Ag (at) Pt (at) PSB composite SERS substrate and preparation method and application thereof, CXB model and establishment method and application thereof

Abstract

The invention relates to an Ag (at) Pt (at) PSB composite SERS substrate and a preparation method and application thereof, and a CXB model and an establishment method and application thereof. The invention discloses a preparation method of an Ag-Pt-PSB composite SERS substrate. The preparation method comprises the following steps: (1) preparing a PSB Bragg reflector; and (2) immersing the PSB Bragg reflector in a silver nitrate solution for 40-60 seconds, taking out the PSB Bragg reflector, immersing the PSB Bragg reflector in a chloroplatinic acid solution for at least 3 minutes, washing the PSB Bragg reflector with water, and drying the washed PSB Bragg reflector to obtain the Ag-Pt-PSB composite SERS substrate. According to the Ag-Pt-PSB composite SERS substrate, the preparation method and the application thereof, the CXB model and the establishment method and the application thereof, the novel Ag-Pt-PSB composite SERS active substrate is successfully developed by adopting electrochemical corrosion and an in-situ reduction method, and the novel Ag-Pt-PSB composite SERS active substrate has a higher SERS enhancement effect and higher sensitivity, so that the novel Ag-Pt-PSB composite SERS active substrate can be applied to celecoxib detection; and the substrate is combined with a deep learning algorithm, so that a CXB model with higher precision can be established.

Description

Technical Field

[0001] The present invention specifically relates to an Ag@Pt@PSB composite SERS substrate and a preparation method and application thereof, a CXB model and an establishment method and application thereof. Background Art

[0002] Celecoxib (CXB) is the first cyclooxygenase-2 (COX-2) specific inhibitor approved for the treatment of osteoarthritis (OA) and rheumatoid arthritis (RA). Many reports have shown that the dose effect of CXB is relatively mild, and the risk of cardiovascular and gastrointestinal events at moderate doses of CXB is significantly lower than that of other similar anti-inflammatory analgesics. Although CXB is considered safe, in high-dose treatment, increased dose levels and duration of use may still cause serious complications such as gastrointestinal bleeding, myocardial infarction, heart failure, and renal damage. At the same time, as a widely used non-steroidal anti-inflammatory drug, CXB has a very large metabolic variability in the plasma concentration of different patients, which amplifies the individual differences in drug exposure. Especially for the middle-aged and elderly people who have been troubled by arthritis for a long time, the decline in metabolic capacity and often accompanied by other chronic diseases such as hypertension and diabetes require more caution when determining the dosing regimen.

[0003] Personalized medicine is the trend of medical development. Clinicians should weigh the patient's own conditions, including age, gender, comorbidities and concomitant drug interactions before using medication. Therapeutic drug monitoring (TDM) relies on measuring drug concentrations in body fluids to adjust the clinical use of drugs, thereby improving the effectiveness and safety of drug treatment. Determining the appropriate celecoxib dose for each individual through TDM can reduce the possibility of adverse events in patients during treatment and maximize the effect of drug treatment. Therefore, the exploration of drug concentration detection methods in patient serum is of great significance to the realization of personalized medicine.

[0004] There are many methods for determining the content of celecoxib, such as high performance liquid chromatography (HPLC), ultraviolet spectroscopy (UV), electrochemical method, etc. Although these methods have advantages in reliability, they usually have problems such as bulky equipment, complicated procedures, long turnaround time, and cumbersome pretreatment. With the increasing demand for highly sensitive and portable detection technology, surface enhanced Raman scattering (SERS) has become an attractive candidate. SERS technology has become an important research topic in the field of drug analysis due to its advantages such as high sensitivity, real-time response, fingerprint recognition characteristics and non-destructive collection. The development of SERS active substrates with stable performance and high sensitivity is the key to the rapid implementation of SERS technology in the biomedical field. According to research, there is currently no report on SERS detection technology for celecoxib.

[0005] In view of this, the present invention proposes a new Ag@Pt@PSB composite SERS substrate and preparation method, application, CXB model and establishment method, application, and synthesizes an Ag@Pt@porous silicon Bragg reflector composite substrate, which is combined with deep learning to assist the SERS detection platform and can be used for label-free detection of celecoxib in serum. Summary of the invention

[0006] The first invention purpose of the present invention is to provide a method for preparing a Ag@Pt@PSB composite SERS substrate, and a new Ag@Pt@PSB composite SERS active substrate is successfully developed by electrochemical corrosion and in-situ reduction. The introduction of platinum elements effectively improves the hot spot distribution of pure silver nanoparticles, so that the substrate exhibits better SERS performance. At the same time, the chemical properties of platinum are more stable, and the distribution of platinum on silver nanoparticles slows down the oxidation rate of silver and improves the long-term stability of the substrate.

[0007] In order to achieve the above purpose, the technical solution adopted is:

[0008] A method for preparing an Ag@Pt@PSB composite SERS substrate comprises the following steps:

[0009] (1) Preparation of PSB Bragg reflectors;

[0010] (2) After immersing the PSB Bragg reflector in a silver nitrate solution for 40-60 seconds, taking it out, and then immersing it in a chloroplatinic acid solution for at least 3 minutes, washing it with water, and drying it to obtain the Ag@Pt@PSB composite SERS substrate.

[0011] Furthermore, in the step (1), the preparation method of the PSB Bragg reflector is: washing the silicon wafer with ethanol, water, and ethanol in sequence, drying, electrochemically etching, and then washing with water and drying;

[0012] In the electrochemical corrosion process, the corrosion liquid contains hydrofluoric acid and ethanol in a volume ratio of 25:25-30;

[0013] In the electrochemical corrosion process, 12 cycles are alternately cycled. One cycle process is: first, 115 mA / cm 2 The high current density was used for corrosion for 1s, and then 65mA / cm 2 The low current density corrosion is carried out for 1.2s, and after an interval of 0.8s, the next cycle of electrochemical corrosion is carried out;

[0014] In the step (2), the concentration of the silver nitrate solution is 0.04-0.06 mol / L, and the concentration of the chloroplatinic acid solution is 0.004-0.006 mol / L.

[0015] Furthermore, in the step (1), during the electrochemical corrosion process, the corrosion solution contains hydrofluoric acid and ethanol in a volume ratio of 25:28;

[0016] In the step (2), the PSB Bragg reflector is immersed in a 0.05 mol / L nitrate solution for 50 seconds, taken out, and then immersed in a 0.005 mol / L chloroplatinic acid solution.

[0017] The second inventive object of the present invention is to provide an Ag@Pt@PSB composite SERS substrate, which is prepared by the above-mentioned preparation method, and the substrate has the advantages of excellent SERS performance, SERS response ability, good reproducibility and high sensitivity.

[0018] The third invention object of the present invention is to provide the use of the above-mentioned Ag@Pt@PSB composite SERS substrate in label-free detection of celecoxib in serum. Based on the SERS response ability of the substrate to CXB, the substrate can be used to detect the concentration of CXB in serum. This is the first time that the drug celecoxib has been detected by SERS.

[0019] The fourth inventive object of the present invention is to provide a method for establishing a CXB model. Based on the above-mentioned SERS substrate and deep learning algorithm, a CXB model with high accuracy can be established, which is helpful to develop a new fast, sensitive, non-invasive and low-cost detection technology based on SERS substrate + deep learning, accurately detect the drug dispersion in patient serum, and provide technical support for the accurate identification of blood drug concentration classification.

[0020] In order to achieve the above purpose, the technical solution adopted is:

[0021] A method for establishing a CXB model comprises the following steps:

[0022] (1) selecting healthy subjects, collecting fresh blood and performing centrifugation to obtain serum; and then using the serum to prepare celecoxib-serum samples of various concentrations;

[0023] (2) collecting data: dropping the celecoxib-serum sample onto the Ag@Pt@PSB composite SERS substrate of claim 4, and obtaining a SERS spectrum after the sample is in full contact with the substrate;

[0024] (3) Data preprocessing: After the SERS spectrum is preprocessed by baseline correction, a deep learning algorithm model, namely the CXB model, is established.

[0025] Furthermore, in step (1), at least 20 samples of healthy subjects are selected;

[0026] Celecoxib-serum samples were at concentrations of 200, 500, 700, 1000, and 5000 ng / mL;

[0027] In the step (3), a GoogleNet, ResNet, or VGG deep learning algorithm model is established.

[0028] Furthermore, in the step (3), a GoogleNet deep learning algorithm model is established.

[0029] The fifth inventive object of the present invention is to provide a CXB model, which is obtained by the above-mentioned establishment method and can be used to monitor the concentration of CXB in serum. It has great detection potential for the concentration of CXB in serum. This technology is expected to develop a portable, economical and highly accurate CXB concentration detection device.

[0030] A sixth object of the present invention is to provide application of the above-mentioned CXB model in a CXB concentration detection device.

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

[0032] 1. In the technical solution of the present invention, an Ag@Pt@PSB composite SERS active substrate was successfully developed by electrochemical corrosion and in-situ reduction method. The combination of Ag@PtNPs not only retains the high plasma effect of silver, but also synergizes the antioxidant ability of platinum, thereby improving the optical stability of AgNPs while maintaining high SERS activity.

[0033] 2. In the technical solution of the present invention, the SERS technology is applied for the first time to the qualitative and quantitative detection of celecoxib in methanol and serum. Figure 1 Ag@Pt alloy nanoparticles are concentrated on the PSB surface, forming a large number of electromagnetic "hot spots" with high plasma enhancement effect. Crystal violet (CV) was used to characterize the substrate performance, and the substrate showed good SERS signal enhancement ability (enhancement factor EF = 5.64×10 5 ), excellent signal reproducibility (relative standard deviation RSD = 6.7%) and sensitive detection performance (linear correlation coefficient R 2 =0.9888). Meanwhile, Ag@PtNPs are beneficial to the long-term stability of the substrate. Ag@Pt@PSB composite substrate was used to detect CXB in a low concentration range, and the application potential of the substrate in serum samples containing drugs was further investigated. The results showed that the LOD of CXB in serum was 20.22 ng / mL, and the average recovery rate was 97.49%-107.65%, which deepened the advantage of SERS technology in TDM.

[0034] 3. In the technical solution of the present invention, based on the composite SERS active substrate of the present invention, the collected CXB blood drug signals are modeled using a deep learning algorithm, and the SERS spectral data sets of different concentrations are classified and identified, which can improve the accuracy of blood drug concentration evaluation. The present invention combines the Ag@Pt@PSB substrate with deep learning, and is expected to develop a fast, simple and low-cost non-invasive blood drug concentration detection device for arthritis patients, thereby promoting the further development of TDM and personalized medicine.

[0035] In an embodiment of the present invention, based on the collected CXB blood drug signal, three kinds of deep learning algorithms are used to build a model and classify and identify SERS spectral data sets of different concentrations. The results show that the GoogleNet algorithm has the best classification effect on the spectral data of five different concentrations of CXB in serum, and the AUC value of the ROC curve reaches 87%. GoogleNet is a deep neural network structure, which has a small model size and parameter amount while having high performance, and can efficiently utilize computing resources. It is a very excellent and practical model, which is widely used in classification applications and has obtained outstanding performance. The analysis of the CXB spectrum makes up for the gap of SERS technology in the field of celecoxib detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The schematic diagram of the substrate preparation process of the present invention and the celecoxib detection based on deep learning.

[0037] Figure 2 Morphological characterization of Ag@Pt@PSB composite SERS substrate; (a) substrate formation process, (b)-(c) surface morphology, cross-sectional structure and reflectance spectrum of porous silicon Bragg reflector, (d) surface morphology of prepared composite SERS substrate, (e) particle size distribution of NPs on the substrate, (f)-(g) distribution of important elements Ag and Pt, (h) EDS energy spectrum scanning electron microscopy detection and component content distribution table, (i) Si 2p, (j) Ag 3d, (k) Pt 4f, (l) O 1s high-resolution XPS spectra.

[0038] Figure 3 The enhancement factor, reproducibility, sensitivity, and long-term stability of the Ag@Pt@PSB composite SERS substrate are shown in the figure below. (a) Comparison of the SERS spectrum of CV and the effect of ordinary Raman spectrum, (b) CV (10 -5 (M) SERS spectra of 20 random points, (c) SERS spectra of 20 points at the characteristic peak 1617 cm -1 The peak intensity distribution at 1617 cm -1The linear relationship between the relative SERS intensity at the characteristic peak and the concentration (the error bar indicates the experimental error of five experiments), (f) comparison of the enhancement effect of different substrates, (g)-(h) SERS spectra and characteristic peak 1617cm of CV on Ag@Pt@PSB and Ag@PSB substrates under different storage cycles -1 Relative intensity comparison chart at .

[0039] Figure 4 Figure 2 shows the test results of Ag@Pt@PSB composite SERS substrate for SERS detection of celecoxib in methanol; (a) SERS spectrum of CXB methanol solution (250 ng / mL), (bd) 3D mapping, 2D mapping and its concentration at 1042 cm -1 and 1030cm -1 Raman peak intensity diagram at the characteristic peak, (e) SERS spectra of CXB at different concentrations in methanol solution, (f) 1042 cm -1 and 1230cm -1 Linear relationship between the relative SERS intensity at the characteristic peak and the concentration (the error bar indicates the experimental error of five experiments).

[0040] Figure 5 Figure 2 shows the test results of Ag@Pt@PSB composite SERS substrate for SERS detection of celecoxib in serum; (a) spectrum of CXB in purified serum (blue line), spectrum of CXB in methanol (green line), spectrum of purified serum (orange line), (b) spectrum of CXB at different concentrations in purified serum, (c) 855cm -1 Concentration gradient fitting diagram at (error bars represent experimental errors of 5 experiments).

[0041] Figure 6 Figure 2 shows the test results of the five-classification model of drug concentration based on deep learning-SERS for SERS detection of celecoxib in serum; (a) Deep learning training based on SERS spectra of CXB in serum, (b) CXB samples with different concentrations in serum at 855 cm -1 、1042cm -1 、1608cm -1 Representative Raman signals at three characteristic peaks, (c) confusion matrix and ROC curve obtained by GoogleNet model classification, (de) confusion matrix and ROC curve of ResNet model and VGG model. DETAILED DESCRIPTION

[0042] In order to further explain the Ag@Pt@PSB composite SERS substrate and preparation method, application, CXB model and establishment method, application of the present invention, and achieve the intended purpose of the invention, the following is a detailed description of the specific implementation methods, structures, features and effects of the Ag@Pt@PSB composite SERS substrate and preparation method, application, CXB model and establishment method, application proposed in the present invention in combination with the preferred embodiments. In the following description, different "one embodiment" or "embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0043] The following will further introduce the Ag@Pt@PSB composite SERS substrate and its preparation method, application, CXB model and its establishment method, application of the present invention in detail in combination with specific embodiments:

[0044] Celecoxib (CXB) is a nonsteroidal anti-inflammatory drug with significant efficacy in the prevention and treatment of arthritis, but overdose or improper use can still cause adverse reactions. Surface enhanced Raman scattering (SERS) is a powerful optical technology. Due to its rapid and non-destructive nature, excellent fingerprint recognition characteristics and trace sensitivity, it is becoming an important tool for therapeutic drug monitoring (TDM). The present invention innovatively proposes a detection scheme based on SERS-deep learning, which realizes the qualitative identification and quantitative detection of CXB at the Raman spectroscopy level for the first time. The present invention successfully synthesized a silver (Ag)@platinum (Pt)@porous silicon Bragg reflector (PSB) composite SERS substrate by electrochemical corrosion (ec) and in situ reduction. The substrate performance was characterized using crystal violet (CV) as a probe, and the substrate showed excellent SERS performance (EF for crystal violet = 5.64×10 5 , RSD=6.7%, R 2 =0.9888). The SERS response of the substrate to CXB was explored. The results showed that the substrate had a significant enhancement effect on CXB (EF = 2.5 × 10 5 ), good reproducibility (RSD < 6%) and high sensitivity (linear detection range of 250-5 ng / mL, R 2=0.984). The substrate was applied to the concentration detection of CXB in serum, and the LOD (20.22ng / mL) and recovery rate (97.49%-107.65%) were both within a reasonable range. Five blood drug concentrations were classified using three deep learning models, GoogleNet, ResNet, and VGG. Among them, the GoogleNet model based on the SERS spectral data set has the best classification effect, with a classification accuracy of 84.17% and an AUC value of the ROC curve of 0.87. The present invention is based on a fast, non-invasive, and simple deep learning-SERS integrated system, which reduces the difficulty of identifying blood drug concentrations by relying solely on SERS spectroscopy. This progress is expected to provide medical practitioners with valuable resources for early diagnosis and formulation of personalized treatment strategies, thereby reducing the incidence of adverse events in patients with arthritis during treatment and maximizing the efficacy of drugs. The technical solution of the present invention is:

[0045] A method for preparing an Ag@Pt@PSB composite SERS substrate comprises the following steps:

[0046] (1) Preparation of PSB Bragg reflectors;

[0047] (2) After immersing the PSB Bragg reflector in a silver nitrate solution for 40-60 seconds, taking it out, and then immersing it in a chloroplatinic acid solution for at least 3 minutes, washing it with water, and drying it to obtain the Ag@Pt@PSB composite SERS substrate.

[0048] Precious metal nanomaterials such as Au, Ag, and Cu are often considered to be ideal choices for constructing SERS substrates. Generally, Ag nanomaterials are widely used in the construction of SERS substrates because of their excellent localized surface plasmon resonance (LSPR) effect and good biocompatibility. However, single-element silver nanomaterials face many challenges: for example, traditional silver nanosols are prone to agglomeration when left static for a long time, making it difficult to provide repeated and reliable signals. Secondly, unmodified AgNPs are easily oxidized and corroded, affecting the stability and repeatability of the substrate. Compared with single-element nanomaterials, composite bimetallic or multimetallic nanocomposites have received increasing attention for their unique properties and multifunctionality. In recent years, more and more reports have expanded SERS to transition metal surfaces such as Fe, Co, Ni, Pd, and Pt. In the above technical scheme, pure platinum (Pt) has good high-temperature oxidation resistance and chemical stability. The combination of Ag@PtNPs not only retains the high plasma effect of silver, but also synergizes the antioxidant ability of platinum, improving the optical stability of AgNPs while maintaining high SERS activity. Porous silicon Bragg reflectors (PSBs) have unique optical properties and the ability to facilitate the dispersion of nanoparticles. They can not only be used as a multifunctional scaffold for loading nanoparticles, but also have a certain auxiliary effect on the enhancement of SERS signals.

[0049] Preferably, in the step (1), the preparation method of the PSB Bragg reflector is: washing the silicon wafer with ethanol, water, and ethanol in sequence, drying, electrochemically etching, and then washing with water and drying;

[0050] In the electrochemical corrosion process, the corrosion liquid contains hydrofluoric acid and ethanol in a volume ratio of 25:25-30;

[0051] In the electrochemical corrosion process, 12 cycles are alternately cycled. One cycle process is: first, 115 mA / cm 2 The high current density was used for corrosion for 1s, and then 65mA / cm 2 The low current density corrosion is carried out for 1.2s, and after an interval of 0.8s, the next cycle of electrochemical corrosion is carried out;

[0052] In the step (2), the concentration of the silver nitrate solution is 0.04-0.06 mol / L, and the concentration of the chloroplatinic acid solution is 0.004-0.006 mol / L.

[0053] Further preferably, in the step (1), during the electrochemical corrosion process, the corrosion solution contains hydrofluoric acid and ethanol in a volume ratio of 25:28;

[0054] In the step (2), the PSB Bragg reflector is immersed in a 0.05 mol / L nitrate solution for 50 seconds, taken out, and then immersed in a 0.005 mol / L chloroplatinic acid solution.

[0055] An Ag@Pt@PSB composite SERS substrate is prepared by the above-mentioned preparation method, and the substrate has the advantages of excellent SERS performance, SERS response ability, good reproducibility and high sensitivity.

[0056] The Ag@Pt@PSB composite SERS substrate is used for label-free detection of celecoxib in serum. Based on the SERS response ability of the substrate to CXB, the substrate can be used for the concentration detection of CXB in serum.

[0057] A method for establishing a CXB model comprises the following steps:

[0058] (1) selecting healthy subjects, collecting fresh blood and performing centrifugation to obtain serum; and then using the serum to prepare celecoxib-serum samples of various concentrations;

[0059] (2) collecting data: dropping the celecoxib-serum sample onto the Ag@Pt@PSB composite SERS substrate of claim 4, and obtaining a SERS spectrum after the sample is in full contact with the substrate;

[0060] (3) Data preprocessing: After the SERS spectrum is preprocessed by baseline correction, a deep learning algorithm model, namely the CXB model, is established.

[0061] Preferably, in step (1), at least 20 samples of healthy subjects are selected;

[0062] Celecoxib-serum samples were at concentrations of 200, 500, 700, 1000, and 5000 ng / mL;

[0063] In the step (3), a GoogleNet, ResNet, or VGG deep learning algorithm model is established.

[0064] Further preferably, in the step (3), a GoogleNet deep learning algorithm model is established.

[0065] A CXB model, obtained by the above-mentioned establishment method, can be used for monitoring the concentration of CXB in serum and has great detection potential for the concentration of CXB in serum. This technology is expected to develop a portable, economical and highly accurate CXB concentration detection device.

[0066] Application of the above CXB model in a CXB concentration detection device.

[0067] Example 1.

[0068] The specific steps are as follows:

[0069] (1) Preparation of PSB

[0070] This embodiment uses P-type boron-doped silicon, and the crystal orientation <100> , resistivity 0.01-0.05Ωcm, thickness 400±25um.

[0071] The experimental preparation procedure was strictly carried out according to the report of Lv Xiaoyi et al. The experimental preparation procedure was as follows: the silicon wafer was cut into 2 cm × 2 cm pieces and ultrasonically cleaned in ethanol for 15 min. The silicon wafer was rinsed again with deionized water and ethanol, and fixed on a single-slot etching device after drying for electrochemical etching.

[0072] Electrochemical etching: The concentration of the etching solution is HF:C2H5OH=25:25 (v / v) to ensure the uniformity of the etching and obtain the photonic crystal with the required central wavelength.

[0073] According to the LabVIEW (National Instruments, USA) program, the corrosion process is: alternately cycle 12 cycles. One cycle process is: first at 115mA / cm -1 The high current density was used for corrosion for 1s, and then 65mA / cm-1 The low current density corrosion was carried out for 1.2s, and the next cycle of electrochemical corrosion was carried out after an interval of 0.8s. To ensure the uniformity of corrosion and sufficient fluoride, each layer had a time interval of 0.8s. The entire corrosion process was carried out in an indoor fume hood with an ambient temperature of room temperature.

[0074] Finally, in order to remove the residual HF on the surface, after the etching, it was rinsed with anhydrous ethanol and dried in a nitrogen atmosphere.

[0075] (2) In situ synthesis of Au@Ag-PSB substrate

[0076] Ag@PtNPs were synthesized in situ on PSB by immersion plating. Since the Si-H bonds on the PSB surface are reductive, silver nanoparticles were reduced on the PSB surface to obtain Ag@PtNPs substrate. Then, the platinum nanoparticles were reduced from H2PtCl6 solution by using the reducing property of AgNPs, thus assembling Ag@PtNPs on the PSB surface.

[0077] The process is as follows: the prepared PSB is immersed in a 0.04M AgNO3 solution for 60 seconds, then taken out to obtain the Ag@PtNPs substrate. The Ag@PtNPs substrate is then immersed in a 0.004M H2PtCl6 solution, rinsed with deionized water after 3.5 minutes, and dried to obtain the final Ag@Pt@PSB composite SERS substrate.

[0078] Example 2.

[0079] The specific steps are as follows:

[0080] (1) Preparation of PSB

[0081] This embodiment uses P-type boron-doped silicon, and the crystal orientation <100> , resistivity 0.01-0.05Ωcm, thickness 400±25um.

[0082] The experimental preparation procedure is as follows: the silicon wafer is cut into 2cm×2cm size and ultrasonically cleaned in ethanol for 15 minutes. The silicon wafer is rinsed again with deionized water and ethanol, and after drying, it is fixed on a single-tank etching device for electrochemical etching.

[0083] Electrochemical etching: The concentration of the etching solution is HF:C2H5OH=25:30 (v / v) to ensure the uniformity of the etching and obtain the photonic crystal with the required central wavelength.

[0084] According to the LabVIEW (National Instruments, USA) program, the corrosion process is: alternately cycle 12 cycles. One cycle process is: first at 115mA / cm -1The high current density was used for corrosion for 1s, and then 65mA / cm -1 The low current density corrosion was carried out for 1.2s, and the next cycle of electrochemical corrosion was carried out after an interval of 0.8s. To ensure the uniformity of corrosion and sufficient fluoride, each layer had a time interval of 0.8s. The entire corrosion process was carried out in an indoor fume hood with an ambient temperature of room temperature.

[0085] Finally, in order to remove the residual HF on the surface, after the etching, it was rinsed with anhydrous ethanol and dried in a nitrogen atmosphere.

[0086] (2) In situ synthesis of Au@Ag-PSB substrate

[0087] The prepared PSB was immersed in 0.06M AgNO3 solution for 40s and then taken out to obtain the Ag@PtNPs substrate. The Ag@PtNPs substrate was then immersed in 0.006M H2PtCl6 solution, rinsed with deionized water after 3min, and dried to obtain the final Ag@Pt@PSB composite SERS substrate.

[0088] Example 3.

[0089] A. Experimental part

[0090] 1. Materials and Reagents

[0091] AgNO3 (99.8%), hydrofluoric acid, methanol, and alcohol were purchased from Sinopharm Chemical Reagent Co., Ltd. Crystal violet, H2PtCl6·6H2O (Pt ≥ 37.5%), and celecoxib (99%) were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. Single crystal silicon was purchased from Tianjin Semiconductor Technology Institute (Tianjin, China). All chemicals were of analytical grade and used without further purification. Deionized water was used for all preparations except crystal violet and celecoxib.

[0092] 2. Preparation of SERS substrate

[0093] (1) Preparation of PSB Bragg reflector

[0094] P-type boron-doped single crystal silicon (crystal orientation <100> , resistivity 0.01-0.05Ωcm, thickness 400±25μm) was used as the material, and the PSB Bragg reflector was prepared by single-slot anode electrochemical etching method.

[0095] The experimental preparation procedure is as follows: the silicon wafer is cut into 2cm×2cm size and ultrasonically cleaned in ethanol for 15 minutes. The silicon wafer is rinsed again with deionized water and ethanol, and after drying, it is fixed on a single-tank etching device for electrochemical etching.

[0096] The operation method is: set the corrosion current density of the high and low refractive index layers to 65mA / cm respectively through the LabVIEW virtual instrument program (National Instruments Corporation, USA) 2 and 115mA / cm 2 , the corrosion time was 1.2s and 1s respectively, and the corrosion cycle was 3s. During the corrosion process, the electrolyte concentration used was HF:C2H5OH=25:28 (v:v) to ensure the uniformity of corrosion and obtain the photonic crystal with the required central wavelength. The entire corrosion process was carried out in an indoor fume hood at room temperature. In order to remove the residual HF on the surface, after the corrosion, it was rinsed with anhydrous ethanol and dried in a nitrogen atmosphere. The reflectance spectrum of PSB was measured using an ultraviolet spectrophotometer (Hitachi U-4100, purchased from Hitachi, Japan).

[0097] (2) In situ synthesis of Ag@PtNPs on PSB

[0098] In this embodiment, Ag@PtNPs were synthesized in situ on PSB by immersion plating. The operation process is: immerse the prepared PSB substrate in AgNO3 solution (0.05M, immersion time 50s). Then, using the reduction characteristics of AgNPs, platinum nanoparticles were reduced from H2PtCl6 solution (0.005M, immersion time 3min), thereby assembling Ag@PtNPs on the PSB surface. After taking out the product, rinse it with deionized water and dry it to obtain the final Ag@Pt@PSB composite SERS substrate.

[0099] 3. SERS spectroscopy measurement

[0100] (1) Detection of crystal violet

[0101] Crystal violet was used as a probe to verify the SERS performance of the Ag@Pt@PSB composite SERS substrate prepared above. -2 M crystal violet-ethanol standard solution and dilute it serially with ethanol to create 10 -5 -10 -11 A series of standard solutions of M. The detection was carried out using a confocal Raman spectrometer (BWS465-532S, BWTEK), using a 532nm laser, and acquiring spectral data under the conditions of a laser power of 35% and an objective lens of 20×. The integration time was 3s, and the average number of acquisitions was 3 times. During the experiment, the substrate was immersed in CV solutions of different concentrations for 3h, then taken out and tested after natural drying at room temperature. Each concentration sample was subjected to 5 independent repeated experiments.

[0102] (2) Detection of celecoxib-methanol solution

[0103] Accurately weigh 10 mg of celecoxib into a 10 mL volumetric flask, and add methanol to make a 1 g / L celecoxib stock solution. The above stock solutions are stored in a -18 °C refrigerator. Take the stock solution and dilute it with methanol step by step to a series of standard solutions with mass concentrations of 250, 100, 50, 20, 10, and 2 ng / mL, respectively.

[0104] Celecoxib was detected using an Optosky 785nm portable refrigerated Raman spectrometer (ATR3110-785, Optosky), with a laser power of 60mW, an average acquisition time of 3 times, and an integration time of 5s. Each time, 20 microliters of sample was dropped onto the Ag@Pt@PSB composite SERS substrate, and after sufficient contact with the substrate, the SERS spectrum of the substrate was obtained. All SERS spectral detections were performed at room temperature.

[0105] (3) Detection of celecoxib in serum

[0106] 20 serum samples were drawn from the blood of healthy subjects. Without using anticoagulants, 3 ml of fresh blood was taken from each patient and placed in a clean environment at 37°C for 1 hour. Centrifuged at 4°C for 5 minutes at high speed (4000r / min), the top transparent layer was extracted, and serum was obtained. The serum was distributed into centrifuge tubes and stored in a refrigerator at -80°C until used in experiments.

[0107] Thaw the serum sample, take 50μL of serum and place it in a 1.5mL centrifuge tube, add 300μL of methanol, centrifuge at 7000r / min for 40min, take the supernatant, dry it and then add methanol to 2mL. Take the celecoxib standard solution separately, add the purified serum, and prepare spiked celecoxib-serum samples equivalent to mass concentrations of 200, 500, 700, 1000 and 5000ng / mL. (Taking the normal daily intake of 400mg as an example, the highest celecoxib blood concentration in the human body is about 700ng / mL. The concentration gradient is set with 700 as the dividing line, and the high, medium and low ranges are selected for testing respectively)

[0108] Optosky 785nm portable refrigerated Raman spectrometer (ATR3110-785, Optosky) was used, with the laser power set to 60mW, the average acquisition times to 3 times, and the integration time to 5s. Each time, 20 microliters of sample was dropped onto the Ag@Pt@PSB composite SERS substrate, and after it was fully in contact with the substrate, the SERS spectrum of the substrate was obtained. All SERS spectral detections were performed at room temperature.

[0109] Serum Sample Approval Statement: All healthy human serum samples were obtained from the People's Hospital of Xinjiang Uygur Autonomous Region. The experimental protocols and serum samples of all subjects were approved by the Ethics Committee of the People's Hospital of Xinjiang Uygur Autonomous Region, and all methods were carried out in accordance with relevant guidelines and regulations. Informed consent was obtained from all subjects and / or their legal guardians.

[0110] 4. Deep Learning

[0111] All SERS spectra were baseline corrected using the spectrometer's built-in algorithm, and three deep learning algorithm models, GoogleNet, ResNet, and VGG, were established to distinguish the SERS spectral data of five different concentrations of celecoxib-spiked serum. Four parallel samples were set for each concentration gradient, and 10 spectral data were collected for each sample, for a total of 200 spectral data. In all models, SERS spectra were randomly selected as training and test spectral sets at a ratio of 4:1. The diagnostic ability of the proposed model was evaluated based on the accuracy, precision, recall rate, and receiver operating characteristic (ROC) curve of the model.

[0112] B. Results and Discussion

[0113] 1. Morphological characterization of SERS substrate

[0114] The morphology of the Ag@Pt@PSB composite SERS substrate was characterized. Figure 2 shown.

[0115] Figure 2 (a) shows the process of single crystal silicon being etched into a porous silicon Bragg reflector and loaded with Ag@PtNPs. AgNPs@PSB is obtained by placing PSB in AgNO3 solution. By placing the PSB substrate with fixed AgNPs in chloroplatinic acid solution, PtNPs will grow on the substrate surface under appropriate conditions.

[0116] Figure 2 (bc) are the SEM micrographs and UV reflectance spectra of PSB. The results show that the surface pore size of PSB is between 20-40nm, and the layer thickness is about 2.8μm, which is consistent with the designed thickness. PSB has a total of 12 periods, with high and low refractive index layers arranged alternately, which appear as alternating light and dark stripes in the cross-sectional view. This unique photonic crystal structure provides more binding sites for the substrate and the molecules to be tested, which is beneficial to promote the enhancement of Raman signals. At the same time, the high porosity and large specific surface area of ​​PSB improve the substrate's carrying capacity for nanoparticles, providing convenience for the in-situ growth of a large number of nanoparticles.

[0117] Figure 2(de) is the SEM morphology characterization of the composite SERS substrate and its particle size distribution. Many small Pt nanoparticles grow on the surface of AgNPs, forming raspberry-shaped Ag@PtNPs with many protrusions. These rough-surfaced nanoparticles grow evenly on the PSB surface, and there are many gaps between adjacent particles. This is conducive to increasing the interaction space with the molecules to be tested and forming a considerable number of "hot spots". The particle size range is concentrated between 100-400nm, and the average particle diameter is 222.95nm. In the SERS system, electromagnetic enhancement is mainly affected by the LSPR effect at the junction of the nanoparticle gap. The particle size and distribution have a significant effect on the "hot spot" area at the junction. The microstructure of the substrate is a key factor affecting its SERS performance. The porous silicon-based composite nanoparticles synthesized in this study have moderate particle size and gap, which is conducive to obtaining drug molecule fingerprint information with high repeatability and sensitivity, and is an excellent SERS substrate.

[0118] Figure 2 (fg) shows the main element analysis of the composite substrate in the mapping detection area. The image shows that the Ag and Pt elements are evenly distributed. This result provides support for the SEM and particle size analysis results.

[0119] Figure 2 (h) is the energy spectrum scanning electron microscope (EDS) detection image during the SEM analysis of the substrate, which further proves the existence and content of the main elements Ag, Pt, Si, and O in the sample. In addition to silicon (Si) (mainly from the substrate), silver (Ag) and platinum (Pt) mainly come from Ag@PtNPs on the PSB substrate, and oxygen (O) mainly comes from SiO2 and AgO.

[0120] XPS further confirmed the surface composition and valence state of the composite substrate. Figure 2 (il) is the XPS fitting diagram of the four main elements Si, Ag, Pt, and O on the substrate surface. Among them, the peaks of the Si element are located at 97.76eV and 103.38eV, respectively. Figure 2 As shown in (i), it complies with Si2p and Si-O. Figure 2 As shown in (j), in the Ag 3d spectrum, 368.08 eV and 374.28 eV correspond to the binding energy of AgO, respectively. 3 / 2 and Ag 3d 5 / 2 The binding energy peaks of are 374.88eV and 368.84eV, which are 6.0eV apart, and match the standard XPS spectrum of Ag. Figure 2 In the Pt 4f spectrum (k), the electron binding energy positions of 71.09 eV and 74.48 eV correspond to metal Pt 0 4f 7 / 2 and Pt0 4f 5 / 2 , while the electron binding energy positions of 72.08eV and 75.48eV correspond to Pt 2+ . Figure 2 In (l), the electron binding energy position of O element 533.23 eV represents CO and Si-O, and the electron binding energy positions 532.58 eV and 530.00 eV correspond to the 1s orbital binding energy of oxygen in C=O and AgO, respectively. The XPS analysis is in good agreement with the EDS results, and the above results indicate that Ag@PtNPs were successfully synthesized on the PSB surface.

[0121] 2. Performance Characterization of SERS Substrates

[0122] (1) Enhancement factor, reproducibility, and sensitivity

[0123] Crystal violet was selected as the probe molecule to evaluate the SERS activity of the prepared Ag@Pt@PSB substrate. -5 The SERS spectra and ordinary Raman spectra of CV of M, such as Figure 3 (a) shows that the composite SERS substrate (Ag@Pt@PSB) has a significant effect on CV enhancement. According to EF=(I SERS / I0)(C0 / C SERS ) estimates the enhancement factor of CV. SERS The concentration is 10 -1 The SERS signal intensity of CV at M, I0 is the concentration of 10 -2 Normal Raman spectrum intensity of CV at M. SERS is the concentration of the analyte during SERS detection, and C0 is the concentration of the test solution when measuring I0. From the formula, we can see that the EF of Ag@Pt@PSB is 5.64×10 5 .

[0124] The signal reproducibility and sensitivity of SERS substrate to probe molecules are key factors in practical applications. -5 The SERS spectra of CV of M at 20 points were analyzed, and the SERS spectra of these 20 points were examined at 1617 cm -1 The relative intensity distribution of characteristic peaks. Figure 3 As shown in (bc), the spectral signals of CV with the same concentration are consistent, with slight intensity fluctuations and a relative standard deviation (RSD) of 6.7%. This fully demonstrates that the prepared substrate SERS has good signal reproducibility and stable optical performance. Based on the signal uniformity of the SERS substrate, in order to evaluate the sensitivity of the substrate, the substrate was recorded at 10 -5 -10 -11 The SERS signal response within the M concentration range is shown in Figure 2. Figure 3 (d) The SERS signal intensity of CV molecules decreases as the concentration decreases. The minimum detection concentration of CV by the substrate is 10 -11 M. Figure 3 As shown in (e), CV at 1617 cm -1 The relative SERS intensity at the site showed a good linear relationship with the concentration (R 2 =0.9888). This indicates that the substrate has high sensitivity and can perform qualitative and quantitative detection of CV at low concentrations.

[0125] (2) Comparison of different substrates and long-term stability

[0126] The enhancement effect of Ag@Pt@PSB composite SERS substrate was compared with that of Ag@PSB and Pt@PSB. -5 The CV of M was used as a probe for detection. Figure 3 As shown in (f), Pt@PSB basically does not show obvious enhancement effect, while Ag@Pt@PSB and Ag@PSB have obvious enhancement effects. -1 Compared with the relative SERS intensity at , the enhancement effect of the composite substrate is 1.7 times that of the Ag@PSB substrate.

[0127] In order to verify the long-term stability of the substrate, such as Figure 3 (gh) shows the SERS signals of Ag@PSB and Ag@Pt@PSB after storage for 0 days and 60 days, respectively. -5 M) as the probe for detection. The experimental results show that at 1617cm -1 At the characteristic peak, the SERS signal of the Ag@Pt@PSB substrate with a storage time of 0 days is the strongest. After a storage period of 60 days, the SERS signals of both Ag@Pt@PSB and Ag@PSB decrease. -1 At the characteristic peak, the Ag@Pt@PSB signal decreased by 2.57 times, while the decrease of Ag@PSB was more significant, decreasing by 5.97 times compared with 0 days. Figure 3 The results of (f) show that the Ag@Pt@PSB substrate has better performance than other traditional substrates, and that the Ag@Pt@PSB substrate has the ability of long-term storage.

[0128] The observation of remarkable SERS responsiveness and long-term stability of Ag@Pt@PSBs demonstrates the feasibility of subsequent substrate applications for the detection of therapeutic drugs for autoimmune diseases.

[0129] 3. SERS detection of celecoxib

[0130] (1) SERS detection of celecoxib in methanol

[0131] The SESR spectrum of celecoxib-methanol standard solution (250 ng / mL) was obtained using the Ag@Pt@PSB prepared above as the substrate. Figure 4 (a) Compared with ordinary Raman, the substrate has a significant SERS enhancement effect on CXB (EF = 2.5 × 10 5 ). Table 1 shows the assignment of the main Raman vibration modes of the CXB samples detected.

[0132] Table 1 SERS peak and vibration mode assignments based on literature

[0133]

[0134] In order to investigate the reproducibility of CXB spectral signals, 20 sets of randomly acquired CXB (250 ng / mL) spectral data were compared. Figure 4 As shown in (bd), the SERS spectra signals obtained by the same concentration of CXB on the substrate are basically the same, and the color bands at the same peak are relatively uniform, indicating that the SERS signal intensity is uniform. At the same time, the SERS spectra of these 20 points at 1042 cm -1 and 1230cm -1 The relative intensity distribution deviation at the characteristic peaks is small, with RSDs of 3.5% and 5.5% respectively. Therefore, this embodiment selects these two characteristic peaks as the stable peak positions of celecoxib in methanol solution, and further investigates the quantitative detection ability of the substrate for CXB molecules. -1 and 1230cm -1 The variation of vibration with concentration and the linear fitting curve are shown in Figure 4 (ef) As shown. In the concentration range of 250-5 ng / mL, the characteristic Raman peak of CXB can be clearly observed, and the Raman signal disappears when the concentration is 2 ng / mL. The SERS signal intensity of CXB is positively correlated with the concentration. -1 (R 2 =0.9840) and 1230cm -1 (R 2 The above results show that the substrate has stable signal reproducibility and highly sensitive quantitative analysis capability for CXB in a low concentration range.

[0135] (2) SERS detection of celecoxib in serum

[0136] In order to investigate the detection potential of Ag@Pt@PSB composite SERS substrate in serum containing CXB, CXB was added to the serum according to the standard addition method to prepare spiked serum samples with different CXB concentrations. Serum proteins in serum occupy the active sites on the surface of Ag@PtNPs, which will affect the SERS detection of serum drug concentration. Methanol purification can destroy the protein surface and precipitate serum proteins, thereby reducing the interference of complex components in serum. Figure 5 As shown in (a), after purifying the protein in serum, the serum characteristic peaks decreased significantly, which will produce less background interference for the detection of CXB. Compared with the detection results in standard methanol solution, the SERS spectrum of celecoxib in the pretreated serum system has slight changes, and some characteristic peaks are quenched or shifted, but several more significant CXB characteristic peaks can still be observed.

[0137] The proposed quantitative analysis method for CXB blood concentration was verified by evaluating the limit of detection (LOD) and recovery rate. For the SERS spectrum of celecoxib in spiked human serum, the concentration detection range was 200 ng / mL to 5000 ng / mL. Five experiments were performed in each concentration gradient. The concentration change spectrum plotted using the mean and standard deviation is shown in Figure 2. Figure 5 As shown in (b), the SERS peak intensity decreases rapidly with the decrease of CXB concentration, and gradually slows down in the low concentration range. In the detection range, the 855cm -1 The linear curve of CXB concentration and SERS peak relative intensity at the characteristic peak is shown in Figure 2. Figure 5 (c) is shown. The linear regression equation after fitting is SERS =0.31c+52.08(R 2 =0.9986). Unspiked healthy human deproteinized serum was used as a blank control. By calculation, the LOD of CXB in serum was 20.22 ng / mL. According to the fitting curve of CXB in the serum system, the recovery experiment was carried out in the linear range to predict the concentration of CXB. The results are shown in Table 2. The average recovery was 97.49%-107.65%, and the RSD was 3.43%-8.89%, indicating that the method has high accuracy and stability. Table 3 lists different detection methods for celecoxib in different solutions, as well as their LOD and recovery. The results show that the SERS detection method based on Ag@Pt@PSB substrate has good detection limit and recovery rate compared with most other methods. This result demonstrates the ability of the SERS detection platform in blood drug concentration detection, and it is expected that the substrate will be applied to the exploration of therapeutic drug monitoring in clinical serum samples.

[0138] Table 2 Average recovery and relative standard deviation of CXB in human serum (n=5)

[0139]

[0140]

[0141] Table 3 Comparison of currently reported detection limits and recoveries of CXB

[0142]

[0143] 4. Five-classification model of drug concentration based on deep learning-SERS

[0144] Based on the verified ability of Ag@Pt@PSB composite SERS substrate to detect CXB blood concentration, the sample data set was expanded to 200. The representative Raman signals of spiked serum samples with different concentrations of CXB were analyzed. Figure 6 As shown in (b), CXB at 855cm -1 、1042cm -1 、1608cm -1 As the concentration difference decreases, the spectral intensities obtained from samples with different concentration gradients overlap significantly, indicating that when the concentration difference is small and the amount of data is large, the accuracy of distinguishing blood drug concentration only by SERS spectrum will decrease. Therefore, this embodiment combines the SERS signal obtained on the substrate with the deep learning model, hoping to improve the accuracy of blood drug concentration evaluation. Figure 6 (a) shows the process of training the deep learning model based on the acquired SERS spectral dataset.

[0145] In order to evaluate the matching degree between different algorithms and SERS substrates, the accuracy, precision, recall rate and AUC value of ROC curve of three classification models, GoogleNet, ResNet and VGG, were further compared, and F1-score was introduced as the harmonic mean of precision and recall rate indicators. The performance indicators of the three models were analyzed according to the confusion matrix and ROC curve, such as Figure 6 (ce) and shown in Table 4. It can be seen from the confusion matrix that the three models have good classification effects on five concentrations of CXB-spiked serum samples. The ROC curve can understand the classification ability of the model and the detection performance of the SERS substrate from another perspective. Among them, the combination of SERS spectral data and GoogleNet algorithm performed best in the classification of celecoxib concentration in serum. Under the premise of high difficulty in the five-category classification, the accuracy rate was 84.17%, the F1-score was 83.29%, and the area under the curve (AUC) reached 0.87, showing a better classification effect than other models. This shows that the model has efficient discrimination ability in accurately identifying different blood drug concentrations after being combined with the Ag@Pt@PSB substrate.

[0146] The above results show that the Ag@Pt@PSB composite SERS substrate prepared in the present invention provides a simple and sensitive strategy to quantify different concentrations of celecoxib in serum by combining with the GoogleNet model. The method can classify five drug concentrations in serum with good accuracy. In the case of large data volume or only a small difference in blood drug concentration, the SERS detection technology driven by the deep learning model effectively solves the problem of difficulty in distinguishing by relying solely on SERS spectra. This proves that integrating the deep learning classification algorithm into the SERS platform is expected to improve the discrimination ability of SERS technology in more complex real serum samples, and has broad application prospects in the fields of TDM and personalized medicine.

[0147] Table 4 Comparison of classification effects of three deep learning models

[0148]

[0149] It can be seen from the examples that the present invention is based on a porous silicon Bragg reflector (PSB) and uses a simple in-situ reduction method to prepare an Ag@Pt@PSB composite SERS substrate. The composite substrate exhibits excellent SERS performance (EFs for crystal violet = 5.64×10 5 , RSD=6.7%, R 2 =0.9888). Because of the introduction of platinum, the substrate has better enhancement performance and long-term stability than pure silver substrate. The substrate was successfully applied to the trace SERS detection of celecoxib in the low concentration range for the first time. The Ag@Pt@PSB composite SERS sensor has a significant enhancement effect on CXB (EFs=2.5×10 5 ), good reproducibility (RSD < 6%) and high sensitivity (linear detection range of 250-5 ng / mL, R 2 =0.9840). The substrate was applied to the concentration monitoring of CXB in serum, and the LOD (20.22 ng / mL) and recovery rate (97.49%-107.65%) were both within a reasonable range.

[0150] Furthermore, the present invention has constructed a new method combining SERS spectroscopy and deep learning algorithms to classify and identify five different concentrations of CXB in serum, achieving an accuracy rate of more than 84%. The integration of this SERS substrate and deep learning provides a non-invasive detection method for the determination and measurement of therapeutic drugs for diseases such as osteoarthritis and rheumatoid arthritis. Therefore, the technical solution of the present invention is expected to provide a reference and basis for more accurate qualitative determination of drug concentrations in the serum of arthritis patients through SERS spectral diagnostic methods and classification models in future clinical tests. This progress is expected to develop a fast, sensitive, highly reliable, and precision detection SERS detection system, driven by spectral diagnostic technology and artificial intelligence, to help medical practitioners develop personalized treatment strategies for patients.

[0151] The above is only a preferred embodiment of the embodiment of the present invention, and does not impose any form of limitation on the embodiment of the present invention. Any simple modification, equivalent changes and modifications made to the above embodiment based on the technical essence of the embodiment of the present invention are still within the scope of the technical solution of the embodiment of the present invention.

Claims

1. A method for preparing a Ag@Pt@PSB composite SERS substrate, characterized in that: The following steps are involved: (1) Preparation of PSB Bragg reflectors; (2) After immersing the PSB Bragg reflector in a silver nitrate solution for 40-60 seconds, taking it out, and then immersing it in a chloroplatinic acid solution for at least 3 minutes, washing it with water, and drying it to obtain the Ag@Pt@PSB composite SERS substrate.

2. The preparation method according to claim 1, characterized in that: In the step (1), the preparation method of the PSB Bragg reflector is: washing the silicon wafer with ethanol, water, and ethanol in sequence, drying, electrochemically etching, and then washing with water and drying; In the electrochemical corrosion process, the corrosion liquid contains hydrofluoric acid and ethanol in a volume ratio of 25:25-30; In the electrochemical corrosion process, 12 cycles are alternately cycled. One cycle process is: first, 115 mA / cm 2 The high current density was used for corrosion for 1s, and then 65mA / cm 2 The low current density corrosion is carried out for 1.2s, and after an interval of 0.8s, the next cycle of electrochemical corrosion is carried out; In the step (2), the concentration of the silver nitrate solution is 0.04-0.06 mol / L, and the concentration of the chloroplatinic acid solution is 0.004-0.006 mol / L.

3. The preparation method according to claim 2, characterized in that: In the step (1), during the electrochemical corrosion process, the corrosion solution contains hydrofluoric acid and ethanol in a volume ratio of 25:28; In the step (2), the PSB Bragg reflector is immersed in a 0.05 mol / L nitrate solution for 50 seconds, taken out, and then immersed in a 0.005 mol / L chloroplatinic acid solution.

4. An Ag@Pt@PSB composite SERS substrate, characterized in that: The preparation method is described in any one of claims 1 to 3.

5. Use of the Ag@Pt@PSB composite SERS substrate according to claim 4 in the detection of celecoxib.

6. A method for establishing a CXB model, characterized in that: The following steps are involved: (1) selecting healthy subjects, collecting fresh blood and performing centrifugation to obtain serum; and then using the serum to prepare celecoxib-serum samples of various concentrations; (2) collecting data: dropping the celecoxib-serum sample onto the Ag@Pt@PSB composite SERS substrate of claim 4, and obtaining a SERS spectrum after the sample is in full contact with the substrate; (3) Data preprocessing: After the SERS spectrum is preprocessed by baseline correction, a deep learning algorithm model, namely the CXB model, is established.

7. The establishment method according to claim 6, characterized in that: In the step (1), at least 20 samples of healthy subjects are selected; Celecoxib-serum samples were at concentrations of 200, 500, 700, 1000, and 5000 ng / mL; In the step (3), a GoogleNet, ResNet, or VGG deep learning algorithm model is established.

8. The establishment method according to claim 7, characterized in that: In the step (3), a GoogleNet deep learning algorithm model is established.

9. A CXB model, characterized in that: Obtained by the establishment method described in any one of claims 6-8.

10. Application of the CXB model according to claim 9 in a CXB concentration detection device.

Images