Biomarker detection system and method

Abstract

The invention provides a biomarker detection system and method, the biomarker detection system comprises: a detection chip, the detection chip contains a detection substance, and after the detection substance and a target marker in a sample to be detected are subjected to physical adsorption combination, a detection area is formed in the detection chip; the image acquisition device is used for acquiring a detection image containing the detection chip, and the image brightness corresponding to a detection area in the detection image is different from the image brightness corresponding to an area outside the detection area; the image processing device is used for calling a first model, and the first model can determine the concentration of the target marker in the sample to be detected based on the brightness of the detection image.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a biomarker detection system and method. Background Technology

[0002] Traumatic brain injury (TBI) is a leading cause of disability and death worldwide, particularly prevalent in high-risk scenarios such as military trauma, traffic accidents, and falls, resulting in extremely high morbidity and a significant medical burden. Some studies have shown that glial cell-derived biomarkers, such as S100B protein, rapidly increase in serum and cerebrospinal fluid after TBI, serving as important indicators for assessing the extent of brain tissue damage and prognosis. Therefore, the development of rapid, sensitive, and automatically identifiable biomarkers is an urgent need in neurosurgery, emergency medicine, and military medicine. Summary of the Invention

[0003] One of the technical problems this application aims to solve is: how to quickly, sensitively and automatically identify biomarkers.

[0004] To address the aforementioned technical problems, embodiments of this application provide a biomarker detection system and method.

[0005] This application provides a biomarker detection system, comprising: a detection chip containing a detection substance, wherein the detection substance physically adsorbs and binds to various substances in a target biomarker in a test sample, forming corresponding detection areas in the detection chip; the target biomarker includes at least one of the following substances: S100B protein and glial fibrillary acidic protein; the detection substance is selenium nanoparticles; an image acquisition device for acquiring a detection image containing the detection chip, wherein the image brightness corresponding to the detection area in the detection image is different from the image brightness corresponding to the area outside the detection area; the brightness of the detection areas formed by the detection substance and different substances in the target biomarker in the detection image is different; and an image processing device for calling a first model, which can determine the concentration of each substance in the target biomarker in the test sample based on the brightness of the detection image; the concentration of the target biomarker is related to the severity of traumatic brain injury; the image processing device is also used to determine the severity of traumatic brain injury based on the concentration of each substance in the target biomarker in the test sample.

[0006] In some embodiments, the detection chip includes: a substrate; a first material layer, wherein a first detection array in the first material layer is arranged on the substrate; an antibody-modified layer, wherein the antibody-modified layer is disposed on the first detection material; a second material layer, wherein a second detection material in the second material layer forms a heterogeneous chain with the first detection material through the antibody-modified layer; the second detection material is capable of physically adsorbing and binding with a target marker in the sample to be tested.

[0007] In some embodiments, the first model includes: a convolutional layer for processing a detection image and extracting a first image feature from the detection image; a residual layer for extracting a second image feature from the detection image based on the first image feature, wherein the first image feature and the second image feature are different; at least one of the first image feature and the second image feature can reflect the brightness of the detection image; and a fully connected layer for determining the concentration of a target marker in the sample to be tested based on the first image feature and the second image feature.

[0008] In some embodiments, the target marker includes a first sub-marker and a second sub-marker; the second detector in the second material layer can physically adsorb and bind with the first sub-marker to form a first scattering signal; the second detector in the second material layer can physically adsorb and bind with the second sub-marker to form a second scattering signal; the first scattering signal and the second scattering signal have different degrees of influence on the brightness of the detected image.

[0009] In some embodiments, the image acquisition device includes a controller, a first acquisition channel, and a second acquisition channel, and the detected image includes a first sub-image and a second sub-image; the controller is configured to control the first acquisition channel to acquire the first sub-image containing the detection chip based on a first scattering signal in the detection chip; and to control the second acquisition channel to acquire the second sub-image containing the detection chip based on a second scattering signal in the detection chip.

[0010] In some embodiments, the image processing apparatus is further configured to invoke a second model, which is capable of simultaneously obtaining the concentration of a first sub-marker based on a first sub-image and the concentration of a second sub-marker based on a second sub-image.

[0011] In some embodiments, the image processing apparatus is also used to predict the development trend of trauma based on the concentration of the target marker in the sample to be tested and the severity of traumatic brain injury.

[0012] This application also provides a method for detecting biomarkers, comprising: acquiring a detection image containing a detection chip; determining the concentration of a target biomarker in a sample to be tested based on the brightness of the detection image using a first model; the concentration of the target biomarker is related to the severity of traumatic brain injury; the target biomarker includes at least one of the following substances: S100B protein and glial fibrillary acidic protein; determining the severity of traumatic brain injury based on the concentration of each of the target biomarkers in the sample to be tested; wherein the target biomarker in the sample to be tested can physically adsorb and bind with the detection substance in the detection chip, forming a detection area in the detection chip; the detection substance is selenium nanoparticles; the image brightness corresponding to the detection area in the detection image is different from the image brightness corresponding to the area outside the detection area; the brightness of each detection area formed by the detection substance and different substances in the target biomarker in the detection image is different.

[0013] In some embodiments, determining the concentration of a target biomarker in a sample to be tested based on the brightness of a detected image using a first model includes: processing the detected image using the first model to extract a first image feature from the detected image; extracting a second image feature from the detected image based on the first image feature using the first model, wherein the first image feature and the second image feature are different; at least one of the first image feature and the second image feature can reflect the brightness of the detected image; and determining the concentration of the target biomarker in the sample to be tested based on the first image feature and the second image feature using the first model.

[0014] This application provides a computer device, including a processor and a memory storing a computer program. When the processor executes the program, it implements the steps of the biomarker detection method provided in the above embodiments.

[0015] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of any of the above-described biomarker detection methods.

[0016] Through the above technical solution, the biomarker detection system and method provided in this application achieve precise control of the optical coupling effect by physically adsorbing and binding the detection substance in the detection chip with the target biomarker, thereby forming a significantly enhanced signal and enabling the formation of recognizable image features in subsequent detection images. The detection image containing the detection chip is acquired by an image acquisition device, and the image is then identified and analyzed by an image processing device to obtain the concentration of the target biomarker. This significantly reduces human interpretation bias and improves diagnostic consistency. Furthermore, the detection method is simple, requiring no labeled or pre-treated nanoprobes, which greatly improves detection speed and has broad application prospects. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the structure of the biomarker detection system disclosed in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of the detection chip disclosed in the embodiments of this application; Figure 3 This is a schematic diagram of the process for preparing the detection chip disclosed in the embodiments of this application; Figure 4 This is a schematic diagram of the detection image disclosed in the embodiments of this application; Figure 5 This is a schematic diagram of the structure of the first model disclosed in the embodiments of this application; Figure 6 This is a schematic diagram illustrating the result verification disclosed in the embodiments of this application; Figure 7 This is a schematic diagram of scatter plots based on dual markers disclosed in an embodiment of this application; Figure 8 This is a schematic flowchart of the biomarker detection method disclosed in the embodiments of this application; Figure 9 This is a schematic diagram of the structure of a computer device disclosed in an embodiment of this application. Detailed Implementation

[0019] The embodiments of this application will be further described in detail below with reference to the accompanying drawings and examples. The detailed description of the following embodiments and the accompanying drawings are used to illustrate the principles of this application by way of example, but should not be used to limit the scope of this application. This application can be implemented in many different forms and is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

[0020] These embodiments are provided to make the application thorough and complete, and to fully express the scope of the application to those skilled in the art. It should be noted that, unless otherwise specifically stated, the relative arrangement of components and steps, material composition, numerical expressions, and values ​​illustrated in these embodiments should be interpreted as merely exemplary and not as limiting.

[0021] Furthermore, the terms "including" or "comprising" as used in this application mean that the element preceding the word covers the element listed after the word, and do not exclude the possibility that it may also cover other elements.

[0022] It should also be noted that, in the description of this application, unless otherwise expressly specified and limited, the terms "installation," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this application depending on the specific circumstances. When a specific device is described as being located between a first device and a second device, an intermediary device may or may not be present between the specific device and the first or second device.

[0023] All terms used in this application have the same meaning as understood by one of ordinary skill in the art to which this application pertains, unless otherwise specifically defined. It should also be understood that terms defined in general dictionaries should be interpreted as having meanings consistent with their meanings in the context of the relevant art, and not as idealized or highly formalized, unless expressly defined herein.

[0024] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, they should be considered part of the specification.

[0025] Traumatic brain injury (TBI) refers to brain tissue damage caused by severe trauma. It is one of the leading causes of disability and death worldwide, especially in high-incidence scenarios such as military trauma, traffic accidents, and falls, where it has an extremely high incidence and medical burden.

[0026] Numerous studies have shown that S100B protein is primarily secreted by glial cells. In cases of traumatic brain injury, subarachnoid hemorrhage, or stroke, glial cells release S100B into the cerebrospinal fluid, which then crosses the blood-brain barrier into the bloodstream. This leads to a rapid increase in glial cell-derived biomarkers such as S100B protein in serum, cerebrospinal fluid, and other bodily fluids after TBI, serving as important indicators for assessing the degree of brain tissue damage and prognosis. Therefore, the development of rapid, sensitive, and automated biomarker detection systems is an urgent need in neurosurgery, emergency medicine, and military medicine.

[0027] Currently, the detection of TBI biomarkers such as S100B mainly relies on traditional methods, such as: (1) Enzyme-linked immunosorbent assay (ELISA): ELISA is one of the standard methods for detecting brain injury markers such as S100B and GFAP (glial fibrillary acidic protein). It is a mature technology with high sensitivity and specificity, but it has the following obvious shortcomings: the experimental procedure is cumbersome and time-consuming (generally 3-6 hours), and it is highly dependent on experimental operation and environment, making it difficult to meet the clinical demand for rapid detection; there are many operation steps, which are easily affected by human error; there is a certain sensitivity bottleneck for the detection of low concentration (pg / mL level) biomarkers; it is difficult to achieve the joint detection of multiple indicators, and it cannot support high-throughput multi-target identification.

[0028] (2) Chemiluminescence and electrochemical sensors: can achieve relatively rapid detection, but the equipment is expensive and the scope of application is limited; the sensitivity decreases in various body fluids (such as urine and saliva) and is easily affected by matrix interference.

[0029] (3) Commercial portable sensors, such as point-of-care testing (POCT) devices: Although convenient, their sensitivity is difficult to meet the ultra-early detection requirements at the pg / mL level; most products still rely on single antigen-antibody pairing and have not achieved intelligent system identification or concentration level judgment.

[0030] (4) Artificial intelligence-assisted analysis system: Artificial intelligence can only be used more in the field of medical image-assisted diagnosis. It is still in the initial stage in the quantitative analysis of nanosensor images. There is a lack of calculation methods that are highly matched with sensor signals, the model training samples are limited, and the versatility is poor.

[0031] In summary, the relevant technologies generally suffer from the following drawbacks: slow detection speed, making it difficult to achieve rapid point-of-care detection; limited sensitivity, making it impossible to accurately identify low-concentration target substances; low level of intelligence, making it impossible to achieve automated and batch image analysis; insufficient parallel detection of multiple markers and dynamic monitoring capabilities; and related data analysis methods are unable to effectively mine complex feature signals in images.

[0032] In view of this, embodiments of this application provide a biomarker detection system, such as... Figure 1 As shown, the detection system includes: a detection chip 110, an image acquisition device 120, and an image processing device 130.

[0033] The detection chip 110 contains detection substances. These substances physically adsorb and bind to various components of the target biomarker 116 in the sample, forming corresponding detection areas within the chip 110. The target biomarker includes at least one of the following: S100B protein and glial fibrillary acidic protein; the detection substance is selenium nanoparticles. For example, the detection chip 110 contains selenium nanoparticles (SeNPs), which can physically adsorb and bind to the target biomarker S100B protein in the sample, thereby forming an immune sandwich structure within the chip 110, inducing signal enhancement, and creating recognizable image features, i.e., forming a detection area corresponding to S100B.

[0034] The image acquisition device 120 is used to acquire detection images containing the detection chip 110. Because the target marker 116 forms an immune sandwich structure with the detection substance in the detection chip 110, signal enhancement is induced. Therefore, in the detection image, the image brightness corresponding to the detection area is different from the image brightness corresponding to the area outside the detection area. Similarly, if the target detector includes multiple substances, such as simultaneously containing S100B protein and glial fibrillary acidic protein, then the brightness of the various detection areas formed by the detection substance and different substances in the target marker in the detection image will also be different.

[0035] The image processing device 130 is used to invoke the first model 131, which can determine the concentration of each substance in the target marker 116 in the sample to be tested based on the brightness of the detection image; the concentration of the target marker is related to the severity of traumatic brain injury. For example, the first model 131 is a convolutional neural network (CNN) model, which can identify immune sandwich structures in the detection image, such as identifying the brightness of each region in the detection image, determining whether there is an immune sandwich structure in the detection chip 110 based on the brightness of each region, and obtaining the concentration of S100B in the target marker 116 by the presence of the immune sandwich structure.

[0036] The image processing device is also used to determine the severity of traumatic brain injury based on the concentration of each substance in the target marker in the sample.

[0037] Based on the concentration of target biomarker 116, image processing device 130 can determine not only the concentration of each substance in the target biomarker, but also the severity of traumatic brain injury (e.g., mild, moderate, severe). For example, based on the combined concentration of S100B and GFAP, the model may classify traumatic brain injury into different types: mild trauma (low S100B, high GFAP); moderate trauma (moderate S100B and GFAP concentrations); and severe trauma (high S100B, very high GFAP).

[0038] In this embodiment, the detection substance in the detection chip 110 is physically adsorbed and bound to the target marker 116 to achieve precise control of the optical coupling effect, forming a significantly enhanced signal, thereby enabling the formation of recognizable image features in subsequent detection images. The detection image containing the detection chip 110 is acquired by the image acquisition device 120, and the detection image is identified and analyzed by the image processing device 130 to obtain the concentration of each substance in the target marker 116 and determine the severity of traumatic brain injury. This significantly reduces human interpretation bias and improves diagnostic consistency. Moreover, the detection method is simple, requiring no labeling or pre-processing of the nanoprobes, which greatly improves detection speed. Clinicians can make more accurate decisions based on the severity of traumatic brain injury. For example, if the traumatic brain injury is severe, doctors may choose more aggressive interventions, such as drug treatment or surgery, demonstrating broad application prospects.

[0039] This application provides a biomarker detection system, which includes a detection chip 110, an image acquisition device 120, and an image processing device 130.

[0040] The detection chip 110 contains detection substances. These substances physically adsorb and bind to various components of the target biomarker 116 in the sample, forming corresponding detection areas within the chip 110. The target biomarker includes at least one of the following: S100B protein and glial fibrillary acidic protein; the detection substance is selenium nanoparticles. For example, the detection chip 110 contains selenium nanoparticles (SeNPs), which can physically adsorb and bind to the target biomarker S100B protein in the sample, thereby forming an immune sandwich structure within the chip 110, inducing signal enhancement, and creating recognizable image features, i.e., forming a detection area corresponding to S100B.

[0041] In one specific embodiment, such as Figure 2 As shown, the detection chip 110 includes: Substrate 112. Exemplarily, the substrate 112 may be a silicon substrate 112. Alternatively, the substrate 112 may also be made of other semiconductor materials. Optionally or additionally, the substrate 112 may include other elemental semiconductor materials such as germanium.

[0042] A first material layer 113 is provided, in which a first detector array is arranged on a substrate 112. For example, the first material layer 113 includes a plurality of regularly arranged polystyrene (PS) nanospheres, the diameter of which may be 500 nm and the spacing between each polystyrene (PS) nanosphere is 5 μm.

[0043] Antibody modification layer 114 is disposed on the first detection substance.

[0044] The second material layer 115 contains a second detectable that forms a heterogeneous chain with the first detectable via an antibody-modified layer 114. The second detectable can physically adsorb and bind to the target marker 116 in the sample. For example, the second material layer 115 includes multiple selenium nanoparticles (SeNPs), which, together with PS nanospheres, form a heterogeneous chain structure via the antibody-modified layer 114. SeNPs possess unique optical properties and can significantly enhance optical signals through the local surface plasmon resonance (LSPR) effect.

[0045] In this embodiment, the absorption and scattering signals of light are amplified when a localized surface plasmon resonance effect occurs on the nanoparticle surface. The heterogeneous chain structure formed by the combination of SeNPs and PS nanospheres enhances this effect, significantly increasing the intensity of the detection signal. This allows the system to maintain high sensitivity in the detection of low-concentration markers (such as S100B protein).

[0046] After the target biomarker is bound to selenium nanoparticles, a label-free reaction occurs between the chip surface and the target substance, and the signal is amplified through the aforementioned optical enhancement effect. This process avoids the complex labeling steps of traditional methods, making the chip easier to use and more efficient.

[0047] In a specific embodiment, such as Figure 3 As shown, the process of preparing the detection chip 110 includes: S11: Substrate 112 cleaning. For example, cutting the silicon wafer to 10×10mm. 2 Afterward, a thorough cleaning is performed to remove surface dirt and impurities. The cleaning process includes ultrasonic cleaning and nitrogen drying.

[0048] S12: Hydrophilization treatment of substrate 112. To improve the adhesion of nanoparticles to the silicon substrate 112, a low-temperature plasma hydrophilization treatment is performed on the silicon wafer surface. This treatment significantly improves the hydrophilicity of the silicon wafer surface, allowing the nanoparticles to be uniformly distributed and providing better surface adsorption conditions for subsequent self-assembly. The hydrophilization process is typically carried out in a low-temperature plasma reactor, where the low-temperature plasma treatment lasts for 8 seconds, reducing the water contact angle to approximately 50–60°.

[0049] S13: A heterogeneous chain is formed on substrate 112.

[0050] Two types of nanoparticles were used in the embodiments of this application: Polystyrene (PS) nanospheres (500 nm in diameter): used as the main material for substrate 112.

[0051] Selenium nanoparticles (SeNPs, 250 nm in diameter): The part used to form the signal enhancement.

[0052] On the surface of a hydrophilically treated silicon wafer, nanoparticles are arranged into a regular array using a liquid-confined self-assembly process. Specifically, a solution of PS nanoparticles of a certain concentration is dropped onto the silicon wafer surface, and template-confined imprinting technology is used to arrange the nanoparticles at a predetermined spacing (e.g., 5 μm). The template is typically made of a transparent material, which can restrict the alignment of the nanoparticles and ensure their uniform distribution. Subsequently, the silicon wafer is heated to 56°C and incubated for 1 hour, allowing the PS nanoparticles to self-assemble into a regular chain-like structure.

[0053] S14: Thermal Curing and Activation. After the self-assembly process, the nanoparticles need to be stably fixed on the silicon wafer surface through thermal curing. The thermal curing temperature is typically 115°C, lasting approximately 35 minutes. This process ensures the stability of the nanoparticles on the surface and prevents them from detaching due to subsequent operations. S15: Antibody Fixation and Blocking: Antibody solution is added and incubated for 1 hour, followed by blocking with BSA (bovine serum albumin) for 30 minutes. The thermally cured chip surface is then activated using a chemical activating agent (EDC (1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride) or NHS (N-hydroxysuccinimide)). This treatment converts the carboxyl groups on the silicon wafer surface into active esters, preparing them for subsequent covalent binding reactions with antibodies.

[0054] In this embodiment, the surface of the detection chip 110 is chemically modified to allow antibodies to be covalently immobilized on the surface, ensuring their specific recognition of the target biomarker. In this way, target proteins such as S100B can accurately bind to the chip surface, forming a label-free immune sandwich structure.

[0055] Unlike traditional immunochromatography, the detection system provided in this application employs a label-free detection strategy. The target protein (such as S100B) binds to selenium nanoparticles through physical adsorption, thereby amplifying the signal. This method eliminates the need for secondary antibodies or complex labeling substances, significantly simplifying the procedure and reducing operational errors.

[0056] This nano-heterogeneous chain structure significantly enhances the detection sensitivity of biomarkers through the enhanced localized surface plasmon resonance (LSPR) effect. Even low concentrations (e.g., 1 pg / mL) of biomarkers can be accurately identified and quantified. This chip eliminates the need for traditional labeling materials and multi-step operations; the label-free detection strategy simplifies and accelerates the process while reducing costs. Furthermore, the detection chip 110 provided in this embodiment is not only suitable for TBI biomarkers such as S100B, but can also be extended to the detection of other neurological disease biomarkers, supporting the early diagnosis of various conditions.

[0057] The image acquisition device 120 is used to acquire detection images containing the detection chip 110. Because the target marker 116 forms an immune sandwich structure with the detection substance in the detection chip 110, signal enhancement is induced. Therefore, in the detection image, the image brightness corresponding to the detection area is different from the image brightness corresponding to the area outside the detection area. Similarly, if the target detector includes multiple substances, such as simultaneously containing S100B protein and glial fibrillary acidic protein, then the brightness of the various detection areas formed by the detection substance and different substances in the target marker in the detection image will also be different.

[0058] In this embodiment, after the sample to be tested is dropped onto the detection chip 110, the target marker 116 in the sample binds to the selenium nanoparticles (SeNPs), resulting in signal enhancement through the localized surface plasmon resonance (LSPR) effect of the nano-heterostructure. During this process, the binding of the target marker 116 significantly enhances the optical signal in the local area, amplifying the optical properties of the selenium nanoparticles (such as scattering or absorption), thereby increasing the signal-to-noise ratio of the imaging. This signal enhancement effect is one of the core characteristics of the nano-heterostructure chip.

[0059] Therefore, in this embodiment, an optical microscope can be used for imaging to generate a detection image containing the detection chip 110. This microscope can acquire optical images of the sample at a specific wavelength and detect the optical signal generated after the target protein binds to the antibody.

[0060] During imaging, images are typically acquired in the near-infrared band to utilize the enhanced signal. The detection area (i.e., the region where the target marker 116 binds to the detection substance) usually shows a stronger signal than the background, while unbound areas show a low background signal. Figure 4 As shown in (a) above, the heterogeneous chains in the negative sample exhibit a uniform weak signal; as Figure 4 As shown in (b), the signal of the target marker 116 binding region 121 in the positive sample was significantly enhanced, with a signal intensity increase of ≥200%, while the unbound region 125 maintained a low background. The bright / dark contrast demonstrated the detection specificity.

[0061] The image processing device 130 is used to invoke a first model 131, which can determine the concentration of each substance in the target marker 116 in the sample to be tested based on the brightness of the detection image; the concentration of the target marker is related to the severity of traumatic brain injury. For example, the first model 131 is a ResNet50 model, which can identify immune sandwich structures in the detection image, such as identifying the brightness of each region in the detection image, determining whether there is an immune sandwich structure in the detection chip 110 based on the brightness of each region, and obtaining the concentration of the target marker 116 by the presence of the immune sandwich structure.

[0062] In one specific embodiment, such as Figure 5 As shown, the first model 131 includes: Convolutional layer 132 is used to process the detection image and extract first image features from the detection image.

[0063] For example, the detected image is processed (e.g., sliced) and then input into the first model 131. The detected image then passes through a series of convolutional layers 132. The function of these convolutional layers 132 is to extract basic visual features from the input image, such as edges, textures, colors, and shapes. Each convolutional layer scans the input image through filters (i.e., convolutional kernels) to generate different feature maps, which represent different levels of detail in the image.

[0064] The residual layer 133 is used to extract a second image feature from the detection image based on a first image feature. The first image feature and the second image feature are different. At least one of the first image feature and the second image feature can reflect the brightness of the detection image.

[0065] For example, the feature map extracted by convolutional layer 132 will be passed to the residual block. This residual block can solve the vanishing gradient and information loss problems that are common in deep network training by introducing "skip connections".

[0066] Each residual block performs the following two steps: Identity Mapping: The residual block allows the input signal to skip part of the convolutional layers 132 and be directly passed to the subsequent layers. This skip connection ensures that the flow of information in the network is not blocked. Residual Learning: By learning the difference between the input and output (i.e., the residual), the network can learn features more efficiently, especially when the number of network layers is very large.

[0067] Thus, in this way, the first model 131 can effectively capture complex features in the image while avoiding overfitting and gradient vanishing problems.

[0068] A fully connected layer 134 is used to determine the concentration of target marker 116 in the sample to be tested based on first image features and second image features.

[0069] After the image is processed by convolutional layer 132 and residual blocks, the extracted features are passed to fully connected layer 134. Through a series of weighted summation processes, the image features are finally converted into quantitative output. In the case of quantitative concentration prediction, the model aims to output a continuous concentration value, representing the concentration of the target marker 116 (such as S100B protein) in the sample to be tested.

[0070] The first model 131 generates numerical values ​​corresponding to the actual concentrations based on previously trained data through regression analysis of the input features. This process is typically optimized using a regression loss function (such as mean squared error) to ensure that the concentration values ​​output by the first model 131 are as accurate as possible.

[0071] pass Figure 6 As shown in the S100B standard curve in the results verification diagram, within the range of 1 pg / mL to 100 ng / mL, the predicted concentration obtained by the detection system provided in this embodiment is in high agreement with the actual value. That is, the detection system provided in this embodiment improves the detection speed while still maintaining high detection accuracy.

[0072] Considering that a single biomarker may not be sufficient to accurately reflect the diverse pathological processes in the clinical assessment of traumatic brain injury (TBI), multiple biomarkers can be detected in some embodiments, such as simultaneous detection of the dual biomarker S100B and GFAP (glial fibrillary acidic protein) to achieve joint typing and concentration prediction of multiple biomarkers.

[0073] Specifically, the target biomarker 116 includes a first sub-biomarker and a second sub-biomarker. For example, the first sub-biomarker is S100B, and the second biomarker is GFAP (glial fibrillary acidic protein).

[0074] The second detector in the second material layer 115 can physically adsorb and bind to the first sub-marker, forming a first scattering signal. For example, S100B forms a red scattering signal with SeNPs. The second detector in the second material layer 115 can physically adsorb and bind to the second sub-marker, forming a second scattering signal; for example, GFAP (glial fibrillary acidic protein) forms a green scattering signal with SeNPs. The first and second scattering signals have different degrees of influence on the brightness of the detected image.

[0075] Based on this, in one embodiment, the image acquisition device 120 includes a controller, a first acquisition channel and a second acquisition channel, and the acquired image includes a first sub-image and a second sub-image; The controller is used to control the first acquisition channel to acquire a first sub-image containing the detection chip 110 based on the first scattering signal in the detection chip 110; and to control the second acquisition channel to acquire a second sub-image containing the detection chip 110 based on the second scattering signal in the detection chip 110.

[0076] Specifically, the image acquisition device 120 can separate light signals of different wavelengths through a dual-channel filter, thereby forming multiple acquisition channels to ensure that the signals of different markers do not interfere with each other.

[0077] The controller in the image acquisition device 120 can switch between dual-channel filters to separate the red and green scattering signals. For example, first, the controller uses the red channel filter to capture the red scattering signal. Then, it switches to the green channel filter to excite the GFAP marker in the sample and capture the green scattering signal.

[0078] Under a red filter, the microscope system excites the red fluorescence of the S100B marker, generating an image that displays the red channel, reflecting the location and concentration of the S100B protein. Switching to a green filter, the microscope system excites the green fluorescence of the GFAP marker, generating an image that displays the green channel, reflecting the location and concentration of the GFAP protein.

[0079] In this embodiment, a multi-channel microscopy system (with dual-channel filters) is used to acquire imaging data of S100B (red channel) and GFAP (green channel) respectively. By selecting appropriate filters, it is ensured that the scattering signals of each marker do not interfere with each other. This provides an efficient and accurate imaging method for multi-marker detection.

[0080] Based on this, in one embodiment, the image processing device 130 is further configured to invoke a second model, which is capable of simultaneously obtaining the concentration of a first sub-marker based on a first sub-image and the concentration of a second sub-marker based on a second sub-image.

[0081] For example, the second model can be a two-stream CNN model, capable of processing image data from two different markers simultaneously. In this process, the two independent image channels correspond to the S100B and GFAP markers, respectively. The S100B channel image originates from the S100B protein marker and is typically displayed as a red light channel (e.g., a red marker). The P channel image originates from the GFAP protein marker and is typically displayed as a green light channel (e.g., a green marker).

[0082] These image data are fed into two independent streams of a two-stream CNN model. Each stream processes its own image data independently, extracting visual features associated with its respective landmarks through convolutional layers 132 and other feature extraction steps.

[0083] In the S100B processing channel, the image input to the S100B channel is processed through convolutional layers 132, pooling layers, and other structures to progressively extract low-level and high-level features from the image. These include, for example, the image's edges, textures, and specific morphological information of the S100B protein.

[0084] Similar to the S100B processing channel, in the GFAP processing channel, GFAP protein-related feature information is extracted independently from the image input to the GFAP channel. At this point, the features recognized by the network are mainly image data related to the binding pattern and region of the GFAP protein.

[0085] Each processing channel can independently process the image features of the corresponding marker, ensuring that the information of each channel is not confused during processing, and extracting key information related to S100B and GFAP respectively.

[0086] After the features of the two processing channels are extracted separately, the features of the S100B and GFAP channels are fused to better understand the relationships between multiple markers involved in the image.

[0087] Specifically, by merging the features obtained from the two processing channels, the two-stream CNN model can comprehensively consider the combined effects of S100B and GFAP in the same image region, thereby obtaining more background information and diagnostic significance. This fusion process is usually performed at the feature level, that is, the convolutional feature maps of the two channels are concatenated, weighted, or summed at a certain level to form a comprehensive feature representation. After feature fusion, the two-stream CNN model processes the fused features through deep convolutional layers 132 and fully connected layers 134. These features are used for composite classification, that is, to classify traumatic brain injury (TBI) based on the joint information of S100B and GFAP markers. The goal of composite classification is to determine the severity of the patient's injury, such as mild, moderate, or severe TBI, based on the concentration information of different markers. For example, Figure 7 As shown, the scatter plot based on dual-marker genotyping can distinguish between mild (○), moderate (△), and severe (□) TBI patient clusters. The dual-stream CNN model can output corresponding genotyping results by learning the distribution patterns of markers for different degrees of lesion severity.

[0088] In some embodiments, the image processing device 130 is also used to predict the development trend of trauma based on the concentration of the target marker 116 in the sample to be tested and the severity of traumatic brain injury.

[0089] Considering that the occurrence and development of trauma is a dynamic process, the concentration of target biomarker 116 will change over time. Therefore, by collecting samples and measuring the concentration at different time points (such as 15 minutes, 30 minutes, 1 hour, 3 hours after trauma), the image processing device 130 can construct a time series dataset.

[0090] Then, based on this time-series dataset, a third model is trained to identify the relationship between changes in biomarker concentration and trauma progression. For example, the third model could be a deep learning model (such as a Residual Gated Recurrent Unit (ResGRU) or a Long Short-Term Memory Network (LSTM)). These models are able to learn the correlation between biomarker concentrations in the time series and trauma progression trends (such as injury worsening, stabilization, or improvement).

[0091] Once the third model is trained, it can predict the future development of trauma based on real-time measured biomarker concentrations. By analyzing the trends in biomarker changes after trauma, the third model can predict whether the trauma will worsen or improve in the next few hours or days. For example, if S100B concentration rises sharply and remains high in a short period, the model may predict that the trauma will develop into moderate or severe brain injury; if GFAP concentration is low and stable, it may indicate milder injury or recovery.

[0092] In some embodiments, based on the concentration of target marker 116, the image processing device 130 can not only predict the severity of trauma (e.g., mild, moderate, severe), but also classify the trauma. For example, based on the combined concentrations of S100B and GFAP, the model may classify trauma into different types: mild trauma (low S100B, high GFAP); moderate trauma (moderate S100B and GFAP concentrations); and severe trauma (high S100B, very high GFAP).

[0093] By correlating biomarker concentrations with trauma recovery trends, the system can perform prognostic assessments. For example, if biomarker concentrations rapidly decrease and stabilize within the first hour after trauma, it may predict a good recovery; conversely, if concentrations continue to rise or fluctuate, it may indicate a risk of trauma worsening. Reports are generated based on the model's output concentration data and predictions, providing the current status of the trauma (mild, moderate, severe) and its development trend (improvement, stability, worsening). For example: "S100B concentration is 25 pg / mL, GFAP concentration is 150 pg / mL, predicting the trauma to develop into moderate, with the injury potentially worsening within the next 72 hours." Clinicians can use these predictions to make more precise decisions. For instance, if a worsening trend is predicted, doctors may choose more aggressive interventions, such as medication or surgery.

[0094] The biomarker detection system provided in this application has the advantages of rapid detection, low sample size, and high temporal resolution; combined with AI sequence modeling algorithms, it can realize multi-time point TBI dynamic analysis and mid-to-long-term prognostic assessment; and provide a high-throughput, intelligent research tool for animal research on traumatic brain injury and drug screening.

[0095] The image processing device, the first model, and the second model in the above embodiments can all be integrated into a server. The server can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0096] The following embodiment of this application provides a biomarker detection method for rapidly detecting the S100B protein level in the serum of patients with traumatic brain injury (TBI). By combining a nanostructure assembled chip and a deep learning model, the method achieves accurate identification and quantitative analysis of sample signals.

[0097] For example, the detection process includes the following: S21: Pre-incubate the S100B standard solution with SeNPs for 20 minutes.

[0098] S22: Add 70 μL of incubation solution to the surface of the functionalized chip, react for 30 minutes, and then wash and dry with ultrapure water.

[0099] S23: Use a microscope for optical imaging to obtain an image of the red scattering signal.

[0100] S24: Use the ResNet50 network to extract features and classify regions containing heterogeneous chains in the image.

[0101] S25: Establish classification models and concentration prediction regression models for six concentration levels (1 pg / mL to 100 ng / mL).

[0102] S26: Using the cross-entropy loss function and R 2 Model evaluation showed an accuracy rate >90%, R0 2 >0.98.

[0103] Comparative analysis of the same batch of samples using the commercial ThermoFisher ELISA showed a Pearson correlation coefficient >0.95. Validation was performed using samples from 30 clinical TBI patients and a mouse model, accurately identifying different degrees of lesion severity. The total testing time was <60 minutes, and the detection limit was 1 pg / mL.

[0104] In the following embodiments, the biomarker detection system can also be applied to multi-marker joint detection and complex injury classification analysis, thereby further verifying the system's multi-throughput recognition capability and the generalization ability of the deep learning model to different biomarker responses. In the clinical assessment of traumatic brain injury (TBI), a single biomarker may not be sufficient to accurately reflect the diverse pathological processes. Therefore, this embodiment introduces a dual-marker detection strategy of S100B and GFAP (glial fibrillary acidic protein) to verify the system's compatibility with multi-marker imaging and AI analysis.

[0105] (1) Material preparation GFAP monoclonal antibody: purchased from Abcam (Cat# ab7260), diluted to a concentration of 10 μg / mL.

[0106] SeNPs complex incubation method: S100B and GFAP mixed standards were incubated with SeNPs solution (0.01 mg / mL) to form complex SeNP particles containing multiple target proteins.

[0107] Chip functionalization method: Using dual antibody sequential modification technology, anti-S100B and anti-GFAP antibodies are immobilized sequentially, and different signal channels are labeled respectively (red light / green light enhanced scattering).

[0108] (2) Imaging and data acquisition A composite SeNPs solution (70 μL) was dropped onto the surface of the functionalized chip, incubated for 30 minutes, washed, and then dried.

[0109] Imaging data from the S100B (red channel) and GFAP (green channel) were acquired using a multi-channel microscope system (with dual-channel filters).

[0110] The system can automatically identify different color-enhanced signals and generate two independent signal channel diagrams.

[0111] (3) AI recognition and model building The model input is a two-channel patch image, with each patch being 224×224px in size.

[0112] A dual-stream CNN architecture is used to process S100B and GFAP signal images simultaneously.

[0113] The network extracts joint features at the fusion layer and outputs composite concentration regression values ​​and damage severity classifications (mild / moderate / severe).

[0114] The model training set consisted of 90 composite concentration samples, and the test set consisted of 30 samples with unknown concentrations. The accuracy reached 92.6%, and the multi-class F1 score was 0.91.

[0115] (4) Systematic review and clinical validation In serum samples from 15 patients with total glial brain injury (TBI), S100B and GFAP levels were simultaneously measured. The AI ​​model accurately identified patient subtypes: high S100B / low GFAP (indicating early glial response) and high GFAP / low S100B (indicating mid-to-late stage damage). Compared with ELISA values, the composite assay results showed a high degree of consistency with the actual concentrations (R0). 2 >0.96).

[0116] In the following embodiments, the biomarker detection system also possesses the capability for dynamic monitoring and prognostic prediction in small animal models, thereby further expanding its application prospects in basic research and translational medicine. For mouse models commonly used in preclinical traumatic brain injury (TBI) research, this detection system is used to perform high-frequency detection of dynamic changes in serum S100B protein, and combined with an AI model to achieve trauma grading and prognostic trend prediction.

[0117] (1) Experimental model Animal model: Male C57BL / 6J mice, 8 weeks old, weighing approximately 22–25g, were used.

[0118] TBI induction method: The controlled cortical impulse (CCI) method was used to establish mild, moderate and severe TBI models, with n=5 in each group.

[0119] Serum sample collection time points: 15 minutes, 30 minutes, 1 hour, 3 hours, 6 hours and 12 hours after surgery, a total of 6 time points.

[0120] (2) Sample processing and detection Approximately 100 μL of mouse serum was collected at each time point, and the supernatant was collected after centrifugation at 10,000 g.

[0121] Serum samples were mixed with SeNPs at a 1:1 ratio and incubated for 20 minutes, resulting in a final SeNP concentration of 0.01 mg / mL.

[0122] 70 μL of the mixture was added to the detection chip, and after standing for 30 minutes, it was cleaned and dried.

[0123] (3) Imaging and AI Modeling Approximately 500 image patches (224×224 px) are acquired for each chip region, and red enhancement signal features are extracted.

[0124] A time-series-based residual regression neural network (ResGRU) is constructed, with the model input being a multi-time point image feature sequence.

[0125] Outputs include: protein concentration fit values; severity score (0–100 points); and prognostic trend prediction (e.g., symptom improvement or worsening trend within 72 hours).

[0126] (4) Performance evaluation The trend of S100B concentration change showed good consistency with the ELISA results (mean square error MSE < 0.04).

[0127] The average accuracy of the mild, moderate and severe models in AI scoring reached 93.2%.

[0128] The model can predict the trend of neurological function score (mNSS) changes over 72 hours after the 3rd hour, with a Spearman correlation coefficient of 0.89.

[0129] Therefore, the biomarker detection system provided in this application has the advantages of rapid detection, low sample size, and high temporal resolution; combined with AI sequence modeling algorithm, it can realize multi-time point TBI dynamic analysis and mid-to-long-term prognostic assessment; and provide a high-throughput, intelligent research tool for animal research on traumatic brain injury and drug screening.

[0130] The biomarker detection system provided in this application has undergone multiple rounds of validation, including animal experiments, human sample testing, and deep neural network analysis simulations, fully demonstrating its feasibility and effectiveness in detecting biomarkers related to brain injury. Specific results are as follows: (1) Animal model experiments demonstrated the system's sensitivity and early diagnostic capability. Serum samples were collected at six time points (15 min, 30 min, 1 h, 3 h, 6 h, 12 h) using mouse models of mild, moderate, and severe total brain injury (TBI). This detection chip achieved accurate detection of S100B protein at the pg / mL level, and showed a high correlation with the severity of injury, exhibiting a good dose-response relationship. The results were highly consistent with commercial ELISA kits (correlation R0). 2 >0.95), and the detection speed is significantly improved.

[0131] (2) Reproducibility and applicability of the preliminary validation system for clinical samples A total of 75 clinical TBI patients (25 mild, 25 moderate, and 25 severe) and healthy controls were included. Serum, urine, and saliva samples were collected for blind testing. The test results obtained based on the detection chip showed high consistency with the disease grading, and the AUC (ROC) reached over 0.91, demonstrating clinical translation potential.

[0132] In summary, the biomarker detection system proposed in this application has undergone multiple rounds of empirical testing, verifying its significant advantages in terms of sensitivity, specificity, throughput, automation, and clinical applicability, and has good technological maturity and commercialization feasibility.

[0133] This application also provides a method for detecting biomarkers, such as... Figure 8 As shown, the method includes: S100: Acquire a detection image containing the detection chip; S200: Using the first model, the concentration of the target biomarker in the sample to be tested is determined based on the brightness of the detected image; the concentration of the target biomarker is related to the severity of traumatic brain injury; the target biomarker includes at least one of the following substances: S100B protein and glial fibrillary acidic protein; S300: Determine the severity of traumatic brain injury based on the concentration of each target biomarker in the sample to be tested; Specifically, the target marker in the sample to be tested can be physically adsorbed and bound to the detection substance in the detection chip, forming a detection area in the detection chip; the detection substance is selenium nanoparticles; the image brightness corresponding to the detection area in the detection image is different from the image brightness corresponding to the area outside the detection area; the brightness of each detection area formed by the detection substance and different substances in the target marker in the detection image is different.

[0134] In this embodiment, the detection substance in the detection chip physically adsorbs and binds to the target marker, achieving precise control of the optical coupling effect and forming a significantly enhanced signal, thereby enabling the formation of identifiable image features in subsequent detection images. By identifying and analyzing the detection image containing the detection chip, the concentration of the target marker is obtained. This significantly reduces human interpretation bias and improves diagnostic consistency. Furthermore, the detection method is simple, requiring no labeling or pretreatment of the nanoprobes, which greatly increases detection speed and has broad application prospects.

[0135] In some embodiments, determining the concentration of a target biomarker in a sample to be tested based on the brightness of a detected image using a first model includes: processing the detected image using the first model to extract a first image feature from the detected image; extracting a second image feature from the detected image based on the first image feature using the first model, wherein the first image feature and the second image feature are different; at least one of the first image feature and the second image feature can reflect the brightness of the detected image; and determining the concentration of the target biomarker in the sample to be tested based on the first image feature and the second image feature using the first model.

[0136] Based on the above embodiments, this application also provides a computer device. Figure 9 A schematic diagram of a computer device structure is provided for an embodiment of this application, such as... Figure 9 As shown, it includes: processor 501, communication interface 502, memory 503 and communication bus 504, wherein processor 501, communication interface 502 and memory 503 communicate with each other through communication bus 504.

[0137] The memory 503 stores a computer program. When the program is executed by the processor 501, the processor 501 performs the following steps: S100: Acquire a detection image containing the detection chip; S200: Using the first model, the concentration of the target biomarker in the sample to be tested is determined based on the brightness of the detected image; the concentration of the target biomarker is related to the severity of traumatic brain injury; the target biomarker includes at least one of the following substances: S100B protein and glial fibrillary acidic protein; S300: Determine the severity of traumatic brain injury based on the concentration of each target biomarker in the sample to be tested; Specifically, the target marker in the sample to be tested can be physically adsorbed and bound to the detection substance in the detection chip, forming a detection area in the detection chip; the detection substance is selenium nanoparticles; the image brightness corresponding to the detection area in the detection image is different from the image brightness corresponding to the area outside the detection area; the brightness of each detection area formed by the detection substance and different substances in the target marker in the detection image is different.

[0138] In some embodiments, the processor 501 further performs the following steps: The first model is used to process the detected image in order to extract the first image features from the detected image.

[0139] Using a first model, based on first image features, a second image feature is extracted from the detection image. The first image feature and the second image feature are different. At least one of the first image feature and the second image feature can reflect the brightness of the detection image.

[0140] The concentration of the target marker in the sample to be tested is determined using the first model, based on the first image features and the second image features.

[0141] The communication bus mentioned in the above computer equipment can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0142] Communication interface 502 is used for communication between the aforementioned computer equipment and other equipment.

[0143] The memory may include RAM (Random Access Memory) or NVM (Non-Volatile Memory), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0144] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be DSPs (Digital Signal Processors), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0145] Based on the above embodiments, this application provides a computer-readable storage medium storing a computer program executable by a computer device. When the program is run on the computer device, the computer device performs the following steps: S100: Acquire a detection image containing the detection chip; S200: Using the first model, the concentration of the target biomarker in the sample to be tested is determined based on the brightness of the detected image; the concentration of the target biomarker is related to the severity of traumatic brain injury; the target biomarker includes at least one of the following substances: S100B protein and glial fibrillary acidic protein; S300: Determine the severity of traumatic brain injury based on the concentration of each target biomarker in the sample to be tested; Specifically, the target marker in the sample to be tested can be physically adsorbed and bound to the detection substance in the detection chip, forming a detection area in the detection chip; the detection substance is selenium nanoparticles; the image brightness corresponding to the detection area in the detection image is different from the image brightness corresponding to the area outside the detection area; the brightness of each detection area formed by the detection substance and different substances in the target marker in the detection image is different.

[0146] In some embodiments, the computer device performs the following steps: The first model is used to process the detected image in order to extract the first image features from the detected image.

[0147] Using a first model, based on first image features, a second image feature is extracted from the detection image. The first image feature and the second image feature are different. At least one of the first image feature and the second image feature can reflect the brightness of the detection image.

[0148] The concentration of the target marker in the sample to be tested is determined using the first model, based on the first image features and the second image features.

[0149] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0150] The embodiments of this application have now been described in detail. To avoid obscuring the concept of this application, some details known in the art have not been described. Those skilled in the art can fully understand how to implement the technical solutions disclosed herein based on the above description.

[0151] While specific embodiments of this application have been described in detail by way of examples, those skilled in the art should understand that the above examples are for illustrative purposes only and are not intended to limit the scope of this application. Those skilled in the art should understand that modifications can be made to the above embodiments or equivalent substitutions can be made to some technical features without departing from the scope and spirit of this application. In particular, as long as there is no structural conflict, the various technical features mentioned in the embodiments can be combined in any manner.

Claims

1. A biomarker detection system, characterized in that, include: The detection chip contains a detection substance, which physically adsorbs and binds to each substance in the target biomarker in the sample to be tested, forming a corresponding detection area in the detection chip; the target biomarker includes at least one of the following substances: S100B protein and glial fibrillary acidic protein. The detection substance is selenium nanoparticles; An image acquisition device is used to acquire a detection image containing the detection chip, wherein the image brightness corresponding to the detection area in the detection image is different from the image brightness corresponding to the area outside the detection area; and the brightness of each detection area formed by the detection substance and different substances in the target marker in the detection image is different. An image processing device is used to invoke a first model, which is capable of determining the concentration of each substance in the target marker in the sample to be tested based on the brightness of the detected image. The concentration of the target biomarker is correlated with the severity of traumatic brain injury; The image processing device is also used to determine the severity of the traumatic brain injury based on the concentration of each substance in the target marker in the sample to be tested.

2. The system according to claim 1, characterized in that, The detection chip includes: Base; A first material layer, wherein a first detector array in the first material layer is arranged on the substrate; An antibody-modified layer is disposed on the first detection substance; The second material layer contains a second detection substance that forms a heterogeneous chain with the first detection substance through the antibody modification layer; the second detection substance is capable of physical adsorption and binding with the target marker in the sample to be tested.

3. The system according to claim 1, characterized in that, The first model includes: A convolutional layer is used to process the detected image and extract a first image feature from the detected image; A residual layer is used to extract a second image feature from the detection image based on the first image feature, wherein the first image feature and the second image feature are different; at least one of the first image feature and the second image feature can reflect the brightness of the detection image. A fully connected layer is used to determine the concentration of the target marker in the sample to be tested based on the first image features and the second image features.

4. The system according to claim 2, characterized in that, The target marker includes a first sub-marker and a second sub-marker; The second detector in the second material layer can physically adsorb and bind with the first sub-marker to form a first scattering signal; The second detector in the second material layer can physically adsorb and bind with the second sub-marker to form a second scattering signal; The first scattering signal and the second scattering signal have different degrees of influence on the brightness of the detected image.

5. The system according to claim 4, characterized in that, The image acquisition device includes a controller, a first acquisition channel, and a second acquisition channel, and the detected image includes a first sub-image and a second sub-image. The controller is used to control the first acquisition channel to acquire a first sub-image containing the detection chip based on the first scattering signal in the detection chip. The second acquisition channel is controlled to acquire the second scattering signal based on the detection chip, which includes the second sub-image of the detection chip.

6. The system according to claim 5, characterized in that, The image processing device is also used to invoke a second model, which is capable of simultaneously obtaining the concentration of the first sub-marker based on the first sub-image and the concentration of the second sub-marker based on the second sub-image.

7. The system according to any one of claims 1-6, characterized in that, The image processing device is also used to predict the development trend of trauma based on the concentration of the target marker in the test sample and the severity of the traumatic brain injury.

8. A method for detecting a biomarker, characterized in that, The method is applied to the system according to any one of claims 1-7, and the method comprises: Acquire a detection image containing the detection chip; Using the first model, the concentration of target markers in the test sample is determined based on the brightness of the detected image; the concentration of the target markers is related to the severity of traumatic brain injury; the target markers include at least one of the following substances: S100B protein and glial fibrillary acidic protein; The severity of the traumatic brain injury is determined based on the concentration of each of the target biomarkers in the sample to be tested. Specifically, the target marker in the sample to be tested can physically adsorb and bind with the detection substance in the detection chip, forming a detection area in the detection chip; the detection substance is selenium nanoparticles; the image brightness corresponding to the detection area in the detection image is different from the image brightness corresponding to the area outside the detection area; the brightness of each detection area formed by the detection substance and different substances in the target marker in the detection image is different.

9. The method according to claim 8, characterized in that, The step of determining the concentration of the target marker in the sample to be tested based on the brightness of the detected image using the first model includes: The first model is used to process the detected image to extract the first image feature from the detected image; Using the first model, based on the first image features, a second image feature is extracted from the detected image. The first image feature and the second image feature are different. At least one of the first image feature and the second image feature can reflect the brightness of the detected image. Using the first model, based on the first image features and the second image features, the concentration of the target marker in the sample to be tested is determined.

10. A computer device, comprising a processor and a memory storing a computer program, characterized in that, When the processor executes the program, it performs the method as described in claim 8 or 9.

Images