection efficiency.

[0033] The model prediction results show that based on a large-scale training data set, the present invention maximizes the identification effect of the first principal component on the TG / GP content in serum, effectively utilizes the refractive index characteristics of serum at different lipid concentrations, and achieves a good prediction effect through the regression model. The correlation coefficients of the regression equations are all greater than 0.94, and the average detection accuracy of the concentration is 94%, indicating that terahertz spectroscopy can accurately detect changes in blood lipid concentration caused by spinal cord injury. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 This is a flow chart of a method for detecting the extent of spinal cord injury based on terahertz spectroscopy;

[0036] Figure 2 Schematic diagram of attenuated total reflection photoelectric terahertz frequency domain spectrometer system and rat spinal cord injury model;

[0037] Figure 3 ATR reflectance spectra obtained by repeated measurements of serum samples of the control group;

[0038] Figure 4 is the box plot of ATR reflectivity at 1.1 THz;

[0039] Figure 5 is the serum refractive index spectrum;

[0040] Figure 6 The histogram of GP and TG concentrations in serum samples measured by mass spectrometer;

[0041] Figure 7 It is a line graph of TG concentration in serum samples measured by mass spectrometer;

[0042] Figure 8 It is a line graph of GP concentration in serum samples measured by mass spectrometer;

[0043] Fig. 9 is the eigenvalues ​​of the first five principal components (PCs) and their cumulative contribution to the total variance of the data;

[0044] Fig.10 Score serum samples on the first 2 PCs;

[0045] Fig.11 The result image is the regression equation fitting between GP / TG concentration and PC1 score;

[0046] Fig.12 It is the scatter distribution between the actual value and the calculated value of GP / TG concentration, and the predicted residual of GP / TG concentration. DETAILED DESCRIPTION

[0047] To make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.

[0048] The technical concept of the present invention is: in response to the problems in existing detection technologies, the present invention is based on terahertz spectroscopy, utilizes the positive correlation between the concentrations of GP and TG molecules and the degree of spinal cord injury, and combines the principal component analysis algorithm and the least squares algorithm to establish a detection model for the GP and TG concentrations of serum samples, so as to achieve effective detection and long-term monitoring of spinal cord injury.

[0049] like Figure 1 and Figure 2 As shown, the embodiment provides a method for detecting the degree of spinal cord injury based on terahertz spectroscopy, comprising the following steps:

[0050] S1, prepare a serum sample, and scan the serum sample using an attenuated total reflection terahertz frequency domain spectrometer to obtain a terahertz spectrum of the serum sample, and calculate the refractive index spectrum of the serum sample after preprocessing the terahertz spectrum, and construct a refractive index spectrum data set.

[0051] In the embodiment, an experimental model with different degrees of spinal cord injury is constructed, and in vitro serum samples at different spinal cord injury moments are extracted. A part of the serum samples collected in each group is used for model training, and a part is used as a model test. In order to eliminate the interference and influence of other factors, it is necessary to control the influencing factors such as the environment and experimental rats. In the present embodiment, all experimental rats were purchased from Hangzhou Medical College, Zhejiang, China. All animal procedures used in this study were carried out in accordance with the provisions of the "Chinese Animal Welfare Law". To control variables, adult female Sprague-Dawley rats weighing 200 to 220 g were uniformly raised in a regulated environment, maintaining a temperature of about 24 ° C and a humidity of about 40%. The living conditions of the rats follow a 12-hour light and dark cycle, and food, water and bedding are regularly replaced.

[0052] In the embodiment, the T10 segment was selected as the position of spinal cord injury in rats for evaluation, and the control group adopted the same procedure as the experimental group, but no spinal cord injury was constructed. The construction of the experimental group of spinal cord injury in rats specifically includes the following steps:

[0053] (1) After the rats were anesthetized by intraperitoneal injection of 1% (w / v) sodium pentobarbital (5 mL / kg), the surgical incision was shaved and disinfected with iodine; (2) After the rat skin was cut, the fascia was exposed, and the muscles were gently dissected. After the muscles on both sides of the spine were cut, the paravertebral tissue was systematically separated layer by layer to expose the T10 segment, and a laminectomy was performed to reveal the dura mater; (3) In order to induce spinal cord injury, the experimental group rats were subjected to spinal cord injury induction by constant pressure vascular clamp compression. The duration of spinal cord compression was changed to change the degree of spinal cord injury. Specifically, the vascular clamp was used for compression for 30 or 120 seconds (15 g force, Oscar, China). (4) After the operation, the muscles and skin were sutured layer by layer and disinfected with povidone iodine. The body temperature was maintained after the operation, and the bladder was artificially drained twice a day.

[0054] In the step, the model of different spinal cord injury degrees is realized by extracting serum samples at different spinal cord injury times. Therefore, in order to study the pathological characteristics of different periods, after the spinal cord injury induction operation, the experimental group rats were divided into five groups according to the serum sample extraction time, with 3 rats in each group, and the spinal cord injury time was: 6 hours, 1 day, 3 days, 7 days, and 14 days. The control group did not undergo spinal cord injury induction operation, with a total of 3 rats.

[0055] The experimental models of different degrees of spinal cord injury are divided by the time of spinal cord injury, which is achieved by extracting serum samples at different times of spinal cord injury. After the experimental rats were induced with spinal cord injury, the spinal cord injury can be divided into acute phase (<12 hours), subacute phase (12 hours to 2 days), intermediate phase (2 days to 7 days), and chronic phase (>7 days) according to the injury time. In order to study the pathological characteristics of different periods, after the spinal cord injury induction operation, the experimental group rats were divided into five groups according to the time of serum sample extraction, with 3 rats in each group: serum samples of rats with spinal cord compression (experimental group) and rats without spinal cord injury (control group) were extracted at 6 hours, 1 day, 3 days, 7 days, and 14 days, respectively.

[0056] Specifically, the process of extracting serum samples from rats with spinal cord compression (experimental group) and rats without spinal cord injury (control group) is as follows: (1) the anesthetized rats are placed in a supine position, and the area around the spine is depilated and disinfected; (2) a blood collection needle is inserted obliquely into the chest cavity under the rat spine, and after the needle tip passes through the diaphragm and enters the ventricle, blood (4-5 ml) is collected using a blood collection tube (367955, BD Vacutainer); (3) the collected plasma is centrifuged (4000 g, 15 min) to obtain a serum sample (1-2 ml).

[0057] In the embodiment, after obtaining the serum sample, the serum sample is scanned using an attenuated total reflection terahertz frequency domain spectrometer to obtain the terahertz frequency domain spectrum of the serum sample. The terahertz frequency domain spectrum includes amplitude information and phase information. Among them, the attenuated total reflection photoelectric terahertz frequency domain spectrometer uses the evanescent wave generated by the attenuated total reflection of the terahertz wave on the surface of the Dove prism of the high-resistance silicon material to perform spectral detection on the serum sample. The evanescent wave interacts with the sample placed on the surface of the attenuated total reflection prism to form a modulated terahertz signal carrying the sample information and then outputs it. The terahertz photoconductive detection antenna then performs coherent mixing on the modulated terahertz signal and the laser signal to generate a down-converted photocurrent signal, and after being amplified by a transimpedance amplifier, it is sampled by the digital signal processing unit, and the sampled oscillating photocurrent signal is Hilbert transformed to obtain a terahertz frequency domain spectrum containing amplitude and phase information.

[0058] Since the working range of the photosensor in the spectrometer and the parasitic interference effect are common at low frequencies, there is no reliable terahertz data below 50 to 100 GHz, so the 300-1200 GHz frequency band of the original spectrum is selected as the frequency range for spectrum analysis, and low smoothing technology is used to reduce the noise interference of the spectral data.

[0059] Since the spectrometer system is easily disturbed by the external environment during operation, it is necessary to perform multiple measurements on the same serum sample of each damage degree to ensure measurement accuracy and eliminate abnormal data. Each time the spectrum of the serum sample is measured, the amplitude and phase spectrum of the terahertz wave with and without the blood sample to be tested need to be tested. In this embodiment, each serum sample is evaluated five times, and the outliers are removed by the box plot method. Figure 3 , Figure 4 The attenuated total reflectance spectrum measurement results of the healthy mouse serum in the control group show that the spectrum of the fourth measurement is much larger than the other four measurements in the high-frequency region. The box plot of the data at 1.1 THz also shows that the fourth measurement value is far more than 1.5 times the quartile, indicating that it is an outlier for this type of data. Finally, the measured data was averaged to obtain a more accurate measurement result.

[0060] In the embodiment, the terahertz frequency domain spectrum is preprocessed to calculate the refractive index spectrum of the serum sample, and a refractive index spectrum data set is constructed. The specific preprocessing process is:

[0061] (1) using the terahertz frequency domain spectrum obtained when no serum sample to be tested is placed as a reference spectrum, and using the terahertz frequency domain spectrum when the serum sample to be tested is placed as a sample spectrum, and using amplitude division and phase subtraction to obtain the amplitude attenuation and phase shift caused by the serum sample to be tested;

[0062] The calculation of amplitude attenuation is based on the following formula:

[0063]

[0064] Where R is the ratio of the total coefficient of the reference spectrum to the sample spectrum after attenuation, r sam (ω) is the total attenuation coefficient of the sample spectrum, r ref (ω) is the total attenuation coefficient of the reference spectrum, is the spectral amplitude measured when the serum sample to be tested is placed, is the spectral amplitude measured without placing the serum sample to be tested, is the photocurrent without absorption loss under ideal conditions, and ω is the frequency of the terahertz wave.

[0065] (2) calculating the refractive index spectrum of the serum sample to be tested based on the amplitude attenuation and phase shift by using the Fresnel attenuation reflection law;

[0066] According to the Fresnel equation for p-polarized radiation, r sam (ω) and r ref (ω) respectively satisfy the following equations:

[0067]

[0068] Among them, n prism =3.42 is the refractive index of the Dove prism of the spectrometer, n air =1 is the refractive index of air, θ=55.7° is the incident angle of the terahertz wave, is the complex refractive index of serum sample n.

[0069] The complex refractive index of serum sample n can be calculated by formula (2) and (3): Satisfy the following formula:

[0070]

[0071] Finally, the refractive index n of the serum sample is obtained sample The expression is:

[0072]

[0073] Among them, Rr ref R×r ref , Re[·] represents the operation of taking the real part of the complex number in the brackets.

[0074] After calculating the terahertz refractive index spectra of all serum samples, a refractive index spectrum data set can be constructed. The terahertz refractive index spectrum obtained after data preprocessing in this embodiment is as follows: Figure 5As shown, Ctrl, 6H, 1D, 3D, 7D, and 14D represent the sham control group and serum samples collected at 6 hours, 1 day, 3 days, 7 days, and 14 days after spinal cord injury, respectively. The shaded area represents the standard error of each frequency point. As the frequency increases, the refractive index of the serum gradually decreases. There is no obvious difference between the low-frequency groups, but at 0.9-1.2 THz, the refractive index of the serum samples in the spinal cord injury group is significantly lower than that in the control group, and decreases with the increase of injury time. For serum samples collected at the same time point, the longer the spinal cord is compressed, the smaller the refractive index. In addition, the effect of spinal cord compression time on the refractive index of blood samples is smaller than that of injury time. The above analysis shows that terahertz spectroscopy can sensitively reflect the changes in the degree and time of spinal cord injury in rats.

[0075] S2, using a mass spectrometer to measure the blood lipid parameter concentration of the in vitro serum sample, and matching the blood lipid parameter concentration with the refractive index spectrum in the refractive index data set, and constructing a database that ensures the mapping relationship between the refractive index spectrum and the blood lipid parameter concentration.

[0076] In the embodiment, the actual blood lipid parameter concentration of the isolated serum sample is also measured by mass spectrometry, and the blood lipid parameters include GP and TG. The specific measured GP concentration and TG concentration are as follows: Figure 6 As shown, Figure 6 It shows that with the increase of injury time and injury degree, GP concentration and TG concentration both gradually increase, and are negatively correlated with the terahertz refractive index spectrum. The embodiment also corresponds the refractive index of the data set to the blood lipid parameters (TG, GP) one by one, that is, the experimental refractive index measurement value and the actual blood lipid parameter value measured by medicine of the same serum sample are specifically sorted into a group to construct a database, that is, each serum sample in the database corresponds to two types of data, the refractive index measurement value and the actual TG and GP concentrations.

[0077] In order to observe the changes in lipid concentration more clearly, line graphs showing the changes in GP concentration and TG concentration over time were also drawn, as shown in Figure 7 , Figure 8 shown. Figure 7 The results showed that in the acute phase of injury (within 24 hours), the TG concentration in serum increased sharply, while in the subacute phase (1 to 7 days), the TG concentration did not increase significantly. On the 14th day, the TG concentration increased significantly to three times that of the 7th day. Figure 8 The results showed that the GP concentration also increased significantly in the acute phase, but then showed a slow upward trend. As the injury time increased, the concentration difference between the 30s compression group and the 120s compression group gradually decreased, indicating that the GP concentration difference caused by the degree of injury would gradually decrease over time. The relationship between blood lipid concentration and terahertz spectral characteristics was confirmed.

[0078] S3, performing principal component analysis on the refractive index spectra in the refractive index data set to obtain principal component analysis results, and performing data fitting based on the first principal component included in the principal component analysis results and the blood lipid parameter concentration corresponding to the first principal component determined according to the database to obtain a detection model between the blood lipid parameter concentration and the first principal component.

[0079] In the embodiment, principal component analysis is performed on the sample refractive index spectrum, and key features in the spectrum data, namely, each principal component, are extracted by linear dimensionality reduction to obtain the principal component analysis result. The specific process is:

[0080] (1) Zero mean matrix construction;

[0081] Calculate the mean vector μ for each observation (i.e., refractive index spectrum data at different frequencies) i , i is the observation frequency index, j is the spectral refractive index index, x ji is the refractive index value of the terahertz frequency domain spectrum at the i-th observation frequency, and the corresponding mean vector is subtracted from each column of the data set to obtain a zero-mean matrix

[0082] (2) Calculation of covariance matrix;

[0083] According to the zero-mean matrix X, the covariance matrix Σ of the refractive index spectrum data is calculated and expressed as:

[0084]

[0085] (3) Perform eigenvalue decomposition on the covariance matrix. After obtaining the eigenvalues ​​and eigenvectors, arrange the eigenvectors from large to small according to the eigenvalues, calculate the proportion of the cumulative variance explained by the eigenvectors, and then select each principal component based on the proportion to obtain the principal component analysis results.

[0086] Specifically, the principal component analysis algorithm was implemented using MATLAB software, and the number of samples n was selected as 30 in the experiment to achieve effective dimensionality reduction. Fig. 9 The eigenvalues ​​of the first five principal components (PCs) and their cumulative contribution to the total variance of the data are shown. As the number of PCs increases, the eigenvalues ​​decrease rapidly; the first and second PCs account for 97.9% and 1.4% of the total variance of the serum spectral data, respectively, while the other PCs account for less than 1% of the total variance of the data. This indicates that the first two PCs already contain most of the spectral data features.

[0087] In the embodiment, based on the principal component analysis results, the confidence ellipse is used to evaluate the effect of the PCA results in distinguishing the degree of damage corresponding to the serum samples. The confidence ellipse is determined using a set of given confidence levels, specifically, by projecting the error range of the data points onto a two-dimensional plane to obtain an elliptical distribution. The calculation of the confidence ellipse requires the use of an error function and a normal distribution function. Drawing the confidence ellipse specifically includes the following steps:

[0088] (1) Determine the error range; in the present invention, the confidence interval is determined to be 95%. (2) Calculate the ellipse parameters; in the principal component space, the ellipse equation corresponding to the centered data is Where λ1 and λ2 are the sizes of the first eigenvalue and the second eigenvalue, respectively, and k is the confidence level. When the confidence level is 95%, the k value calculated according to the chi-square distribution is 5.991. (3) The final confidence ellipse is formed by superposition; in order to restore the true confidence ellipse, the ellipse needs to be translated and rotated, and the confidence ellipse obtained from the spectral data of serum samples at different injury times is drawn in the same coordinate system to obtain the final confidence ellipse.

[0089] Specifically, the samples were divided into 6 groups according to the time of spinal cord injury, and a confidence ellipse was drawn for each group of data according to the 95% confidence interval. Fig.10 The serum sample scores on the first two PCs are shown. The scatter plot shows that the six groups of samples have good clustering characteristics, especially the first PC score of each group of data, which shows an increasing trend with the increase of injury time. However, there is no significant difference in the second PC score among the groups.

[0090] In the embodiment, based on the first principal component PC1 included in the principal component analysis result and the blood lipid parameter concentration corresponding to the first principal component PC1 determined according to the database, data fitting is performed to obtain a detection model between the blood lipid parameter concentration and the first principal component. Specifically, based on the scatter distribution between the GP / TG concentration and the PC1 score, the least squares method is used to establish a regression equation between the PC1 score and the GP and TG concentrations in the serum, and the specific distribution of the data points is specifically observed. It is preliminarily determined that the regression equation between the GP concentration and the PC1 score is a linear equation y=b m *x i m +...+b1x+b0, the regression equation between TG concentration and PC1 score is the exponential equation y=a1e bx +a0. Based on this assumption, the least squares algorithm is used to establish the regression equation, which includes the following steps:

[0091] (1) Use PC1 score to establish the data matrix, GP / TG concentration to establish the observation value vector, and the corresponding coefficient vector β to be determined GP and β TG; (2) The coefficient vector is obtained by minimizing the sum of squared errors using the least squares method, and then a prediction model is obtained that constrains the mapping relationship between PC1 score and GP / TG concentration through the coefficient vector; (3) The correlation coefficient of the corresponding prediction model of GP and TG is calculated: R GP =0.979, R TG =0.942, and the regression equation between GP / TG concentration and PC1 score was plotted (including 95% confidence band and 95% prediction band), as shown in the attached figure. Fig.11 As shown; (4) Calculate the calibration root mean square standardized error RMSSEC of the two prediction models respectively GP =0.70, RMSSEC TG =2.23, cross-validation root mean square standardized error RMSSECV GP =1.46,RMSSECV TG =7.61, and the root mean square standardized error of prediction RMSSEP GP =1.20, RMSSEP TG =25.17, draw the image as shown in the attached Fig.12 As shown. Among them, Fig.12 (a) shows the scatter distribution between the actual and calculated values ​​of GP concentration. Fig.12 (b) shows the predicted residuals of GP concentration. Fig.12 (c) shows the scatter distribution between the actual and calculated values ​​of TG concentration. Fig.12 (d) shows the predicted residuals of TG concentration.

[0092] Among them, the root mean square standardized error of calibration (RMSSEC) is used to evaluate the degree of fit of the model to the training data. It calculates the root mean square error between the model prediction of the training set and the actual observation; the root mean square standardized error of cross validation (RMSSECV) is used to evaluate the prediction performance of the model in the cross validation process. It calculates the root mean square error between the model prediction of the retained validation set and the actual observation; the root mean square standardized error of prediction (RMSSEP) is used to evaluate the model's predictive ability for new samples. It calculates the root mean square error between the model prediction of new samples and the actual observation.

[0093] S4, calculate the first principal component of the refractive index of the terahertz spectrum corresponding to the blood sample to be tested with unknown spinal cord injury degree, and predict the blood lipid parameter concentration based on the first principal component based on the constructed detection model, evaluate the spinal cord injury degree based on the blood lipid parameter concentration, and realize the spinal cord injury degree detection based on terahertz spectroscopy.

[0094] In the embodiment, the trained GP and TG detection models are used to test the test set as unknown serum samples to obtain the analysis results of blood lipid concentration, and the measured values ​​and actual values ​​of GP and TG are compared to verify the measurement accuracy of the model. The trained model is used to test the test set, and the average measurement accuracy of the model is 94%, indicating that terahertz spectroscopy can accurately detect changes in blood lipid concentration caused by spinal cord injury.

[0095] The embodiment also provides a spinal cord injury degree detection system based on terahertz spectroscopy, including an attenuated total reflection terahertz frequency domain spectrometer, a mass spectrometer, and a computing device. Among them, the attenuated total reflection terahertz frequency domain spectrometer is used to scan the serum sample to obtain the terahertz frequency domain spectrum of the serum sample. The mass spectrometer is used to measure the blood lipid parameter concentration of the in vitro serum sample. The computing device is mainly used for various analysis and calculations in the process of spinal cord injury degree detection, specifically including various calculations in S1-S4 in the above method, to realize spinal cord injury degree detection.

[0096] The computing device, in addition to the processor and memory, also includes hardware required for other services such as internal bus, network interface, memory, etc. at the hardware level. The memory is a non-volatile memory, and the processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the various computing processes in S1-S4 above. Of course, in addition to the software implementation, the present invention does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0097] The specific implementation methods described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting the degree of spinal cord injury based on terahertz spectroscopy, characterized in that: The following steps are involved: Preparing a serum sample, and scanning the serum sample using an attenuated total reflection terahertz frequency domain spectrometer to obtain a terahertz frequency domain spectrum of the serum sample, and calculating the refractive index spectrum of the serum sample after preprocessing the terahertz frequency domain spectrum, and constructing a refractive index spectrum data set; The blood lipid parameter concentration of the in vitro serum sample is measured by a mass spectrometer, and the blood lipid parameter concentration is matched with the refractive index spectrum in the refractive index data set, so as to construct a database that ensures the mapping relationship between the refractive index spectrum and the blood lipid parameter concentration; Performing principal component analysis on the refractive index spectrum in the refractive index data set to obtain a principal component analysis result, and performing data fitting based on a first principal component included in the principal component analysis result and a blood lipid parameter concentration corresponding to the first principal component determined according to a database to obtain a detection model between the blood lipid parameter concentration and the first principal component; The first principal component of the refractive index of the terahertz spectrum corresponding to the blood sample to be tested with unknown spinal cord injury degree is calculated, and the blood lipid parameter concentration is predicted based on the first principal component based on the constructed detection model. The spinal cord injury degree is evaluated based on the blood lipid parameter concentration, thereby realizing the spinal cord injury degree detection based on terahertz spectroscopy.

2. The method for detecting the degree of spinal cord injury based on terahertz spectroscopy according to claim 1, characterized in that: The refractive index spectrum of the serum sample is calculated after preprocessing the terahertz frequency domain spectrum, including: The terahertz frequency domain spectrum obtained when no serum sample to be tested is placed is used as a reference spectrum, and the terahertz frequency domain spectrum when the serum sample to be tested is placed is used as a sample spectrum, and the amplitude attenuation and phase shift caused by the serum sample to be tested are obtained by using amplitude division and phase subtraction; The refractive index spectrum of the serum sample to be tested is calculated based on the amplitude attenuation and phase shift using the Fresnel attenuation reflection law.

3. The method for detecting the degree of spinal cord injury based on terahertz spectroscopy according to claim 1, characterized in that: Perform principal component analysis on the refractive index spectra in the refractive index dataset, including: The refractive index spectrum data is subjected to zero-mean matrix construction, covariance matrix calculation, eigenvalue and eigenvector calculation, and then the eigenvectors are arranged from large to small according to the eigenvalues, and the proportion of the cumulative variance explained by the eigenvectors is calculated. Then, each principal component is selected according to the proportion to obtain the principal component analysis results.

4. The method for detecting the degree of spinal cord injury based on terahertz spectroscopy according to claim 1, characterized in that: The blood lipid parameters include glycerophospholipids GP and triglycerides TG; For glycerophospholipid GP, a linear equation was used for fitting the data between glycerophospholipid GP and the first principal component.

5. The method for detecting the degree of spinal cord injury based on terahertz spectroscopy according to claim 4, characterized in that: For triglyceride TG, when fitting the data of triglyceride TG and the first principal component, an exponential equation is used for fitting.

6. The method for detecting the degree of spinal cord injury based on terahertz spectroscopy according to claim 4, characterized in that: When fitting the equation, the least square method is used to solve the coefficients of the corresponding equation, and a detection model consisting of the determination coefficients and the equation is obtained.

7. The method for detecting the degree of spinal cord injury based on terahertz spectroscopy according to claim 1, characterized in that: Also includes: Based on the principal component analysis results, confidence ellipse was used to evaluate the principal component analysis results to distinguish the damage degree effects corresponding to the serum samples.

8. A terahertz spectroscopy spinal cord injury degree detection system, characterized in that: Includes attenuated total reflection terahertz frequency domain spectrometer, mass spectrometer, and computing equipment; The attenuated total reflection terahertz frequency domain spectrometer is used to scan the serum sample to obtain the terahertz frequency domain spectrum of the serum sample; The mass spectrometer is used to measure the concentration of blood lipid parameters of the isolated serum sample; The computing device is used to calculate the refractive index spectrum of the serum sample after preprocessing the terahertz frequency domain spectrum, and construct a refractive index spectrum data set; it is also used to correspond the blood lipid parameter concentration with the refractive index spectrum in the refractive index data set, and construct a database that ensures the mapping relationship between the refractive index spectrum and the blood lipid parameter concentration; it is also used to perform principal component analysis on the refractive index spectrum in the refractive index data set to obtain the principal component analysis result, and based on the first principal component included in the principal component analysis result and the blood lipid parameter concentration corresponding to the first principal component determined according to the database, perform data fitting to obtain a detection model between the blood lipid parameter concentration and the first principal component; it is also used to calculate the first principal component of the refractive index of the terahertz spectrum corresponding to the blood sample to be tested with unknown spinal cord injury degree, and predict the blood lipid parameter concentration based on the first principal component based on the constructed detection model, evaluate the spinal cord injury degree based on the blood lipid parameter concentration, and realize spinal cord injury degree detection based on terahertz spectroscopy.

9. A terahertz spectroscopy spinal cord injury degree detection system, characterized in that: The computing device is also used to evaluate the principal component analysis results using a confidence ellipse based on the principal component analysis results to distinguish the damage degree effects corresponding to the serum samples.

Images