OCT-SERS combined multi-mode gastric cancer tissue diagnosis system and method

Abstract

The invention belongs to the technical field of medical image processing, and discloses an OCT-SERS combined multi-mode gastric cancer tissue diagnosis system and method. A set of OCT-SERS combined system light path which shares a sample arm and is high in registration precision realizes high-precision registration of an OCT imaging range and an SERS detection position; the surface of a biological tissue is covered with a hydrogel film doped with silver nanoparticles, then SERS signal detection is carried out, and non-destructive SERS enhancement of the tissue is achieved; multi-modal gastric cancer diagnosis is carried out by adopting an artificial intelligence algorithm through collection of normal and cancerous gastric wall tissue OCT images and SERS spectral data and establishment of a database. According to the invention, by establishing the normal and cancer tissue OCT image and SERS spectrum database and carrying out artificial intelligence learning, real-time accurate diagnosis of pathological tissues can be realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical image processing, and particularly relates to a multimodal gastric cancer tissue diagnosis system and method combining OCT and SERS. Background Art

[0002] The diagnosis process of cancer tissue generally includes four levels. The first level is clinical diagnosis, which relies on symptoms and signs to analyze the possibility of a patient having gastric cancer. The second is auxiliary examination. By means of physical and chemical methods and using instrument equipment, on the basis of clinical suspicion of a tumor, imaging, laboratory and other examinations are carried out to make a diagnosis. Through various imaging examinations, such as CT, magnetic resonance imaging, X-ray, ultrasound examination, etc., abnormal masses in the patient's body can be found, and through laboratory examinations, abnormal tumor markers can be found. Auxiliary examinations can only determine that a mass has appeared in the patient, but cannot confirm the tumor. The third is surgical diagnosis. Through surgery or various endoscopic examinations, especially laparoscopy, tumor nodules on the peritoneum and pleura that cannot be seen by some images can be directly seen with the naked eye or on the screen image to make a diagnosis. The fourth is pathological diagnosis that can confirm the benign and malignant nature of the tumor, which is the gold standard for the diagnosis of current tumor-related diseases. Its disadvantage is that the time required for the examination is long, and it takes nearly an hour to complete the frozen pathological diagnosis. The current problem encountered in clinical practice is that there is no other effective method except for the gold standard of frozen pathological examination of suspicious lesions.

[0003] Optical methods are the most promising auxiliary diagnostic means after traditional ultrasound imaging, nuclear magnetic resonance imaging and other technologies. In recent years, optical imaging and spectroscopy analysis have been widely used in the fields of biomedicine and others. The microscope magnifies the surgical field of view, reduces the invasion of normal tissues during cancer tissue resection, and at the same time, provides a clearer field of view for doctors, making the surgical operation more accurate and safe, but it can only provide the surface information of tissues. In contrast, Optical Coherence Tomography (OCT) can provide high-resolution tomographic images of tissues in a non-contact manner, with the characteristics of non-invasiveness, high detection sensitivity, fast imaging speed, etc., especially with a resolution on the order of micrometers and a penetration depth on the order of millimeters. Raman Spectrum (RS) technology generates the inherent molecular vibration spectrum of a substance through inelastic collisions between photons and matter molecules to reflect the structural information of molecules, and can provide information on changes in the structure, conformation and quantity of the main components such as proteins and nucleic acids inside cells. However, the weak Raman signal characteristics inherent in biological tissues greatly limit its application in the field of disease diagnosis. Surface-enhanced Raman spectroscopy (SERS) can solve this problem. When a laser interacts with the surface of nanostructures, an electric field enhancement effect will occur. The localized surface plasmons on the surface of nanostructures can localize the electromagnetic field and enhance the vibration signals of nearby molecules. This technology can overcome the limitation that the weak Raman signal inherent in tissues is easily masked by its own fluorescence background and can be used for cancer detection.

[0004] If the OCT and SERS technologies are combined, it is expected to judge in real time whether the tissue is cancerous and the type of cancer tissue according to the morphological changes of the tissue provided by the OCT technology and the composition and conformational changes of the main components of cells provided by the SERS technology during the operation, so as to achieve precise resection or treatment. In the past ten years or so, some research groups have begun to attempt to develop the combined technology of OCT and RS. This technology can simultaneously perform structural imaging and biochemical examination of tissues, providing a more comprehensive means for clinical diagnosis. Although the SERS technology shows great advantages in Raman signal enhancement, all current research is only limited to the reports on the combination of OCT and RS technologies.

[0005] Through the above analysis, the problems and defects of the existing technology are as follows: The Raman scattering signal of biological tissues themselves is relatively weak, and combining the OCT and RS optical paths will inevitably lead to a further reduction in the collection efficiency of the Raman scattering signal, and there is still room for improvement in the registration accuracy of the existing combined technology of OCT and RS in the optical path, providing guidance for accurately locating the cancerous area of tissues. Summary of the Invention

[0006] To overcome the problems existing in the related technologies, the disclosed embodiments of the present invention provide a multimodal gastric cancer tissue diagnosis system and method using OCT-SERS. Specifically, it relates to a multimodal gastric cancer tissue diagnosis method using OCT-SERS technology.

[0007] The technical solution is as follows: A multimodal gastric cancer tissue diagnosis method using OCT-SERS includes:

[0008] S1. Establish the optical path of an OCT-SERS combined system with a shared sample arm to achieve the registration of the OCT imaging range and the SERS detection position, and the registration error is less than 30 micrometers; the shared sample arm uses a spatial optical path form to make the OCT imaging optical path and the SERS detection optical path share a galvanometer and a scanning lens;

[0009] S2. Use a hydrogel film doped with silver nanoparticles to cover the surface of biological tissue and then detect the SERS signal to achieve non-invasive SERS enhancement of gastric cancer tissue;

[0010] S3. Obtain the OCT images and SERS spectral data of normal and cancerous gastric wall tissues and establish a database;

[0011] S4. Perform multimodal gastric cancer diagnosis on the OCT images and SERS spectra of normal and gastric cancer tissues.

[0012] In step S1, to achieve the registration of the OCT imaging range and the SERS detection position, it includes: the registration of the moving directions of the scanning galvanometer and the electric displacement platform and the registration of the focus positions of OCT imaging and SERS detection;

[0013] The registration of the moving directions of the scanning galvanometer and the electric displacement platform includes: placing polystyrene microspheres on the stage and setting the OCT imaging range; moving the stage, adjusting the position of the polystyrene microspheres in the OCT imaging range so that the imaging range appears in the upper, lower, left, and right parts of the scanning range respectively, and using a program to calculate the center coordinates; comparing the horizontal and vertical angle changes of the microsphere center in the image before and after movement, calculating the angle difference between the galvanometer scanning direction and the stage movement direction, and adjusting the scanning angle of the scanning galvanometer according to the existing angle difference to achieve the registration of the two;

[0014] The registration of the focus positions of OCT imaging and SERS detection includes: placing polystyrene microspheres on the stage and performing OCT imaging, adjusting the OCT image of the microspheres to the center position of the scanning range, and performing spectral detection; adjusting the two-dimensional adjustment frame at the laser incident light to the position where the spectral signal is the largest to achieve the registration of the OCT and SERS optical paths;

[0015] During the process of testing the registration accuracy, first fix the stage and place four polystyrene microspheres. Move the scanning galvanometer to collect spectral data, perform mapping on the collected spectra, and use a program to calculate the center coordinates of the circle. Control the scanning range of the scanning galvanometer to perform OCT imaging on the plastic microspheres, use a program to calculate the center coordinates of the circle, and calculate the registration error based on the difference in the center coordinates between the spectral mapping and the OCT image.

[0016] In step S2, the preparation of the hydrogel film doped with silver nanoparticles includes:

[0017] Drug preparation and solution preparation: The cross-linking agent for preparing the hydrogel is polyethylene glycol diacrylate (PEGDA); the photoinitiator is hydroxymethyl phenyl ketone (HMPP); weigh silver nitrate powder and add water to prepare a silver nitrate solution; weigh hydroxylamine hydrochloride powder and add water to prepare a hydroxylamine hydrochloride solution; weigh sodium hydroxide powder and prepare a sodium hydroxide solution; mix the above-prepared hydroxylamine hydrochloride solution and sodium hydroxide solution, shake well and ultrasonicate to prepare a mixed solution of hydroxylamine hydrochloride and sodium hydroxide;

[0018] Preparation process: Stir the prepared silver nitrate solution, add the above-prepared mixed solution of hydroxylamine hydrochloride and sodium hydroxide, stir and react to obtain silver nanocolloid; extract the silver nanocolloid solution, centrifuge and take the lower-layer high-concentration silver nanocolloid; take an appropriate amount of PEGDA and mix it with the silver nanocolloid, add an appropriate amount of photoinitiator HMPP and then perform ultrasonic treatment to make them fully mixed; finally, place the uniformly mixed hydrogel precursor solution in a mold and perform ultraviolet light irradiation for curing treatment;

[0019] In step S2, the detection of SERS signals includes: Place the tissue on an aluminum sheet, cover the surface with a hydrogel film doped with silver nanoparticles, and perform SERS detection with a 785nm laser, with a laser power of 30mW and an integration time of 3s.

[0020] In step S3, the acquisition and database establishment of OCT images and SERS spectral data of normal and cancerous gastric wall tissues include:

[0021] S301, Collect gastric cancer tissue specimens, cover the surface of the tissue with a hydrogel film doped with silver nanoparticles for SERS signal acquisition, use 1.5mm×1.5mm as the OCT imaging range, and perform SERS signal acquisition within the range, with an interval of 300μm between each sampling point;

[0022] S302, Collect OCT images and SERS spectral data within the range of 1.5mm×1.5mm;

[0023] S303, Display the OCT images and SERS spectral information of normal and cancerous gastric tissues.

[0024] In step S4, for the multimodal gastric cancer diagnosis of OCT images and SERS spectra of normal and gastric cancer tissues, it includes:

[0025] S401. Extract the features of two-dimensional image data using ResNet50;

[0026] S402. Extract the features of one-dimensional spectral data using ResNet50;

[0027] S403. Concatenate the two 1×2048 one-dimensional vectors obtained by the two-dimensional ResNet50 and the one-dimensional ResNet50 in the fully connected layer, and then send them into the Transformer network for classification and recognition to achieve the fusion and diagnosis of multimodal data.

[0028] In step S401, extracting the features of two-dimensional image data using ResNet50 includes:

[0029] S4011. Input data;

[0030] S4012. Convolution operation: The convolution kernel slides on the image, and extracts the important features of each local area to generate a feature map;

[0031] S4013. The first layer of convolution and pooling;

[0032] S4014. Use the residual block to learn the residuals of the convolutional layer;

[0033] S4015. The change in the size of the feature map;

[0034] S4016. Use global average pooling to average the values in each feature map to obtain an output of 1×1×2048;

[0035] S4017. Use the fully connected layer to flatten the feature map into a one-dimensional vector with a length of 2048;

[0036] In step S402, extracting the features of one-dimensional spectral data using ResNet50 includes: The convolution kernel slides in one-dimensional space, extends the convolution operation from the two-dimensional plane to the one-dimensional sequence, and flattens the feature map into a one-dimensional vector with a length of 2048.

[0037] Another object of the present invention is to provide a multimodal gastric cancer tissue diagnosis system using OCT-SERS in combination. This system implements the multimodal gastric cancer tissue diagnosis method using OCT-SERS in combination. This system realizes the partial sharing of the spatial optical paths of the two modalities through self-designed and built optical paths, and realizes the registration of the imaging and spectral acquisition positions, specifically including:

[0038] Optical path of an OCT-SERS combined system with a high registration accuracy sharing a sample arm, which is used to achieve high-precision registration of the OCT imaging range and the SERS detection position; the shared sample arm adopts a spatial optical path form to share a galvanometer and a scanning lens for the OCT imaging optical path and the SERS detection optical path;

[0039] Non-invasive SERS signal enhancement module for gastric cancer tissue, which is used to detect SERS signals after covering the surface of biological tissue with a hydrogel film doped with silver nanoparticles, so as to achieve non-invasive SERS enhancement of the tissue;

[0040] Database establishment module, which is used to acquire and establish a database for OCT images and SERS spectral data of normal and cancerous gastric wall tissues;

[0041] Multi-modal diagnosis module, which is used for multi-modal gastric cancer diagnosis of OCT images and SERS spectra of normal and gastric cancer tissues.

[0042] Furthermore, the OCT imaging optical path is used for non-invasive and high-resolution tomography of biological tissues, and the depth information of biological tissues is obtained by using the coherence of the reference light and the sample light;

[0043] The OCT imaging optical path includes an OCT basic control unit, an optical fiber collimator, a beam splitter, a high reflector, an electrically controlled flip mirror, a high-pass dichroic mirror, a scanning galvanometer, a scanning lens, and an electric displacement platform;

[0044] The broadband light source emitted by the OCT basic control unit is collimated by the optical fiber collimator and then split into a sample light beam and a reference light beam by the beam splitter; the sample light is reflected by the electrically controlled flip mirror and enters the scanning lens through the high-pass dichroic mirror and the scanning galvanometer, and finally focuses on the sample surface;

[0045] The reference light is reflected back to the beam splitter by the high reflector, and after interfering with the backscattered light of the sample, it is transmitted back to the OCT basic control unit; the backscattered light of the sample is the scattered light of the sample on the electric displacement platform, which passes through the scanning lens, the scanning galvanometer, the high-pass dichroic mirror, and the electrically controlled flip mirror and finally coheres with the reference light at the beam splitter.

[0046] Furthermore, the alignment wavelength of the optical fiber collimator is 900 nm, the output light diameter is 2 mm, and the reflection-transmission ratio of the beam splitter is 3:7.

[0047] Furthermore, the SERS detection optical path is used to detect the chemical components of the suspicious lesion area of the tissue to judge whether other chemical substances are secreted due to the lesion;

[0048] Specifically, it includes a laser, a collimating lens, a first beam expander and a second beam expander, a band-pass filter, a high-pass dichroic mirror, a scanning galvanometer, a scanning lens, a high-pass filter, a collection lens, and a spectrometer;

[0049] The emitted light of the laser passes through a collimating lens, a first beam expander, and a second beam expander, and then the laser with wavelengths other than 785 nm is filtered out by a band-pass filter. Subsequently, the laser is reflected by a high-pass dichroic mirror, and finally reflected by a scanning galvanometer into a scanning lens and focused on the surface of the sample. At this time, a hydrogel film doped with silver nanoparticles is covered on the surface of the biological tissue. Both the SERS signal and the Rayleigh scattering signal of the sample are collected by the scanning lens, and then reflected by the scanning galvanometer, the high-pass dichroic mirror, and the high-pass filter in sequence to filter out the Rayleigh light, and finally collected by the collecting lens into the spectrum;

[0050] The laser has a central wavelength of 785 nm and is used for the excitation of SERS signals;

[0051] The spectrometer is used to collect SERS signals.

[0052] The Raman scattering signal of the biological tissue itself is weak, and combining the OCT and RS optical paths will inevitably lead to a further reduction in the collection efficiency of the Raman scattering signal. However, through the surface-enhanced Raman scattering of the present invention, the Raman scattering signal of the tissue is greatly enhanced, enabling OCT and SERS to be combined without affecting their signals. Combining all the above technical solutions, the beneficial effects of the present invention are as follows: The present invention solves the defect of the inherently weak Raman scattering signal of biological tissues by covering a hydrogel film doped with silver nanoparticles on the surface of biological tissues, significantly enhancing its Raman scattering signal. This enables OCT and SERS to be combined without affecting the collection of their signals. In addition, after the integration of the two optical paths, the Mapping of SERS and the perfect matching of the OCT tomographic images can be achieved, realizing the dual-modal observation of the morphology and composition of tissues. By establishing an OCT image and SERS spectrum database of normal and cancer tissues and performing artificial intelligence learning, the present invention can achieve real-time and accurate diagnosis of diseased tissues. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with the present disclosure, and are used together with the specification to explain the principles of the present disclosure;

[0054] Figure 1 It is a schematic diagram of the optical path of an OCT-SERS combined system with high registration accuracy of the common sample arm of the multi-modal gastric tissue diagnosis system provided by the embodiment of the present invention;

[0055] Figure 2 It is a schematic diagram of the registration situation between the galvanometer scanning direction and the moving direction of the stage provided by the embodiment of the present invention;

[0056] Figure 3It is a schematic diagram of the registration of the optical path focusing positions of OCT and SERS provided by an embodiment of the present invention. Among them, (a) is the OCT image and the center position coordinates of four polystyrene microspheres, and (b) is the Mapping image and the center position coordinates of four polystyrene microspheres;

[0057] Figure 4 It is a flowchart of a multimodal gastric cancer tissue diagnosis method using OCT-SERS combination provided by an embodiment of the present invention;

[0058] Figure 5 They are the OCT images and pathological section effect diagrams of normal and cancerous gastric wall tissues provided by the present invention; Figure A is the OCT tomographic image of normal gastric wall tissue, Figure B is the pathological section image of normal gastric wall tissue, Figure C is the OCT tomographic image of cancerous gastric wall tissue, Figure D is the section image of glandular structure in cancer cells, and Figure E is the pathological section image of cancerous gastric wall tissue;

[0059] Figure 6 It is a comparison diagram of RS and SERS spectra of normal and cancerous gastric wall tissues provided by the present invention;

[0060] Figure 7 It is a schematic diagram of a multimodal neural network provided by the present invention;

[0061] In the figure: 1. Fiber collimator; 2. Beam splitter; 3. High reflector; 4. Electrically controlled flip mirror; 5. High-pass dichroic mirror; 6. Scanning galvanometer; 7. Scanning lens; 8-1. Collimating lens; 8-2. First beam expander; 8-3. Second beam expander; 8-4. Collection lens; 9. Band-pass filter; 10. High-pass filter; 11. Electric displacement platform. Detailed implementation manners

[0062] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the detailed implementation manners of the present invention in conjunction with the accompanying drawings. Many specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific implementations disclosed below.

[0063] The innovation of the present invention lies in that: optical technology can achieve non-invasive and non-destructive lesion detection of tissues. Optical Coherence Tomography (OCT) can rapidly and depth-resolve tissue morphological imaging, but lacks detailed molecular contrast; although Raman Scattering (RS) can quantitatively analyze the biomolecular content of tissues, the inherently weak Raman scattering characteristics of biological tissues greatly limit its application, while Surface-Enhanced Raman Scattering (SERS) technology can perfectly solve this problem. Therefore, combining OCT and SERS technologies can determine in real time whether tissues are diseased and improve the treatment efficiency. The present invention solves the problems of high-precision registration of the optical path of the combined system, non-destructive SERS signal enhancement of biological tissues, and multimodal data analysis of SERS spectra and OCT images based on artificial intelligence technology.

[0064] In the prior art, when OCT (Optical Coherence Tomography) and RS (Raman Scattering) technologies are combined, the problem of inherently weak Raman scattering of biological tissues cannot be solved. The present invention can solve this problem, and improves the signal by fabricating a hydrogel film doped with silver nanoparticles. In the combined optical path of the prior art, the high-precision registration work of the optical path described in the present invention and the subsequent multimodal data analysis based on artificial intelligence are not carried out. Therefore, the positive effects of the present invention cannot be achieved.

[0065] Example 1. The multimodal gastric cancer tissue diagnosis system (OCT-SERS combined system optical path) provided by the embodiment of the present invention solves the problem of high-precision registration of imaging and spectral acquisition positions through an independently designed and built partial spatial optical path sharing and adjustment method. Specifically, it includes:

[0066] The OCT-SERS combined system optical path with high registration accuracy of the shared sample arm is used to achieve high-precision registration of the OCT imaging range and the SERS detection position; the shared sample arm adopts the form of a spatial optical path to share the galvanometer and scanning lens 7 for the OCT imaging optical path and the SERS detection optical path;

[0067] The non-destructive SERS signal enhancement module for gastric cancer tissue is used to detect the SERS signal after covering the surface of biological tissues with a hydrogel film doped with silver nanoparticles, realizing non-destructive SERS enhancement of tissues;

[0068] The database establishment module is used for obtaining and establishing the OCT image and SERS spectral data of normal and cancerous gastric wall tissues;

[0069] The multimodal diagnosis module is used for multimodal gastric cancer diagnosis of the OCT images and SERS spectra of normal and gastric cancer tissues.

[0070] As Figure 1As shown, the OCT imaging optical path is used for non-invasive and high-resolution tomography of biological tissues, and the depth information of biological tissues is obtained by using the coherence of the reference light and the sample light;

[0071] The OCT imaging optical path includes an OCT basic control unit, a fiber collimator 1, a beam splitter 2, a high reflector 3, an electronically controlled flip mirror 4, a high-pass dichroic mirror 5, a scanning galvanometer 6, a scanning lens 7, and an electric displacement platform 11;

[0072] The broadband light source emitted by the OCT basic control unit is collimated by the fiber collimator 1 and then split by the beam splitter 2 into a sample light beam and a reference light beam; the sample light is reflected by the electronically controlled flip mirror 4 and then enters the scanning lens 7 through the high-pass dichroic mirror 5 and the scanning galvanometer 6, and finally focuses on the sample surface;

[0073] The reference light is reflected back to the beam splitter 2 (beam splitting cube) by the high reflector 3, and after interfering with the backscattered light of the sample, it is transmitted back to the OCT basic control unit; the backscattered light of the sample is the scattered light of the sample on the electric displacement platform 11, which passes through the scanning lens 7, the scanning galvanometer 6, the high-pass dichroic mirror 5, and the electronically controlled flip mirror 4 and finally interferes with the reference light at the beam splitter 2.

[0074] It can be understood that the backscattered light of the sample is the OCT broadband light reflected by the sample, and this light returns to the beam splitter 2 along the original path through the scanning lens 7, the scanning galvanometer 6, the high-pass dichroic mirror 5, and the electronically controlled flip mirror 4, where it interferes with the reference light.

[0075] The backscattered light of the sample is the reflected light (i.e., the backscattered light) of the sample on the electric displacement platform 11, which returns to the beam splitter 2 along the original path through the scanning lens 7, the scanning galvanometer 6, the high-pass dichroic mirror 5, and the electronically controlled flip mirror 4, where it interferes with the reference light.

[0076] Among them, the broadband light emitted by the OCT basic control unit is split by the beam splitter 2 into sample light and reference light. The reference light is reflected by the high reflector 3 and interferes with the light (backscattered light) reflected from the sample surface at the beam splitter 2, and then is transmitted back to the OCT basic control unit through the collimator 1.

[0077] The OCT basic control unit (GAN621, Thorlabs, USA) can both emit OCT broadband light and collect the coherent signals of the reference light and the sample light. The emitted broadband light is a near-infrared broadband light source with a central wavelength of 900 nm and a bandwidth of 150 nm. The fast response spectrometer therein can achieve high-speed acquisition of coherent signals.

[0078] Exemplarily, the alignment wavelength of the fiber collimator 1 is 900 nm, the output light diameter is 2 mm, and the reflection-transmission ratio of the beam splitter 2 is 3:7.

[0079] Exemplarily, as Figure 1 shown, the SERS detection optical path is used to detect the chemical components of the suspicious lesion area of the tissue to determine whether other chemical substances are secreted due to the lesion;

[0080] Specifically, it includes a laser, a collimating lens 8-1, a first beam expander 8-2 and a second beam expander 8-3, a band-pass filter 9, a high-pass dichroic mirror 5, a scanning galvanometer 6, a scanning lens 7, a high-pass filter 10, a collection lens 8-4, and a spectrometer;

[0081] The light emitted from the laser passes through the collimating lens 8-1, the first beam expander 8-2 and the second beam expander 8-3, and then the laser with wavelengths other than 785 nm is filtered by the band-pass filter 9; subsequently, the laser is reflected by the high-pass dichroic mirror 5, and finally reflected by the scanning galvanometer 6 into the scanning lens 7 and focused on the sample surface on the electric displacement platform 11; at this time, a hydrogel film doped with silver nanoparticles is used to cover the sample surface on the electric displacement platform 11; both the SERS signal and the Rayleigh scattering signal of the sample are collected by the scanning lens 7, and then reflected by the scanning galvanometer 6 in sequence, and the Rayleigh light is filtered by the high-pass dichroic mirror 5 and the high-pass filter 10, and finally collected by the collection lens 8-4 into the spectrometer;

[0082] The laser has a central wavelength of 785 nm and is used for the excitation of SERS signals;

[0083] The spectrometer is used to collect SERS signals.

[0084] In order to avoid fluorescence signal interference as much as possible, the laser is a semiconductor continuous laser with a central wavelength of 785 nm and adjustable power, which is used for the excitation of SERS signals.

[0085] The focal length of the collimating lens 8-1 is 40 mm, and the focal lengths of the first beam expander 8-2 and the second beam expander 8-3 are 20 mm and 40 mm respectively.

[0086] The central wavelength of the band-pass filter 9 is 785 nm, and the spectral half-width at half maximum is: 10 nm ± 2. The cut-off wavelength of the high-pass filter 10 is 810 nm. The focal length of the aspherical collection lens 8-4 is 20 mm.

[0087] The spectral range of the spectrometer is: 150 nm - 1500 nm, and the resolution is 1.2 nm. The spectral range for collecting SERS signals is: 400 nm - 1800 nm.

[0088] Among them, the hardware devices that require software control, such as the OCT basic control unit and the spectrometer, are electrically connected to the PC side. In addition, the OCT basic control unit is electrically connected to the scanning galvanometer 6.

[0089] Among them, the collimating lens 8-1, the first beam expander lens 8-2, the second beam expander lens 8-3, and the collection lens 8-4 all adopt the structural form of plano-convex lenses.

[0090] Example 2. The multi-modal gastric cancer tissue diagnosis method provided by the embodiment of the present invention uses an intelligent diagnosis algorithm for gastric cancer tissue, which solves the non-invasive SERS signal enhancement of biological tissues and the accurate intelligent diagnosis of gastric cancer.

[0091] As Figure 4 shown, it specifically includes:

[0092] S1. Establish an optical path of an OCT-SERS combined system with a high registration accuracy sharing a sample arm to achieve high-precision registration of the OCT imaging range and the SERS detection position; the shared sample arm adopts a spatial optical path form to make the OCT imaging optical path and the SERS detection optical path share the galvanometer and the scanning lens 7;

[0093] S2. Non-invasive SERS signal enhancement of gastric cancer tissue: Use a hydrogel film doped with silver nanoparticles to cover the surface of biological tissue and then detect the SERS signal to achieve non-invasive SERS enhancement of the tissue;

[0094] S3. Acquisition and database establishment of OCT images and SERS spectral data of normal and cancerous gastric wall tissues;

[0095] S4. Multi-modal gastric cancer diagnosis of OCT images and SERS spectra of normal and gastric cancer tissues.

[0096] Exemplarily, in step S1, for the SERS detection optical path, the present invention uses a hydrogel film doped with silver nanoparticles as the SERS enhancement substrate. When the film covers the surface of the gastric mucosa, its secretions exchange liquid with the hydrogel film and are adsorbed on the surface of the silver nanoparticles. When light irradiates the surface of the silver particles, the surface plasmon resonance of the particles will generate a strong electromagnetic field enhancement effect. This enhancement effect usually occurs in the "hot spots" of metal nanostructures, such as the gaps between nanoparticles, the gaps between nanoparticles and the substrate, and nanoscale tips. The electromagnetic field intensity in these hot spot regions can be several orders of magnitude higher than that in the surrounding regions, thus greatly enhancing the Raman scattering signal of nearby molecules and overcoming the weak Raman scattering characteristics of biological tissues themselves;

[0097] If the OCT imaging optical path and the SERS detection optical path are combined, it can not only perform micron-level high-resolution tomographic imaging of the lesion area of biological tissue but also have the ability to determine chemical components. The two modalities complement each other and can greatly improve the diagnostic accuracy and robustness of the lesion tissue;

[0098] For the targeted SERS detection of suspicious lesion areas found after OCT imaging, it is necessary to accurately register the OCT imaging position and the spectral acquisition position. Although the partial sharing of the optical path can greatly reduce the position gap between the focused spots of the two modalities, it is far from enough for this high-resolution imaging and spectral detection;

[0099] Exemplarily, for the incident light parts of the OCT imaging optical path and the SERS detection optical path, two-dimensional displacement adjustment frames are used to finely adjust the incident light positions of the two, and the specific registration steps are as follows:

[0100] (1) Registration of the moving directions of the scanning galvanometer 6 and the electric displacement platform 11;

[0101] (1.1) Perform OCT imaging on polystyrene microspheres with a diameter of about 300 microns (other samples are also acceptable). The imaging range is 3 mm × 3 mm. Move the stage to adjust the position of the microspheres within the OCT imaging range so that they appear in the upper, lower, left, and right parts of the scanning range respectively;

[0102] (1.2) At this time, move the stage and compare the horizontal and vertical angular changes of the microspheres in the image before and after movement to calculate the angular difference between the galvanometer scanning direction and the stage movement direction. According to the existing rotational angle difference, adjust the scanning angle of the scanning galvanometer 6 to achieve their registration;

[0103] As Figure 2 shown, the registration effect of the moving directions of the scanning galvanometer 6 and the electric displacement platform 11.

[0104] Exemplarily, (2) The registration of the focus positions of OCT imaging and SERS detection includes:

[0105] Place polystyrene (plastic) microspheres with a diameter of about 300 microns on the stage, perform OCT imaging, adjust them to the center position of the image, and then perform spectral detection;

[0106] Adjust the two-dimensional adjustment frame at the laser incident light until the spectral signal reaches the strongest, and the high-precision registration of the OCT and SERS optical paths can be achieved;

[0107] During the process of testing its registration accuracy, first fix the stage and place four polystyrene (plastic) microspheres. Move the scanning galvanometer 6 to collect spectral data and perform mapping on the collected spectra;

[0108] Control the scanning range of the scanning galvanometer 6 to reconstruct the OCT image of the plastic microspheres, and calculate the registration error through spectral mapping and OCT images;

[0109] The registration effect is as Figure 3 shown, and the registration error is less than 30 microns.

[0110] Exemplarily, in step S2, the innovation of the non-invasive SERS signal enhancement method for gastric cancer tissue lies in: in current biological tissue diagnosis based on SERS technology, the vast majority of them detect pathological information from the secretions of biological tissues, including liquids such as blood, sweat, tissue fluid, and homogenate supernatant. There are few methods that can collect signals without processing the biological tissue. There are a few studies that can directly collect the SERS signal of the tissue, but the tissue needs to be sectioned, and the section thickness is only about ten-odd micrometers. If the tissue is not sectioned and the silver nanoparticle-enhanced substrate is directly dropped onto the surface of the whole tissue. At this time, the silver nanoparticles are very easy to penetrate into the tissue interior, and their enhancement effect is very unstable and it is inevitable to cause toxic harm to the tissue. However, by using the method of the present invention, after the silver nanoparticles are concentrated, they are doped into the hydrogel film. In this way, the silver nanoparticles are fixed in the hydrogel film, which not only will not easily penetrate into the tissue interior, but also can evenly cover the tissue surface to achieve the enhancement of the SERS signal;

[0111] Exemplarily, the synthesis process of silver nanoparticles:

[0112] Drug preparation and solution preparation: Weigh 1.5 g of PEGDA and 0.05 g of HMPP with an electronic balance; weigh 0.0170 g of silver nitrate powder with an electronic balance, and add 90 mL of water to prepare a silver nitrate solution with a concentration of 1.11×10 -3 mol / L; weigh 0.0104 g of hydroxylamine hydrochloride powder with an electronic balance, and add 5 mL of water to prepare a hydroxylamine hydrochloride solution with a concentration of 3×10 -2 mol / L; weigh 0.0120 g of sodium hydroxide powder, and add 5 mL of water to prepare a sodium hydroxide solution with a concentration of 6×10 -2 mol / L; mix the above-prepared hydroxylamine hydrochloride solution and sodium hydroxide solution, shake well and ultrasonicate for 10 min to prepare a 10 mL mixed solution with the concentrations of hydroxylamine hydrochloride and sodium hydroxide being 1.5×10 -2 mol / L and 3×10 -2 mol / L respectively.

[0113] Preparation process: Add the prepared silver nitrate solution into a 150 mL conical flask containing a rotor, place it on a magnetic stirrer, stir at 500 rpm / min for 10 min, then quickly add the above-prepared mixed solution of hydroxylamine hydrochloride and sodium hydroxide into the conical flask. The solution quickly turns black and then quickly turns brownish-gray. Continue to stir and react for 10 min to obtain silver nanocolloid. Extract 100 mL of the silver colloid solution and divide it into 10 centrifuge tubes. After centrifuging at a speed of 4500 revolutions per minute for 10 min, take 1 mL of the lower-layer high-concentration silver nanocolloid from each tube, and a total of 10 mL of concentrated silver nanocolloid is obtained. Place 1.5 g of PEGDA in a centrifuge tube, add 0.05 g of HMPP dropwise, add the concentrated silver nanocolloid to 5 mL, mix and then ultrasonicate for 10 min. Finally, drop the fully mixed solution into a quartz cell of 1.5 mm × 1.5 mm × 0.2 mm and perform a curing treatment using ultraviolet light irradiation.

[0114] Exemplarily, for the detection of SERS signals to achieve non-invasive SERS enhancement of tissues, it includes: placing the tissue on an aluminum sheet, covering the surface with a hydrogel film doped with silver nanoparticles, and performing SERS detection using a 785 nm laser with a laser power of 30 mW and an integration time of 3 s.

[0115] Exemplarily, in step S3, the acquisition and database establishment of OCT images and SERS spectral data of normal and cancerous gastric wall tissues include:

[0116] S301, Cooperate with a municipal hospital in a certain city to collect a large number of gastric cancer tissue specimens. These specimens are carefully cut into two parts (normal and cancerous) by experienced doctors and then recorded in detail. First, apply a trace amount of enhancement substrate on the tissue surface for SERS signal acquisition. Take 1.5 mm × 1.5 mm as the OCT imaging range and perform SERS signal acquisition within this range, with an interval of 300 μm between each sampling point;

[0117] S302, Perform OCT imaging and SERS signal acquisition within the set range of 1.5 mm × 1.5 mm;

[0118] S303, Display the OCT images and SERS spectral information of normal and cancerous gastric tissues, such as Figure 5 The OCT B-scan images and pathological section effect diagrams of normal and cancerous gastric wall tissues;

[0119] Among them, the OCT of normal gastric wall tissue shows a clear gastric glandular tube structure ( Figure 5 Figure A in Figure 5 ), which is consistent with the pathological structure shown in Figure 5 Figure B. The gastric tubes are arranged regularly, forming dense gastric pits on the mucosal surface (

[0120] Compared with normal gastric wall tissue, gastric cancer shows obvious differences in tissue morphology and structure, as shown in Figures D and F of Figure 5 . The absence or rare formation of glandular structures in cancer cells results in an obviously homogeneous appearance. Figure 5 The abnormal cell behavior observed in the E pathological diagram is consistent with the OCT image. Essentially, OCT technology can accurately depict the morphological and structural changes of cancer tissue.

[0121] The glandular ducts of normal gastric wall tissue generate a dense and uniform gastric pit structure on the surface of the gastric mucosa. In the OCT tomography of cancerous gastric wall tissue, it is shown as a dense and uniform state without morphological structure.

[0122] As shown in Figure 5 the OCT tomography images and pathological section effect diagrams of normal and cancerous gastric wall tissues, where Figure 5 Figures A and C are the OCT tomography images of normal and cancerous gastric wall tissues, Figure 5 and Figures B and E are the pathological section diagrams of normal and cancerous gastric wall tissues.

[0123] The SERS spectrum of gastric wall tissue has a greatly enhanced Raman spectral intensity after SERS enhancement compared to the Raman spectra c and d of the tissue itself. As shown in Figure 6 the SERS spectrum diagram of gastric wall tissue.

[0124] Exemplarily, in step S4, in the multimodal gastric cancer diagnosis of OCT images and SERS spectra of normal and gastric cancer tissues, CNN can effectively extract local features in the image through convolution operations and has good effects on tasks such as image recognition and object detection. As shown in Figure 7 Schematic diagram of a multimodal neural network. For OCT images and SERS spectral data, a convolutional neural network structure (multimodal neural network) of one-dimensional and two-dimensional ResNet is respectively adopted, and their outputs, that is, the final fully connected layers, are used as the input of the Transformer. The Transformer can capture the dependencies between positions in the input sequence through the self-attention mechanism and can better model long-range dependencies.

[0125] Among them, the feature of the multimodal gastric cancer diagnosis of OCT images and SERS spectra of normal and gastric cancer tissues is that the fusion of two-dimensional image data and one-dimensional spectral data is realized by an integrated learning method, improving the recognition accuracy and robustness of gastric cancer tissue.

[0126] Exemplarily, ResNet50 is a deep convolutional neural network that solves the problem of gradient disappearance in deep networks through residual blocks (Residual Blocks) and efficiently extracts features.

[0127] The multimodal gastric cancer diagnosis method for OCT images and SERS spectra of normal and gastric cancer tissues includes:

[0128] S401. Extract the features of two-dimensional image data using ResNet50;

[0129] S4011. Input data: The two-dimensional data is usually a grayscale image of 200×200×1, and the number of channels is 1. In ResNet50, the input data first passes through the convolutional layer and pooling layer, and then extracts features through multiple residual blocks.

[0130] S4012. Convolution operation: The convolution operation is the core of feature extraction in ResNet50. Through convolution, the model can learn local patterns (such as edges, textures, etc.) from the image. The convolutional kernel slides on the image and extracts important features of each local area to generate a feature map.

[0131] S4013. The first layer of convolution and pooling, including:

[0132] 7×7 convolution: The first layer of ResNet50 is a convolutional kernel of size 7×7, using 64 convolutional kernels, and the stride is 2. Therefore, the input image size changes from 200×200×1 to 100×100×64 after convolution, and each channel corresponds to 64 different features.

[0133] Max Pooling: Use 3×3 max pooling with a stride of 2 to reduce the feature map size to 50×50×64, thereby reducing the computational amount and retaining important features.

[0134] S4014. Residual Block:

[0135] In ResNet50, the core part is the residual block. Each residual block consists of multiple convolutional layers and has a skip connection to ensure that the gradient can be directly transmitted. The structure of a residual block includes:

[0136] Convolutional layer Each residual block usually includes two or three convolutional layers, with a convolutional kernel size of 3×3 and a stride of 1 or 2. These convolutional layers are used to extract features and change the number of channels.

[0137] Skip connection: The input directly skips multiple convolutional layers and connects to the subsequent layers to prevent the problem of gradient disappearance.

[0138] The purpose of the residual block is to learn the residual of the convolutional layer, that is, the convolutional layer learns the difference between the input features and the output features. The skip connection enables the model to directly optimize this residual part.

[0139] S4015. Changes in the size of the feature map, including:

[0140] Stage2: The input size is 50×50×64. After convolution and pooling operations, the number of channels increases to 256, and the size becomes 50×50×256.

[0141] Stage3: The stride is 2. The input size is 50×50×256 and becomes 25×25×512.

[0142] Stage4: The stride is 2. The input size is 25×25×512 and becomes 13×13×1024.

[0143] Stage5: The stride is 2. The input size is 13×13×1024 and becomes 7×7×2048.

[0144] S4016. Global average pooling:

[0145] After the features are extracted in the convolutional layer, ResNet50 uses global average pooling to average the values in each feature map, obtaining an output of 1×1×2048. This step reduces the spatial dimension (7×7) to 1 while retaining the important features in the channels.

[0146] S4017. Fully connected layer: Finally, the feature map is flattened into a one-dimensional vector of length 2048.

[0147] S402. Extract features of one-dimensional spectral data using ResNet50;

[0148] The processing of one-dimensional data (such as time series, signals, etc.) is similar to two-dimensional data, except that the convolutional kernel slides in one-dimensional space instead of two-dimensional space. The operation of convolution can be extended from a two-dimensional plane to a one-dimensional sequence. Specifically, it includes:

[0149] S4021. Input data: The input data is a one-dimensional vector with a size of 1×748.

[0150] S4022. Convolution operation: In one-dimensional convolution, the convolutional kernel slides along the time steps or feature dimensions of the sequence. For one-dimensional data, the size of the convolutional kernel becomes 7. The input data will be processed through a one-dimensional convolutional layer and a pooling layer.

[0151] S4023. The first layer of convolution and pooling, including:

[0152] Convolutional layer: Use a convolutional kernel of size 7, a stride of 2, and 64 filters. The length of the input data changes from 748 to (748 - 7) / 2 + 1 = 371, and the number of channels becomes 64. The output size is 1×371×64.

[0153] Max Pooling: Use max pooling with a size of 3 and a stride of 2, and the output size becomes 1×185×64.

[0154] S4024. Residual Block:

[0155] The structure of the one-dimensional residual block is similar to that of the two-dimensional residual block. Each residual block includes multiple one-dimensional convolutional layers and skip connections. The residual block extracts features in the time series through one-dimensional convolution and increases the number of channels.

[0156] S4025. Changes in the size of the feature map, including:

[0157] Stage2: The input is 1×185×64. After convolution and pooling, the output becomes 1×185×256.

[0158] Stage3: The stride is 2, and the output size becomes 1×93×512.

[0159] Stage4: The stride is 2, and the output size becomes 1×47×1024.

[0160] Stage5: The stride is 2, and the output size becomes 1×24×2048.

[0161] S4026. Global Average Pooling: Similar to the two-dimensional convolutional network, after passing through the residual block, use the global average pooling operation to average the features of each channel, and the output becomes 1×1×2048.

[0162] S4027. Fully Connected Layer: Finally, flatten the feature map into a one-dimensional vector with a length of 2048.

[0163] S403. Concatenate the two 1×2048 one-dimensional vectors obtained from the fully connected layer of the two-dimensional ResNet50 and the one-dimensional ResNet50, and then send them into the Transformer network for classification and recognition, finally realizing the fusion and diagnosis of multi-modal data.

[0164] The above is only a relatively optimal specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A multimodal gastric cancer tissue diagnosis method using OCT-SERS, characterized in that: The method comprises the following steps: S1, establishing an optical path of an OCT-SERS combined system with a shared sample arm to achieve registration of the OCT imaging range and the SERS detection position, wherein the registration error is less than 30 microns; the shared sample arm adopts a spatial optical path so that the OCT imaging optical path and the SERS detection optical path share a galvanometer (6) and a scanning lens (7); S2, using a hydrogel film doped with silver nanoparticles to cover the surface of biological tissues to detect SERS signals, achieving non-destructive SERS enhancement of gastric cancer tissues; S3, acquisition of OCT images and SERS spectral data of normal and cancerous gastric wall tissues and establishment of database; S4, Multimodal gastric cancer diagnosis based on OCT images and SERS spectra of normal and gastric cancer tissues.

2. The multimodal gastric cancer tissue diagnosis method using OCT-SERS according to claim 1, characterized in that: In step S1, the OCT imaging range and the SERS detection position are aligned, including: the alignment of the moving directions of the scanning galvanometer (6) and the electric displacement platform (11) and the alignment of the OCT imaging and the SERS detection focus position; The alignment of the moving directions of the scanning galvanometer (6) and the electric displacement platform (11) includes: placing polystyrene microspheres on a stage and setting an OCT imaging range; moving the stage and adjusting the position of the polystyrene microspheres in the OCT imaging range so that the imaging range appears in the upper, lower, left and right parts of the scanning range respectively, and using a program to calculate the coordinates of the center of the circle; comparing the horizontal and vertical angle changes of the center of the microsphere in the image before and after the movement, calculating the angle difference between the scanning direction of the galvanometer and the moving direction of the stage, and adjusting the scanning angle of the scanning galvanometer (6) according to the existing angle difference to achieve the alignment of the two; The OCT imaging and SERS detection focus position alignment includes: placing polystyrene microspheres on a stage and performing OCT imaging, adjusting the OCT image of the microspheres to the center of the scanning range, and performing spectral detection; adjusting the two-dimensional adjustment frame at the laser incident light to the maximum spectral signal to achieve the alignment of the OCT and SERS optical paths; In the process of testing the registration accuracy, the stage is first fixed and four polystyrene microspheres are placed, the scanning galvanometer (6) is moved to collect spectral data, and the collected spectrum is mapped, and the center coordinates of the circle are calculated using a program; the scanning range of the scanning galvanometer (6) is controlled to perform OCT imaging on the plastic microspheres, and the center coordinates of the circle are calculated using a program, and the registration error is calculated through the difference between the spectral mapping and the center coordinates in the OCT image.

3. The multimodal gastric cancer tissue diagnosis method using OCT-SERS according to claim 1, characterized in that: In step S2, the preparation of the hydrogel film doped with silver nanoparticles includes: Drug preparation and solution preparation: the crosslinking agent for preparing the hydrogel is polyethylene glycol diacrylate; the photoinitiator is dihydroxydimethyl acetone; weigh silver nitrate powder, add water to prepare a silver nitrate solution; weigh hydroxylamine hydrochloride powder, add water to prepare a hydroxylamine hydrochloride solution; weigh sodium hydroxide powder, prepare a sodium hydroxide solution; mix the above-prepared hydroxylamine hydrochloride solution and sodium hydroxide solution, shake well and then sonicate to prepare a mixed solution of hydroxylamine hydrochloride and sodium hydroxide; Preparation process: Stir the prepared silver nitrate solution, add the prepared mixed solution of hydroxylamine hydrochloride and sodium hydroxide, stir and react to obtain nano silver colloid; extract the nano silver colloid solution, centrifuge and take out the lower layer of high concentration nano silver colloid; mix polyethylene glycol diacrylate with the nano silver colloid, add dihydroxydimethylacetone and then perform ultrasonic treatment to fully mix them; place the uniformly mixed hydrogel precursor solution in a mold and perform ultraviolet light irradiation for curing; In step S2, SERS signal detection is performed, including: placing the tissue on an aluminum sheet, the surface of which is covered with a hydrogel film doped with silver nanoparticles, and performing SERS detection using a 785nm laser with a laser power of 30mW and an integration time of 3s.

4. The multimodal gastric cancer tissue diagnosis method using OCT-SERS according to claim 1, characterized in that: In step S3, the acquisition of OCT images and SERS spectrum data of normal and cancerous gastric wall tissues and the establishment of a database include: S301, collecting gastric cancer tissue specimens, covering the surface of the tissue with a hydrogel film doped with silver nanoparticles for SERS signal collection, using 1.5 mm × 1.5 mm as the OCT imaging range, and collecting SERS signals within the range, with the interval between each sampling point being 300 μm; S302, collecting OCT images and SERS spectrum data within a range of 1.5 mm×1.5 mm; S303, OCT images and SERS spectral information display of normal and cancerous gastric tissues.

5. The multimodal gastric cancer tissue diagnosis method using OCT-SERS according to claim 1, characterized in that: In step S4, multimodal gastric cancer diagnosis based on OCT images and SERS spectra of normal and gastric cancer tissues includes: S401, extracting features of two-dimensional image data using ResNet50; S402, extracting features of one-dimensional spectral data using ResNet50; S403, concatenate the two 1×2048 one-dimensional vectors obtained by the two-dimensional ResNet50 and the one-dimensional ResNet50 in the fully connected layer and send them to the Transformer network for classification and recognition to achieve fusion and diagnosis of multimodal data.

6. The multimodal gastric cancer tissue diagnosis method using OCT-SERS according to claim 5, characterized in that: In step S401, ResNet50 is used to extract features of the two-dimensional image data, including: S4011, input data; S4012, convolution operation: the convolution kernel slides on the image and extracts the important features of each local area to generate a feature map; S4013, first layer convolution and pooling; S4014, using the residual block to learn the residual of the convolutional layer; S4015, change of feature map size; S4016, use global average pooling to average the values ​​in each feature map to obtain a 1×1×2048 output; S4017, using the fully connected layer feature map is flattened into a one-dimensional vector of length 2048; In step S402, ResNet50 is used to extract features of one-dimensional spectral data, including: sliding the convolution kernel in one-dimensional space, extending the convolution operation from the two-dimensional plane to the one-dimensional sequence, and flattening the feature map into a one-dimensional vector with a length of 2048.

7. An OCT-SERS combined multimodal gastric cancer tissue diagnosis system, characterized in that: The system implements the multimodal gastric cancer tissue diagnosis method combined with OCT-SERS as claimed in any one of claims 1 to 6. The system realizes the partial spatial light path sharing of the two modalities by independently designing and constructing the light path, and realizes the alignment of the imaging and spectral acquisition positions, specifically including: The optical path of the OCT-SERS combined system with high registration accuracy of the shared sample arm is used to achieve high-precision registration of the OCT imaging range and the SERS detection position; the shared sample arm adopts the form of a spatial optical path so that the OCT imaging optical path and the SERS detection optical path share a galvanometer (6) and a scanning lens (7); The non-destructive SERS signal enhancement module for gastric cancer tissue is used to detect SERS signals after covering the surface of biological tissue with a hydrogel film doped with silver nanoparticles, thereby achieving non-destructive SERS enhancement of tissue; Database establishment module, used for acquisition and database establishment of OCT images and SERS spectrum data of normal and cancerous gastric wall tissues; The multimodal diagnosis module is used for multimodal gastric cancer diagnosis based on OCT images and SERS spectra of normal and gastric cancer tissues.

8. The OCT-SERS combined multimodal gastric cancer tissue diagnosis system according to claim 7, characterized in that: The OCT imaging optical path is used to perform non-destructive, high-resolution tomographic imaging of biological tissues, and to obtain depth information of biological tissues by utilizing the coherence of reference light and sample light; The OCT imaging optical path comprises an OCT basic control unit, a fiber optic collimator (1), a beam splitter (2), a high-reflection mirror (3), an electrically controlled flip mirror (4), a high-pass dichroic mirror (5), a scanning galvanometer (6), a scanning lens (7), and an electric displacement platform (11); The broadband light source emitted by the OCT basic control unit is collimated by a fiber collimator (1) and then split into a sample light beam and a reference light beam by a beam splitter (2); After being reflected by the electrically controlled flip mirror (4), the sample light enters the scanning lens (7) through the high-pass dichroic mirror (5) and the scanning galvanometer (6), and is finally focused on the sample surface; The reference light is reflected by the high-reflection mirror (3) back to the beam splitter (2), and is transmitted back to the OCT basic control unit after interfering with the backscattered light of the sample; the backscattered light of the sample is the scattered light of the sample on the electric displacement platform (11), which passes through the scanning lens (7), the scanning galvanometer (6), the high-pass dichroic mirror (5), and the electrically controlled flip mirror (4), and finally interferes with the reference light at the beam splitter (2).

9. The OCT-SERS combined multimodal gastric cancer tissue diagnosis system according to claim 7, characterized in that: The alignment wavelength of the optical fiber collimator (1) is 900 nm, the light output diameter is 2 mm, and the reflection and transmission ratio of the beam splitter (2) is 3:

7.

10. The OCT-SERS combined multimodal gastric cancer tissue diagnosis system according to claim 7, characterized in that: The SERS detection optical path is used to detect the chemical composition of the lesion area of ​​the tissue to determine whether other chemical substances are secreted due to the lesion; Specifically comprising: a laser, a collimating lens (8-1), a first beam expanding lens (8-2) and a second beam expanding lens (8-3), a band pass filter (9), a high pass dichroic mirror (5), a scanning galvanometer (6), a scanning lens (7), a high pass filter (10), a collecting lens (8-4) and a spectrometer; The output light of the laser passes through a collimating lens (8-1), a first beam expanding lens (8-2) and a second beam expanding lens (8-3), and then passes through a bandpass filter (9) to filter out laser light with wavelengths other than 785 nm; the laser light is reflected by a high-pass dichroic mirror (5), and finally passes through a scanning galvanometer (6) to be reflected into a scanning lens (7), and is focused on the surface of a sample; a hydrogel film doped with silver nanoparticles is covered on the surface of a biological tissue; a SERS signal and a Rayleigh scattering signal of the sample are both collected by a scanning lens (7), and then successively reflected by a scanning galvanometer (6), a high-pass dichroic mirror (5) and a high-pass filter (10) to filter out the Rayleigh light, and finally collected into a spectrum by a collecting lens (8-4); The laser has a center length of 785 nm and is used for the excitation of SERS signals; The spectrometer is used to collect SERS signals.

Images