Surface-enhanced raman spectroscopy detection system based on gold nanostir micro-nanorobot

Abstract

The application provides a surface-enhanced Raman spectrum detection system based on a gold nanosticker micro-nano robot, which comprises a micro-fluidic chip module, a magnetic spiral micro-nano robot with controllable growth of gold nanostickers on the tip, a magnetic field control module for driving the micro-nano robot to perform three-dimensional targeted movement, a Raman spectrum acquisition module for acquiring SERS signals, and a data processing and classification module internally integrated with a one-dimensional convolutional neural network (1D-CNN) model for spectrum preprocessing and intelligent classification. The application integrates precise control of the micro-nano robot, efficient SERS enhancement of the gold nanostickers, and intelligent analysis of the deep learning model, has the advantages of high detection precision, stable signals, adaptation to complex environments, easy deployment to portable devices, and the like, and can be used in the fields of in-vitro biochemical analysis, drug screening, environmental monitoring and the like.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of micro-nano robots and surface-enhanced Raman spectroscopy, and in particular to a surface-enhanced Raman spectroscopy detection system based on gold nanoparticle-based micro-nano robots. Background Technology

[0002] Combining magnetic helical micro / nano robots with surface-enhanced Raman spectroscopy (SERS) to achieve remote, precise manipulation of micro / nano probes and in-situ spectroscopic detection using external magnetic fields is a current research hotspot in biosensing and precision medicine. Existing techniques utilize gold nanospikes grown on the surface of microrobots as SERS substrates to simulate targeted motion and molecular detection processes in microfluidic chips. However, current technologies still have several limitations:

[0003] (1) Insufficient stability and controllability: The uniformity of the rotating magnetic field is easily disturbed, affecting the motion accuracy; gold nanoparticles are prone to agglomeration or detachment in complex fluid environments, resulting in poor repeatability of SERS signals.

[0004] (2) Lack of standardization and intelligence: The influence of operating parameters (such as magnetic field and viscosity) is not systematically evaluated, and there is a lack of standardized processes. Data processing relies on traditional methods, making it difficult to achieve high-precision and automated classification and diagnosis.

[0005] (3) High system cost and difficult conversion: It relies on high-precision magnetic field equipment and spectrometers, and lacks verification of movement and detection in real complex samples (such as high-viscosity biological mucus).

[0006] Therefore, developing an integrated diagnostic method that combines high-precision motion control, stable SERS signal enhancement, and intelligent data analysis capabilities has significant application value. Summary of the Invention

[0007] To address the above technical problems, this invention discloses a surface-enhanced Raman spectroscopy detection system based on gold nanoparticle-based micro-nano robots. This system integrates controllable preparation, precise magnetic drive, in-situ SERS sampling, and intelligent spectral analysis, improving stability, sensitivity, and practicality. It overcomes the shortcomings of existing technologies, such as insufficient motion control precision of micro-nano robots, unstable SERS substrate signals, and low level of intelligent data analysis.

[0008] The technical solution adopted by this invention is as follows:

[0009] A surface-enhanced Raman spectroscopy detection system based on gold nanoparticle-based microrobots, characterized in that it comprises:

[0010] Microfluidic chip modules are used to contain samples to be tested and provide a microchannel environment;

[0011] Magnetic spiral micro-nano robots, with gold nanospikes grown controllably at their tip, exhibit magnetic responsiveness;

[0012] The magnetic field control module is used to generate and control the rotating magnetic field to drive the micro-nano robot to perform three-dimensional motion in the microfluidic chip and target the target area.

[0013] The Raman spectroscopy acquisition module is used to excite and acquire surface-enhanced Raman spectral signals on the gold nanospikes at the tip of the micro-nano robot;

[0014] The data processing and classification module integrates a spectral classification model based on a one-dimensional convolutional neural network (1D-CNN) to preprocess and classify the acquired SERS signals and output the classification results of the samples.

[0015] Using this technical solution, gold nanospikes achieve locally controllable growth at their tips, resulting in a stable structure that is not prone to aggregation and detachment. They exhibit high SERS enhancement efficiency and good signal repeatability. The three-dimensional magnetic drive provides precise navigation, adapting to complex fluids, high-viscosity media, and narrow channels, ensuring reliable targeted positioning. The deep learning algorithm is lightweight, features comprehensive preprocessing, and boasts high classification accuracy. It can be embedded and is compatible with portable devices.

[0016] As a further improvement of the present invention, the magnetic helical micro / nano robot is prepared by the following steps:

[0017] Step S1: A spiral structure array is prepared using micron-level 3D printing technology. The printing material contains magnetic particles, image developer, and photosensitive resin. After cleaning, a nickel layer and a platinum layer are deposited sequentially on the surface of the spiral structure to obtain a microrobot.

[0018] Step S2: The spiral array tip of the microrobot is immersed in the reaction solution containing the gold precursor by the micro-motion platform to achieve localized controllable growth of gold nanospikes; the reaction solution containing the gold precursor includes chloroauric acid, hydrogen peroxide and silver nitrate.

[0019] As a further improvement of the present invention, in step S1, the 3D printing exposure parameters are 55 mW / cm² and the exposure time is 20s.

[0020] As a further improvement of the present invention, the nickel layer thickness is 500 nm.

[0021] As a further improvement of the present invention, the printing material comprises photosensitive resin, ethylene dimethacrylate, PVP, barium sulfate and neodymium iron boron magnetic particles.

[0022] As a further improvement of the present invention, in step S2, the immersion depth and time are precisely controlled by a micro-displacement platform through the gold nanospike growth process to ensure that the reinforcing structure is formed only at the tip of the robot.

[0023] As a further improvement of the present invention, the magnetic field control module includes a three-axis Helmholtz electromagnetic coil, which changes the intensity, rotation frequency, spatial tilt angle and orientation angle of the rotating magnetic field by real-time adjustment of the input current signal, so as to realize three-dimensional trajectory navigation, target positioning and attitude control of the micro-nano robot within the channel of the microfluidic chip.

[0024] As a further improvement of the present invention, the surface of the gold nanoparticles is grafted with Raman reporter molecular probes; when the surface-enhanced Raman spectroscopy detection based on gold nanoparticle micro-nano robots is used to detect hydrogen peroxide, the concentration of hydrogen peroxide is determined by detecting the change in the intensity ratio of the characteristic peaks of the Raman reporter molecular probes, thereby achieving quantitative detection for non-diagnostic purposes.

[0025] As a further improvement of the present invention, the preprocessing in the data processing and classification module includes:

[0026] Cosmic ray removal steps: Cosmic ray interference peaks are identified based on adaptive peak aspect ratio detection and half-width threshold, and linear interpolation is used for repair.

[0027] Baseline correction steps: Adaptive iterative reweighted penalized least squares method based on Cholesky decomposition optimization is used to remove fluorescence background;

[0028] Smoothing steps: Apply Savitzky-Golay filter to smooth the signal.

[0029] As a further improvement of the present invention, the structure of the one-dimensional convolutional neural network model specifically includes: three one-dimensional convolutional layers with output channels of 8, 64 and 64 respectively, each convolutional layer followed by a ReLU activation layer, and the convolutional kernel size is 15×1.

[0030] Two one-dimensional max pooling layers, with a pooling window and stride of 2;

[0031] The network consists of three fully connected layers with output dimensions of 512, 512, and K, respectively, where K is the total number of classification categories; a Softmax layer at the end of the network; and a training module that uses the cross-entropy loss function.

[0032] As a further improvement of the present invention, the data processing and classification module is deployed on a portable edge computing device or a local terminal that communicates with a cloud server, supporting offline operation; the one-dimensional convolutional neural network model is optimized by quantization or pruning to adapt to the computing resources of the embedded device.

[0033] This invention discloses a sample detection method based on surface-enhanced Raman spectroscopy using gold nanoparticles / microrobots, employing the surface-enhanced Raman spectroscopy detection system based on gold nanoparticles / microrobots as described in any of the preceding claims, and comprising the following steps:

[0034] Step S10: Place the magnetic spiral micro-nano robot into the microfluidic chip containing the sample to be tested;

[0035] Step S20: A controllable rotating magnetic field is generated by an external magnetic field control system to drive the micro-nano robot to perform three-dimensional trajectory movement within the chip channel and target the detection area.

[0036] Step S30: Using the gold nanospikes grown at the tip of the micro-nano robot as a surface-enhanced Raman spectroscopy substrate, the target marker in the detection area is sampled in situ and the signal is enhanced to acquire the corresponding SERS spectral signal.

[0037] Step S40: Preprocess the acquired SERS spectral signal and input it into a pre-trained one-dimensional convolutional neural network model to output the classification result of the sample;

[0038] The method is used for in vitro sample analysis for non-diagnostic purposes.

[0039] As a further improvement of the present invention, in step S20, the intensity, frequency, spatial tilt angle and orientation angle of the rotating magnetic field are changed by real-time control of the input current of the triaxial Helmholtz electromagnetic coil device to achieve precise navigation and target positioning of the three-dimensional motion trajectory of the micro-nano robot; Raman reporter molecular probes are grafted onto the surface of the gold nanospikes; when the target marker is hydrogen peroxide, the concentration of hydrogen peroxide is determined by detecting the change ratio of the characteristic peak of the Raman reporter molecular probe.

[0040] This invention discloses the application of the sample detection method based on surface-enhanced Raman spectroscopy of gold nanoparticles microrobots as described above, for microfluidic in vitro sample analysis, laboratory detection of biomarkers, detection of environmental substances, or material analysis.

[0041] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0042] First, high precision and high stability: By optimizing the robot's three-layer metal structure (nickel layer + platinum layer) and the controllable growth process of gold nanospikes, the magnetic responsiveness and structural stability of the SERS substrate are significantly improved. Combined with precise magnetic field control of a three-dimensional Helmholtz coil, accurate targeting in complex microfluidic environments is achieved.

[0043] Second, integrated intelligent diagnosis: "Sampling-enrichment-detection-analysis" is integrated onto a single micro-nano robot. By introducing an improved 1D-CNN deep learning model, automatic denoising, baseline correction, and end-to-end high-precision classification of the original SERS signal are achieved (test accuracy reaches 94.29%), overcoming the problems of traditional analysis methods that rely on human experience and have low processing efficiency.

[0044] Third, adaptability to complex environments: This method and system are suitable for high-viscosity, non-uniform simulated biological fluid environments. The gold nanoparticle structure improves local sampling efficiency, while the optimized magnetocontrol strategy ensures the robot's ability to penetrate complex media.

[0045] Fourth, lightweight and easy to deploy: The designed 1D-CNN model has a small number of parameters (about 16,000), high computational efficiency, and a single spectrum processing time of about 3 minutes. It can be easily deployed in portable Raman spectrometers or embedded edge devices, supports offline operation, reduces system costs, and enhances clinical translation potential. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the photopolymerization process for preparing a spiral array in an embodiment of the present invention.

[0047] Figure 2 This is a schematic diagram illustrating the in-situ growth of nano-gold spikes in a localized region at the tip of a spiral using a micro-motion platform in an embodiment of the present invention.

[0048] Figure 3 This is a schematic diagram illustrating the motion control of a magnetic helical microrobot in a microfluidic chip according to an embodiment of the present invention.

[0049] Figure 4 This is a schematic diagram illustrating Raman sensing detection of a sample on a gold nanoparticle-reinforced substrate in an embodiment of the present invention.

[0050] Figure 5 In this embodiment of the invention, based on the VGG-16 deep learning model architecture, Raman data is input into an initial convolutional layer with 64 filters. Detailed Implementation

[0051] The preferred embodiments of the present invention will be described in further detail below.

[0052] Example 1

[0053] The fabrication of gold nanoparticle-like microrobots involves the following steps:

[0054] (1) Fabrication of helical structure: Helical arrays were fabricated using a micron-scale 3D printer (exposure parameters: 55 mW / cm², 20 s).

[0055] To endow the biopsy microrobot with imaging and strong magnetic response capabilities, this embodiment employs a mixture of 12 g of photosensitive resin and 4 mL of ethylene dimethacrylate monomer pre-dissolved with 3 g of PVP, along with 6 g of barium sulfate contrast agent and 8 g of neodymium iron boron magnetic particles. After thorough homogenization using a planetary centrifugal mixer, the mixture is subjected to secondary 3D printing to obtain a magnetic helical micro / nano robot with both magnetic response and imaging capabilities. After printing, the robot is sequentially immersed in ethanol and deionized water for 1 hour each to remove uncured resin. A 500 nm nickel layer and a platinum layer are then sequentially deposited on the surface of the helical array using an electron beam evaporation system. Figure 1 As shown. Further, the photosensitive resin is Rubik's Cube S130.

[0056] (2) Growth of gold nanospike tips: A spiral array coated with nickel and platinum layers was fixed on a micro-displacement platform. The process of the spiral structure immersing in the reagent solution was observed using a contact angle meter. The reaction reagent was prepared by mixing 600 μL of chloroauric acid, 600 μL of hydrogen peroxide, and 120 μL of silver nitrate in 1 mL of deionized water. Figure 2 As shown, the depth of the helical array tip immersed in the reagent solution was adjusted using a micro-displacement platform. After different reaction times, the helical array was removed from the reagent solution, rinsed with deionized water, dried in air, and the microrobots were collected, yielding magnetic micro / nanorobots with gold nanospike structures at their tips.

[0057] The motion control of the magnetic micro / nano robots with gold nanospike structures at their tips, prepared as described above, in a microfluidic chip includes the following steps:

[0058] In a triaxial Helmholtz electromagnetic coil device with integrated optical imaging, a fabricated micro-nano robot is placed in a microfluidic chip channel. The speed and direction of the fluid inside the chip are precisely controlled by an injection pump system to simulate different physiological flow environments.

[0059] The three-axis Helmholtz electromagnetic coil device is activated to generate a uniform rotating magnetic field, driving the robot's stable helical motion. By adjusting the input current, the magnetic field strength, rotation frequency, and spatial angle can be precisely controlled. Through real-time adjustment of the magnetic field's tilt and orientation angles, precise navigation and targeted positioning of the microrobot's three-dimensional motion trajectory can be achieved. Figure 3 As shown, the magnetic micro / nano robot with gold nanospikes at its tip obtained in this embodiment can complete stable and controllable motion and attitude adjustment in media with different flow rates and viscosities and in narrow channel structures.

[0060] Using the magnetic micro / nano robot obtained in this embodiment, combined with Raman detection, fingerprint peak detection of sampled substances can be achieved. Gold nanoparticles are not only high-performance three-dimensional samplers but also excellent Raman signal enhancement substrate materials, forming a substrate for constructing Raman reporter molecular probes, thereby enabling the detection of markers, such as... Figure 4 As shown, Raman reporter molecules are efficiently grafted onto the surface of gold nanoparticles to construct a Raman reporter molecule detection probe (Au-4MBA). Au-4MBA will be oxidized by H2O2 in the microenvironment, causing changes in the Raman characteristic peaks before and after 4MBA. The ratio of the two characteristic peaks is linearly related to the H2O2 concentration.

[0061] In-situ detection and intelligent analysis of SERS spectra for non-diagnostic purposes includes the following steps:

[0062] (1) SERS detection: Raman reporter molecules (such as 4MBA (4-mercaptobenzoic acid)) were grafted onto the surface of gold nanoparticles as probes (Au-4MBA). When 4MBA was oxidized by H2O2 in the microenvironment, its Raman characteristic peaks changed, and the proportion of characteristic peaks was linearly correlated with the H2O2 concentration. After robot targeting and positioning, SERS signals were acquired by Raman spectroscopy (see...). Figure 4 ).

[0063] (2) Spectral preprocessing:

[0064] Cosmic ray removal: Interference peaks are identified by calculating the peak height to half-width ratio (>180) and the half-width threshold (<6), and linear interpolation over a range of 2.5 times the half-width is used for repair.

[0065] Baseline correction: Adaptive iterative reweighted penalized least squares (arPLS) based on Cholesky decomposition optimization is used to remove fluorescence background.

[0066] Smoothing: A Savitzky-Golay filter with a window size of 11 points is used for smoothing.

[0067] Z-score standardization eliminates sample differences.

[0068] (3) Deep learning classification:

[0069] Raman spectroscopy analysis employs a convolutional neural network based on an improved version of AlexNet, implemented using the PyTorch framework, with the aim of optimizing Raman spectroscopy classification performance.

[0070] The model comprises three one-dimensional convolutional layers with output channels of 8, 64, and 64, respectively. Each convolutional layer is followed by a ReLU activation layer to introduce non-linearity and enhance feature learning capability. To improve classification accuracy, a longer convolutional kernel (15×1) is used to better capture the spectral features of the Raman data, such as... Figure 5 As shown, the model also includes two one-dimensional max-pooling layers with a pooling window and stride of 2, preserving key information by progressively reducing the feature map dimensionality. At the end of the network architecture are three fully connected layers that integrate the extracted features for final classification. During training, the model output is converted from logistic values ​​to probabilities via a softmax layer, using the cross-entropy loss function, the Adam optimizer with default parameters, a batch size of 128, and 200 training epochs. A weight balancing strategy is used to mitigate data imbalance, adjusting the weights of each training sample during loss calculation to ensure equal contribution of each class to the loss.

[0071] The above method achieved an average accuracy of 94.29% on the test set in a 5-class classification task for lung cancer-related biomarkers. The Integrated Gradients method allows for interpretability analysis of model decisions and identification of key Raman characteristic peaks.

[0072] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A surface-enhanced Raman spectroscopy detection system based on gold nanoparticle-based micro / nano robots, characterized in that, include: Microfluidic chip module for containing samples and providing a microchannel environment; magnetic spiral micro-nano robot with gold nanospikes grown controllably at its tip, exhibiting magnetic responsiveness; The magnetic field control module is used to generate and control the rotating magnetic field to drive the micro-nano robot to perform three-dimensional motion in the microfluidic chip and target the target area. The Raman spectroscopy acquisition module is used to excite and acquire surface-enhanced Raman spectral signals from the gold nanoparticles at the tip of the micro-nano robot; the data processing and classification module integrates a spectral classification model based on a one-dimensional convolutional neural network, which is used to preprocess and classify the acquired SERS signals and output the classification results of the samples.

2. The surface-enhanced Raman spectroscopy detection system based on gold nanoparticle-based microrobots according to claim 1, characterized in that, The magnetic spiral micro-nano robot is prepared by the following steps: Step S1, a spiral structure array is prepared by micron-level 3D printing technology. The printing material contains magnetic particles, image developer and photosensitive resin. After cleaning, a nickel layer and a platinum layer are deposited sequentially on the surface of the spiral structure to obtain the micro-robot. Step S2: The spiral array tip of the microrobot is immersed in the reaction solution containing the gold precursor by the micro-motion platform to achieve localized controllable growth of gold nanospikes; the reaction solution containing the gold precursor includes chloroauric acid, hydrogen peroxide and silver nitrate.

3. The surface-enhanced Raman spectroscopy detection system based on gold nanoparticle-based microrobots according to claim 2, characterized in that: In step S1, the 3D printing exposure parameters are 55 mW / cm², the exposure time is 20 s, the nickel layer thickness is 500 nm, and the printing material contains photosensitive resin, ethylene dimethacrylate, PVP, barium sulfate and neodymium iron boron magnetic particles; in step S2, the morphology is controlled by the immersion depth and reaction time.

4. The surface-enhanced Raman spectroscopy detection system based on gold nanoparticle-based microrobots according to claim 1, characterized in that: The magnetic field control module includes a three-axis Helmholtz electromagnetic coil. By adjusting the input current signal in real time, it changes the intensity, rotation frequency, spatial tilt angle, and orientation angle of the rotating magnetic field, so as to realize three-dimensional trajectory navigation, targeted positioning, and attitude control of the micro-nano robot within the channel of the microfluidic chip.

5. The surface-enhanced Raman spectroscopy detection system based on gold nanoparticle-based microrobots according to claim 1, characterized in that: The gold nanoparticles are grafted with Raman reporter molecular probes on their surface. When the surface-enhanced Raman spectroscopy detection based on gold nanoparticle micro-nano robots is used to detect hydrogen peroxide, the concentration of hydrogen peroxide is determined by detecting the change in the intensity ratio of the characteristic peaks of the Raman reporter molecular probes, thus achieving quantitative detection for non-diagnostic purposes.

6. The surface-enhanced Raman spectroscopy detection system based on gold nanoparticle-based microrobots according to claim 1, characterized in that: The preprocessing in the data processing and classification module includes: cosmic ray removal step: identifying cosmic ray interference peaks based on adaptive peak aspect ratio detection and half-peak width threshold, and repairing them using linear interpolation; baseline correction step: removing fluorescence background using an adaptive iterative reweighted penalized least squares method based on Cholesky decomposition optimization; and smoothing step: applying a Savitzky-Golay filter to smooth the signal. The structure of the one-dimensional convolutional neural network model specifically includes: three one-dimensional convolutional layers with output channels of 8, 64, and 64 respectively, each followed by a ReLU activation layer with a kernel size of 15×1; two one-dimensional max pooling layers with a pooling window and stride of 2; three fully connected layers with output dimensions of 512, 512, and K respectively, where K is the total number of classification categories; and a Softmax layer at the end of the network and a training module using the cross-entropy loss function.

7. The surface-enhanced Raman spectroscopy detection system based on gold nanoparticle-based microrobots according to claim 1, characterized in that: The data processing and classification module is deployed on a portable edge computing device or a local terminal that communicates with a cloud server, supporting offline operation; the one-dimensional convolutional neural network model is optimized by quantization or pruning to adapt to the computing resources of embedded devices.

8. A sample detection method based on surface-enhanced Raman spectroscopy of gold nanoparticles / microrobots, characterized in that, The surface-enhanced Raman spectroscopy (SERS) detection system based on gold nanoparticle-based microrobots as described in any one of claims 1 to 7 includes the following steps: Step S10: placing a magnetic spiral microrobot into a microfluidic chip containing the sample to be tested; Step S20: generating a controllable rotating magnetic field through an external magnetic field control system to drive the microrobot to perform three-dimensional trajectory movement within the chip channel and target the detection area; Step S30: using the gold nanoparticles grown at the tip of the microrobot as a surface-enhanced Raman spectroscopy substrate to perform in-situ sampling and signal enhancement of the markers to be tested in the detection area, and acquiring the corresponding SERS spectral signals; Step S40: preprocessing the acquired SERS spectral signals and inputting them into a pre-trained one-dimensional convolutional neural network model to output the classification results of the samples; wherein, the method is used for in vitro sample analysis for non-diagnostic purposes.

9. The sample detection method based on surface-enhanced Raman spectroscopy of gold nanoparticles / microrobots according to claim 8, characterized in that, In step S20, the input current of the triaxial Helmholtz electromagnetic coil device is adjusted in real time to change the strength, frequency, spatial tilt angle, and orientation angle of the rotating magnetic field, so as to achieve precise navigation and target positioning of the three-dimensional motion trajectory of the micro-nano robot; Raman reporter molecular probes are grafted onto the surface of the gold nanospikes; when the target marker is hydrogen peroxide, the concentration of hydrogen peroxide is determined by detecting the change ratio of the characteristic peak of the Raman reporter molecular probe.

10. The application of the sample detection method based on surface-enhanced Raman spectroscopy of gold nanoparticles microrobots as described in claim 8 or 9, characterized in that: Used for microfluidic in vitro sample analysis, laboratory detection of biomarkers, detection of environmental substances, or material analysis.

Images