SERS (Surface Enhanced Raman Scattering) optical fiber detection device for lung nodule diagnosis and use method thereof

Abstract

The invention discloses an SERS (Surface Enhanced Raman Scattering) optical fiber detection device for lung nodule diagnosis and a use method of the SERS optical fiber detection device. The SERS optical fiber detection device comprises an SERS optical fiber, a needle-shaped detection element, a Raman spectrometer and an upper computer, wherein the signal acquisition end of the SERS optical fiber is modified with a Raman scattering enhancement material, and the signal output end of the SERS optical fiber is coupled with a Raman spectrometer to form an SERS optical fiber-spectrometer system; the signal acquisition end of the SERS optical fiber is integrated with the needle-shaped detection element; the needle-shaped detection element is inserted into the pulmonary nodule tissue, so that the pulmonary nodule tissue is in contact with the signal acquisition end of the SERS optical fiber, and an obtained Raman scattering signal is transmitted to the spectrograph; and after the spectrometer converts the Raman scattering signal into digital spectrum data, the digital spectrum data are transmitted to an upper computer for pulmonary nodule typing diagnosis, and a diagnosis result is obtained. The method has the advantages of simple device, low cost, high sensitivity, high signal-to-noise ratio and the like, and is suitable for being applied to other intraoperative diagnosis scenes which need quick, real-time and non-marking detection and are difficult to contact.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Raman detection, in particular to a SERS optical fiber detection device for lung nodule diagnosis and a use method thereof. BACKGROUND

[0002] The incidence and mortality of lung cancer are the highest among malignant tumors, and the early diagnosis and treatment of lung cancer is the key to improving the survival rate of patients. The current diagnosis and treatment process of lung nodules usually first screens and locates through imaging examination, and then obtains tissue samples through puncture biopsy for pathological analysis. However, the specificity and accuracy of this method for malignant lesion determination are limited. As the gold standard for diagnosis, pathological biopsy has the problem of too long result time (usually 3-7 days), and the puncture operation may need multiple sampling, increasing the risk of complications such as pneumothorax and bleeding. Although a few hospitals can use rapid pathological section technology, each sample still needs to be processed for 20-30 minutes, greatly prolonging the duration of the operation.

[0003] Surface-enhanced Raman scattering (SERS) technology, as a fast and sensitive detection technology without the need for labeling, can provide the chemical fingerprint of complex biological samples, present the characteristic spectrum of different states of tissues and be used for cell classification and tissue typing, thus having the potential for intraoperative diagnosis of lung puncture operation and being able to provide tissue chemical fingerprint for typing analysis.

[0004] However, traditional Raman spectrum equipment is complex and expensive, requiring higher technical support. The collection speed is slow, and long-time integration is usually required. Oxidation and other irreversible changes may occur during the clinical detection process, which may affect the reliability of the detection results. Higher laser power is also required, which may cause damage to the lung nodules or fluorescence interference. SUMMARY

[0005] Therefore, the present application provides a SERS optical fiber detection device for lung nodule diagnosis and a use method thereof.

[0006] The first aspect of the present application provides a SERS optical fiber detection device for lung nodule diagnosis, comprising a SERS optical fiber, a needle-shaped detection element, a Raman spectrometer and an upper computer. The signal collection end of the SERS optical fiber is modified with a Raman scattering enhancement material, and the signal output end is coupled with the Raman spectrometer to form a SERS optical fiber-spectrometer system. The signal collection end of the SERS optical fiber is integrated with the needle-shaped detection element. By inserting the needle-shaped detection element into the lung nodule tissue, the lung nodule tissue is in contact with the signal collection end of the SERS optical fiber, and the Raman scattering signal is transmitted to the spectrometer. The spectrometer converts the Raman scattering signal into digital spectral data, and then transmits the digital spectral data to the host computer for lung nodule typing diagnosis to obtain a diagnosis result.

[0007] Further, the SERS optical fiber is a multimode optical fiber, and the multimode optical fiber includes a glass optical fiber and a quartz optical fiber; and the Raman scattering enhancement material is a gold-coated silver nanocube.

[0008] Further, the Raman spectrometer includes a laser, a CCD detector, and a fiber coupler; the fiber coupler is used for coupling with the SERS optical fiber; the laser is used for generating laser of a specific wavelength and emitting to the SERS optical fiber; and the CCD detector is used for collecting the Raman scattering signal transmitted by the SERS optical fiber and converting the Raman scattering into digital spectral data.

[0009] Further, the needle-shaped detection element includes a clamp-type needle cavity, a channel, and a needle tube tip; the channel is covered by the clamp-type needle cavity; the SERS optical fiber-spectrometer system is placed in the channel, and one end of the SERS optical fiber modified with the Raman scattering enhancement material is located inside the needle tube tip; by inserting the needle-shaped detection element into the lung nodule tissue, the lung nodule tissue is contacted with the detection end of the SERS optical fiber, and then the Raman scattering signal of the lung nodule tissue is collected.

[0010] Further, the needle-shaped detection element is an ablation needle; the SERS optical fiber is arranged in the channel of the ablation needle; and the ablation needle is used for ablation treatment of the lung nodule according to the diagnosis result of the lung nodule typing diagnosis.

[0011] Further, in the host computer, the lung nodule typing diagnosis includes the following steps: receiving digital spectral data, the digital spectral data being a mixed Raman spectral data containing a fiber background signal and a lung nodule tissue Raman signal; performing background signal identification and deduction on the mixed spectral data to obtain lung nodule tissue spectral data, i.e., sample Raman spectral data; performing Raman spectral denoising and identification on the sample Raman spectral data to obtain a Raman spectral feature vector; performing lung nodule typing and lesion degree evaluation according to the Raman spectral feature vector to obtain a diagnosis result.

[0012] Further, the background signal identification and deduction on the mixed spectral data specifically includes the following steps: obtaining fiber background Raman spectral data, the fiber background Raman spectral data being converted from the Raman scattering signal generated when the SERS optical fiber does not contact the lung nodule tissue; training a background signal deduction model according to the fiber background Raman spectrum data; inputting the mixed spectrum data into the background signal deduction model, identifying the characteristics of the background signal, and deducting the background signal from the mixed spectrum data according to the characteristics of the background signal to obtain sample spectrum data.

[0013] Further, the sample Raman spectrum data is subjected to Raman spectrum identification, specifically including the following steps: preprocessing the sample spectrum data to improve the data quality of the sample spectrum data; training a Raman spectrum identification model; inputting the sample Raman spectrum data into the Raman spectrum identification model, extracting high-dimensional feature data of the hidden layer through the Raman spectrum identification model, and outputting to obtain a Raman spectrum feature vector.

[0014] Further, the lung nodule typing and lesion degree evaluation according to the Raman spectrum feature vector specifically includes the following steps: inputting the Raman spectrum feature vector into the developed and trained lung nodule evaluation model; performing lung nodule typing and lesion degree evaluation in the lung nodule evaluation model, and outputting the classification probability of each type and the lesion degree score; determining the type with the largest probability as the diagnostic result of lung nodule typing, and determining the diagnostic result of lung nodule lesion degree according to the preset scoring rule.

[0015] Another aspect of the present application provides a use method of a SERS optical fiber detection device for lung nodule diagnosis, including the following steps: modifying a Raman scattering enhancement material at the signal collection end of the SERS optical fiber, coupling the signal output end of the SERS optical fiber with a Raman spectrometer to form a SERS optical fiber-spectrometer system; integrating the signal collection end of the SERS optical fiber-spectrometer system with a needle-shaped detection element; by inserting the needle-shaped detection element into the lung nodule tissue, the lung nodule tissue is in contact with the signal collection end of the SERS optical fiber, and the Raman scattering signal is transmitted to the spectrometer; performing lung nodule typing diagnosis through the upper computer to obtain a diagnostic result.

[0016] The embodiment of the application has the following beneficial effects: the SERS optical fiber detection device for lung nodule diagnosis and the use method thereof utilize an optical fiber as an optical path and a Raman signal transmission medium, have the advantages of remote analysis capability, strong adaptability, and small probe size, are suitable for application in intraoperative diagnosis scenes that require rapid, real-time, label-free detection and other difficult-to-contact scenes, do not need to puncture and take out lung nodule tissue biopsy, can collect high signal-to-noise ratio Raman spectra of lung nodules in vivo through the SERS optical fiber in a rapid and real-time manner, detect full-spectrum information to comprehensively analyze the typing of the tissue, simultaneously perform real-time data analysis through an artificial intelligence model to complete real-time and accurate diagnosis of the benignity and malignancy of the lung nodule, and design and manufacture the SERS optical fiber detection device that can be compatible with an ablation needle, so as to guide the performance of a minimally invasive lung nodule ablation operation. The application has the advantages of simplicity, low cost, high sensitivity, and high signal-to-noise ratio, and has good universality and high compatibility.

[0017] Additional aspects and advantages of the application will be set forth in the description that follows, and in part will be obvious from the description, or can be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0019] Figure 1 is a basic structure schematic diagram of the SERS optical fiber detection device for lung nodule diagnosis of the application; Figure 2 is a physical structure schematic diagram of the SERS optical fiber detection device for lung nodule diagnosis of the application; Figure 3 is a step flow schematic diagram of lung nodule typing diagnosis performed by the host computer in the application; Figure 4 is a Raman spectrum data difference data graph of normal lung tissue and diseased lung nodules; Figure 5 is a Raman spectrum data graph of normal lung nodules detected by using a common optical fiber; Figure 6 is a Raman spectrum data graph of normal lung nodules detected by using the SERS optical fiber detection device for lung nodule diagnosis of the application; Figure 7 is a SERS optical fiber Raman spectrum feature vector difference data graph of normal lung tissue and diseased lung nodules; Figure 8is a summary receiver operating characteristic (ROC) curve plot of the lung nodule assessment model under five-fold cross-validation.

[0020] Figure 9 is a summary classification confusion matrix plot of the lung nodule assessment model under five-fold cross-validation.

[0021] Figure 10 is a Grad-CAM feature attention region plot of the lung nodule assessment model.

[0022] Figure 11 is a statistical significant difference region (Mann-Whitney U test) analysis plot of the optical fiber Raman spectrum feature vector.

[0023] Figure 12 is a prediction confidence distribution histogram plot of the lung nodule assessment model under five-fold cross-validation.

[0024] Reference signs: 1-SERS optical fiber and / or ablation needle, 2-clamped needle body cavity, 3-channel, 4-shaft, 5-needle tube tip. DETAILED DESCRIPTION

[0025] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0026] Optical fiber Raman spectrum detection uses optical fiber as the light path and Raman signal transmission medium, has the advantages of remote analysis capability, strong adaptability, and small probe size, and is suitable for application in intraoperative diagnosis scenes that require rapid, real-time, label-free detection and other difficult-to-access scenes. Optical fiber Raman sensing technology has been gradually applied to clinical intraoperative detection, and the advantages of remote signal transmission, small volume, and flexible travel of optical fiber provide more scenes and solutions for intraoperative diagnosis.

[0027] However, the existing Raman spectrum technology using optical fiber Raman detection of lung tissue has low sensitivity and poor accuracy. Due to the design principle of the system, which is to collect the spontaneous Raman scattering signal of the tissue, and the attenuation and loss of laser energy in the integration process of the distribution of optical elements and the optical fiber probe, the collected tissue Raman spectrum characteristic peaks are few and the signal-to-noise ratio is low, resulting in certain information loss and increased difficulty in spectral data processing and analysis.

[0028] In addition, the existing Raman spectrum technology using optical fiber Raman detection of lung tissue usually only has a detection function and does not realize the development and expansion of further treatment.

[0029] In view of these shortcomings, the embodiment of the present application provides a SERS optical fiber detection device for lung nodule diagnosis. As shown inFigure 1 As shown in the figure, the device comprises a SERS optical fiber, a needle-shaped detection element, a Raman spectrometer and a host computer; The signal collection end of the SERS optical fiber is decorated with a Raman scattering enhancement material, and the signal output end is coupled with the Raman spectrometer to form a SERS optical fiber-spectrometer system; The signal collection end of the SERS optical fiber of the SERS optical fiber-spectrometer system is integrated with the needle-shaped detection element; by inserting the needle-shaped detection element into the lung nodule tissue, the lung nodule tissue is brought into contact with the signal collection end of the SERS optical fiber, and the Raman scattering signal is transmitted to the spectrometer; after the spectrometer converts the Raman scattering signal into digital spectral data, the digital spectral data are transmitted to the host computer for lung nodule typing diagnosis to obtain a diagnosis result.

[0030] The physical structure diagram of the embodiment of the present application is as shown in the figure Figure 2 The provided Raman fiber detection device has the characteristics of simple device, low cost, high sensitivity and signal-to-noise ratio. Compared with directly using ordinary optical fiber, using the SERS optical fiber of the present application to detect lung nodules, and through the process of detecting lung nodules by fiber Raman spectrum, a Raman spectrum with higher signal-to-noise ratio can be collected, which significantly improves the accuracy and reliability of diagnosis.

[0031] The implementation process of each part of the present application will be described in detail below: SERS optical fiber: the SERS optical fiber is an elongated light guide wire, and contacting the lung nodule with the SERS optical fiber can obtain high signal-to-noise ratio Raman spectrum of the nodule for subsequent analysis. In the embodiment of the present application, the SERS optical fiber is a multimode optical fiber, which includes glass optical fiber and quartz optical fiber, etc. In order to enhance the Raman spectrum collection effect, the signal collection end of the SERS optical fiber is decorated with a Raman scattering enhancement material, which can be gold-coated silver nanocubes (Ag@AuNCs) or the like. After decorating the Raman scattering enhancement material, the SERS optical fiber can enhance the Raman signal through the localized surface plasmon resonance (LSPR) effect of the metal nanostructure of the Raman scattering enhancement material. When the excitation laser irradiates the surface of the metal nanoparticles, LSPR will be excited, and a strong local electric field enhancement will be generated. This electric field enhancement can significantly amplify the Raman scattering signal of the nearby molecules (such as proteins, nucleic acids and lipids in the lung nodule tissue), thereby improving the sensitivity and signal-to-noise ratio of the detection. Through the design of the Raman scattering enhancement material, the SERS optical fiber of the present application can collect Raman spectrum with higher signal-to-noise ratio, which significantly improves the accuracy and reliability of diagnosis.

[0032] Raman spectrometer: the Raman spectrometer in the embodiment of the application comprises a laser, a CCD detector and a fiber coupler. The fiber coupler is used for coupling with the SERS optical fiber, serving as a light path hub. On one hand, the fiber coupler efficiently couples the excitation light from the laser to the SERS optical fiber; on the other hand, the fiber coupler can separate the Raman scattering signals collected by the SERS optical fiber and guide the Raman scattering signals to the CCD detector. In some embodiments, the fiber coupler adopts a filter or a circulator for splitting light, so as to ensure that the excitation light and the Raman signals are bidirectionally transmitted in the single SERS optical fiber without interfering with each other.

[0033] The laser is used for generating laser light of a specific wavelength (such as 785 nm) as an excitation light source. The laser power is usually set to 20-30 mW, so as to avoid tissue thermal damage and reduce fluorescence interference. The laser is transmitted to the fiber coupler through the excitation optical fiber and is finally guided to the end of the SERS optical fiber, so as to excite the lung nodule tissue to generate Raman scattering signals. The CCD detector is used for collecting the Raman scattering signals transmitted by the SERS optical fiber and converting the Raman scattering into a spectrum with intensity varying with Raman shift, as digital spectrum data.

[0034] In the embodiment of the application, the exposure time of the Raman spectrometer is set to 50-200 ms, the integral times are set to 20-50 times, and the laser power is set to 20-30 mW. After collecting the dark current, the Raman spectrometer starts to collect the Raman spectrum of the lung nodule and obtains digital spectrum data, which is transmitted to the upper computer.

[0035] Needle-shaped detection element: the needle-shaped detection element in the embodiment of the application comprises a clamp-type needle body cavity, a channel and a needle tube tip; the channel is covered by the clamp-type needle body cavity; the SERS optical fiber-spectrometer system is placed in the channel, and one end of the SERS optical fiber modified with a Raman scattering enhancement material is located inside the needle tube tip; after the needle tube tip is inserted into the lung nodule tissue, the Raman scattering signals of the lung nodule tissue are collected by inserting the needle-shaped detection element into the lung nodule tissue, so that the lung nodule tissue is in contact with the signal collection end of the SERS optical fiber. The needle-shaped detection element can be an ablation needle or other medical device.

[0036] Since the SERS optical fiber has poor safety in operation, it is easy to be broken during the lung nodule tissue puncture process, resulting in damage of the optical fiber or failure of detection. In the embodiment of the application, the SERS optical fiber is protected by the design of the needle tube tip. After the prepared SERS optical fiber is placed in the channel of the needle-shaped detection element, the needle tube tip is inserted into the lung nodule tissue. The needle tube tip is located at the front end of the needle-shaped detection element and has an open design, so as to facilitate insertion into the tissue and allow the optical fiber to be in contact with the tissue; the tip is designed to be blunt, so as to reduce the puncture damage to the lung nodule.

[0037] The clamp-type needle cavity is designed as an openable clamping structure, consisting of two semi-circular or symmetrical parts that enclose the channel and close via a mechanical structure (such as snaps or threads) to secure and protect the SERS fiber. The internal design of the cavity allows some lung nodule tissue to come into contact with the fiber optic signal acquisition end during the insertion of the needle-shaped detection element into the lung nodule tissue. This reduces mechanical stress on the fiber, lowering the risk of breakage; it also further ensures that the special material on the fiber surface is not damaged or left on the tissue. Furthermore, it enhances the contact effect between the lung nodule and the lung nodule sample within the device.

[0038] The embodiments of the present invention, through the design of the needle-shaped structure, can adapt to the actual operating environment of medical devices, improving the convenience and accuracy of testing; through the design of the channel and clamp-type needle cavity, the optical fiber is effectively protected from damage during operation, further ensuring the reliability of the test results.

[0039] In some embodiments, the needle-shaped detection element is an ablation needle; a SERS optical fiber is disposed in the channel of the ablation needle; the ablation needle is used to perform ablation treatment on the lung nodules based on the diagnostic results of the lung nodule classification diagnosis.

[0040] In this embodiment, the design of the ablation needle enables the Raman fiber optic detection device of this invention to achieve integrated diagnosis and treatment. After acquiring Raman spectra through SERS fiber, computer software analyzes and outputs diagnostic results (such as benign or malignant) in real time. If the diagnosis is malignant, the operator can immediately insert the ablation needle for ablation treatment without the need for a secondary surgery.

[0041] Host computer: such as Figure 3 As shown, the process of diagnosing lung nodules in the host computer includes the following steps: S1. Receive digital spectral data, which is a hybrid Raman spectral data containing fiber background signal and lung nodule tissue Raman signal; S2. Background signal identification and subtraction are performed on the mixed spectral data to obtain the spectral data of lung nodule tissue, i.e., the sample Raman spectral data; S3. Perform Raman spectral denoising and identification on the sample Raman spectral data to obtain the Raman spectral feature vector; S4. Based on the Raman spectral characteristic vectors, lung nodules are classified and the degree of lesion is assessed to obtain diagnostic results.

[0042] In this embodiment of the invention, a host computer performs real-time data analysis to achieve real-time and accurate diagnosis of the benign or malignant nature of lung nodules, thereby guiding the minimally invasive lung nodule ablation surgery.

[0043] like Figure 4The diagram illustrates the differences in Raman spectral data between normal lung tissue and diseased lung nodules. It shows that the Raman spectra of normal lung tissue and diseased lung nodules differ in the position, width, and intensity of characteristic peaks. These differences primarily stem from the fact that Raman spectroscopy can reflect characteristic spectra that indicate disease progression in tissues, such as changes at the molecular level of nucleic acids, lipids, and proteins. Therefore, Raman spectroscopy can be used to detect whether lung nodules have become diseased.

[0044] The following details the implementation process for each step in the diagnosis of pulmonary nodules: S1. Receives digital spectral data, specifically a hybrid Raman spectral data, which includes fiber background signal and lung nodule tissue Raman signal.

[0045] A needle-shaped detection element is inserted into the lung nodule tissue, bringing the tissue into contact with the signal acquisition end of the SERS fiber optic cable. The spectrometer exposure time is set to 50-200 ms, the integration count to 20-50, and the laser power to 20-30 mW. After acquiring the dark current three times and confirming the spectrometer system is functioning correctly, Raman spectra of the lung nodule can be acquired. In this embodiment of the invention, setting the spectrometer exposure time to 200 ms, the integration count to 50, and the laser power to 30 mW yields the best Raman spectral signal. The Raman spectral data obtained when detecting lung nodule tissue is a mixed spectrum, containing both the fiber background signal and the Raman spectrum of the lung nodule tissue.

[0046] S2. Background signal identification and subtraction are performed on the mixed spectral data to obtain sample spectral data.

[0047] Figure 5 It involves directly using ordinary optical fibers to detect the Raman spectrum of normal lung tissue, particularly in the 500-1300 cm⁻¹ range. -1 The inverted peaks are due to the large real-time fluctuations in the fiber optic background signal and the weak sample signal with a low signal-to-noise ratio in the mixed spectrum during detection. This causes classical fiber background subtraction methods (such as least squares and polynomial fitting) to fail to accurately distinguish between the fiber optic background signal and the Raman signal from lung nodule tissue, leading to inaccurate fiber optic background signal identification and subtraction. When the sample signal is weak, the fluctuations in the fiber optic background have a significant impact on the spectrum, which is a problem that still exists when fiber Raman spectroscopy is directly applied to biological tissue detection. Therefore, methods such as enhancing the Raman scattering signal of the sample during detection and developing and training a real-time accurate fiber optic background subtraction model can effectively reduce the errors caused by fluctuations in the fiber optic background signal.

[0048] S2. Background signal identification and subtraction are performed on the mixed spectral data to obtain the spectral data of lung nodule tissue, i.e., the sample Raman spectral data.

[0049] In step S2, background signal identification and subtraction are performed on the mixed spectral data, specifically including the following steps: S2-1. Acquire the background Raman spectrum data of the optical fiber. The background Raman spectrum data of the optical fiber is obtained by converting the Raman scattering signal generated when the SERS optical fiber is not in contact with the lung nodule tissue. This step involves acquiring the Raman signal generated only by the optical fiber itself in the absence of a sample (e.g., placing the SERS optical fiber in a sample standard reference such as artificial tissue fluid, using the same parameters as the detection: laser power 20-30 mW, integration times 20-50, exposure time 50-200 ms).

[0050] S2-2. Training the background signal model based on fiber background Raman spectral data. More than 1300 pure fiber background signals were obtained through at least 1300 dark current acquisitions. These signals were used to train a deep learning-based algorithm, resulting in a model that can accurately identify and subtract the fiber background signal.

[0051] As a specific implementation, a U-net cascaded model based on convolutional neural networks can be used. The background signal model used in this embodiment of the invention is UNet_CA7, which has an encoder-decoder structure and includes a downsampling path (contraction path) and an upsampling path (expansion path). The model extracts high-dimensional features through 7 layers of downsampling and utilizes the CBAM attention mechanism to enhance the model's ability to capture key spectral features.

[0052] S2-3. Input the mixed spectral data into the background signal subtraction model, identify the characteristics of the background signal, and perform background signal subtraction on the mixed spectral data according to the characteristics of the background signal to obtain sample spectral data.

[0053] In this step, the background signal subtraction model can be applied to the developed software to identify and subtract the fiber background signal in the mixed spectrum in real time, thereby obtaining data containing only the lung nodule tissue spectrum. Specifically, in this step, the background signal subtraction model predicts and fits the fiber background curve of the mixed spectrum based on the background signal characteristics in each mixed spectrum, and then subtracts the curve from the mixed spectral data, so that the sample spectral data after subtraction retains only the lung nodule-related signal, improving the signal-to-noise ratio of the spectral data.

[0054] Raman spectra of normal lung nodules after background signal subtraction using embodiments of the present invention are shown below. Figure 6 As shown, its signal-to-noise ratio is significantly improved compared to the Raman spectrum detected using ordinary optical fibers.

[0055] S3. Perform Raman spectral identification on the sample spectral data to obtain the Raman spectral feature vector.

[0056] In step S3, Raman spectroscopy identification is performed on the sample spectral data, which specifically includes the following steps: S3-1. Preprocess the sample spectral data to improve the data quality.

[0057] As a preferred embodiment, the preprocessing methods include smoothing, baseline correction, and normalization. Smoothing reduces random noise and avoids spurious peaks using Savitzky-Golay (SG) filtering; baseline correction further corrects the tissue autofluorescence background using polynomial fitting or asymmetric least squares; and normalization uses a standard score (Z-score) to eliminate the influence of intensity fluctuations caused by fiber contact angle and instantaneous laser power fluctuations.

[0058] S3-2. Training the Raman spectroscopy recognition model; The Raman spectroscopy recognition model constructed in this embodiment of the invention is a one-dimensional convolutional neural network (1D-CNN) model. The model structure includes, in sequence: Input layer: used to receive one-dimensional spectral vectors; Feature extraction module: Composed of multiple alternating one-dimensional convolutional layers and max pooling layers, used to extract local waveform features of the spectrum layer by layer and reduce dimensionality; Feature embedding layer: Located before the output layer, it consists of multiple neurons, such as 128 or 256, and serves as a convergence layer for high-dimensional features; Output layer: Used for classification or regression output during training.

[0059] The Raman spectroscopy recognition model uses a sample set with known labels to supervise the training of the model, and updates the network weights through the backpropagation algorithm until the loss function converges.

[0060] S3-3. Input the sample Raman spectrum data into the Raman spectroscopy recognition model, extract the high-dimensional feature data of the hidden layer through the Raman spectroscopy recognition model, and output the Raman spectral feature vector.

[0061] In this embodiment of the invention, the sample Raman spectral data is input into a trained 1D-CNN model for forward propagation calculation. After passing through convolutional and pooling layers in the 1D-CNN model, the sample Raman spectral data enters the feature embedding layer. The output data of the nodes in the feature embedding layer before the output layer is truncated; finally, the array composed of the truncated output values ​​from the feature embedding layer of the Raman spectral recognition model is defined as the sample's Raman spectral feature vector.

[0062] In another embodiment, Raman spectral feature vectors can also be obtained through peak detection.

[0063] Specifically, peak detection is first performed on the sample spectral data to identify the maxima points of the sample spectral data, which are then used as primary characteristic peaks. Preferably, the peak detection algorithm includes local maximum detection, derivative method, and threshold method. This embodiment of the invention uses the derivative method to calculate the first derivative of the spectrum and determine the peak position based on the zero-crossing point of the derivative (slope change); the second derivative is used to distinguish between true peaks and shoulder peaks, and the calculated primary characteristic peaks include the peak position (Raman shift) and peak height (intensity) of each peak.

[0064] Secondly, curve fitting is performed on the initial characteristic peaks to obtain secondary characteristic peaks. The purpose of curve fitting is to accurately calculate the parameters of each characteristic peak. In this embodiment of the invention, Gaussian fitting is used, which is expressed by the following formula: ;in x 0 represents the peak position. I 0 represents the peak height. σ The peak width is calculated, and the peak area can be further calculated based on the peak height and peak width. Finally, the precise peak position, peak height, peak width, and peak area are obtained through fitting, serving as secondary characteristic peaks.

[0065] Then, the secondary characteristic peaks are matched with their position, shape, and width using the pre-stored Raman characteristic peak data in the database to determine the chemical substances corresponding to the secondary characteristic peaks.

[0066] like Figure 7 The image shows the difference in SERS fiber Raman spectral characteristic vectors between normal lung tissue and diseased lung nodule tissue. The main characteristic peaks for the two lung tissues are shown in the table below, where the common characteristic peaks differ in intensity and peak width:

[0067] The pre-stored Raman characteristic peak data are shown in the table above. The chemical substance corresponding to each peak is identified by matching the fitted peaks with the reference table.

[0068] Finally, the characteristic peak data is output as the Raman spectral eigenvector. The output characteristic peak data is presented in structured data form, including the Raman shift, intensity, peak width, corresponding chemical substance identifier, and data confidence level of the characteristic peak.

[0069] S4. Based on the Raman spectral characteristic vectors, lung nodules are classified and the degree of lesion is assessed to obtain diagnostic results.

[0070] In step S4, lung nodule classification and lesion severity assessment are performed based on Raman spectral feature vectors, specifically including the following steps: S4-1. Input the Raman spectral eigenvectors into the developed and trained lung nodule assessment model; S4-2. In the lung nodule assessment model, lung nodule classification and lesion severity assessment are performed, and the classification probability and lesion severity score for each classification are output; S4-3. The classification with the highest probability is determined as the diagnostic result of the lung nodule classification, and the diagnostic result of the degree of lung nodule lesion is determined according to the preset scoring rules.

[0071] In this embodiment of the invention, pulmonary nodule classification and lesion severity assessment are achieved through a machine learning model. The structure of the machine learning model can be selected from support vector machines, random forests, neural networks, etc. Raman spectroscopy feature vectors are input into the pulmonary nodule assessment model, and the model outputs classification results as classification probabilities (such as normal, benign, malignant), and lesion severity assessment results as scores (such as numerical values / grades reflecting the invasiveness or evolution stage of the lesion, used to assist in judging the severity of the lesion).

[0072] As a preferred embodiment, the lung nodule assessment model is trained through the following steps: S4-0-1. Establish a machine learning model as the initial lung nodule assessment model; S4-0-2. Obtain historical feature vectors from historical diagnosis records of lung nodules; establish a feature dataset based on historical feature vectors; S4-0-3. Label the historical feature vectors in the feature dataset, including lung nodule classification and lesion severity; S4-0-4. Divide the feature dataset into a training dataset, a validation dataset, and a test dataset; use the training dataset to train the machine learning model, and use the validation dataset to evaluate the performance of the machine learning model; optimize the hyperparameters of the machine learning model based on the validation results. S4-0-5. Use the test dataset to calculate the accuracy, sensitivity, and specificity of the machine learning model's output. When the accuracy of the machine learning model's output reaches a preset threshold, the machine learning model is used as the preset lung nodule assessment model.

[0073] In this embodiment of the invention, each historical feature vector is labeled as "normal," "benign lesion," or "malignant lesion," and includes indicators of the degree of malignancy (such as atypical adenomatous hyperplasia (AAH) and adenocarcinoma in situ (AIS), with adenocarcinoma including minimally invasive adenocarcinoma (MIA) and invasive adenocarcinoma (IA)). The feature dataset is presented as structured data, with each row representing a historical feature vector and columns including feature peak parameters and labels. The model is trained using the training set, and hyperparameters (such as the kernel function of SVM, the number of trees in a random forest, the number of hidden layers in a neural network, or the learning rate) are optimized and adjusted based on the validation set results. Finally, metrics (such as accuracy, recall, and F1 score) are calculated using the test set to ensure the model's generalization ability.

[0074] This invention utilizes a machine learning model for lung nodule classification and lesion severity assessment. It effectively integrates multiple characteristic peak parameters (such as peak intensity, peak width, and ratio) and captures continuous changes in lesion severity (such as benign / malignant probability scores) through nonlinear mapping, thereby improving classification reliability. The machine learning model significantly enhances diagnostic accuracy and efficiency. Through training, the model learns Raman spectral characteristic peaks, automatically identifies subtle differences between normal and diseased tissues, and performs real-time analysis, meeting the needs for rapid intraoperative decision-making.

[0075] To illustrate the effectiveness, robustness, and generalization ability of the lung nodule assessment model in step S4 of this invention in practical applications, the following provides a specific example of the preferred model based on a one-dimensional convolutional neural network (1D-CNN) using a 5-fold cross-validation strategy to illustrate the method of this invention.

[0076] Dataset Construction and Preprocessing: The example uses Raman spectral data of clinically collected lung nodules (covering adenocarcinoma in situ, minimally invasive adenocarcinoma, and invasive adenocarcinoma) and paired normal tissues. Using the S1-S3 steps, a dataset containing 875 sample spectral feature vectors was constructed, with a feature length L=530. To simulate a real clinical testing environment and prevent model overfitting, data augmentation techniques were employed during the training phase, including random shifting and adding Gaussian noise to increase the diversity of the training samples.

[0077] Model Construction and Validation Strategy: This example uses a one-dimensional convolutional neural network (1D-CNN) as the lung nodule assessment model. The model structure includes multiple convolutional layers to extract deep features from Raman spectroscopy, and introduces a Dropout layer to improve the model's generalization ability. Finally, a Softmax layer outputs the classification probability. The validation process employs a five-fold cross-validation method: the dataset is randomly divided into five equal parts, with four parts selected alternately as the training set and the remaining part as the test set. Five rounds of independent training and evaluation are performed, and the average of the five test results is taken as the performance metric. This validation method effectively avoids evaluation bias caused by the randomness of data partitioning, proving that the model has stable diagnostic capabilities under different data distributions.

[0078] The model proposed in this invention exhibits extremely high diagnostic efficacy in five-fold cross-validation. For example... Figure 8 As shown in the receiver operating characteristic (ROC) curve, the area under the ROC curve (AUC) for all test folds is as high as 0.963. This metric is significantly close to 1.0, indicating that the model has a very strong ability to distinguish between benign and malignant tumors. Figure 9 As shown in the confusion matrix, the model's average classification accuracy reached 89.60% (standard deviation ± 2.90%). The low standard deviation confirms the model's excellent stability and its resilience to sample fluctuations. In the summarized test results, the model's average sensitivity for identifying lung nodules reached 91.40%. This means that the method in this embodiment can identify minute early lesions with a very high probability, significantly reducing the risk of missed diagnoses during surgery and meeting the high safety requirements for rapid intraoperative decision-making. Figure 10 In the 1D-CNN model, high-weight spectral regions (Grad-CAM peak positions) and Figure 11 The regions of statistical significance (P-value < 0.05) calculated by the Mann-Whitney U test highly overlap. This not only verifies that the Raman spectral feature vector contains a specific biochemical fingerprint that can distinguish between normal and diseased tissues, but also confirms that the AI ​​model of this invention successfully captured and utilized these key biological features for accurate typing. At the same time, it proves that the machine learning model makes judgments based on real biophysical features, rather than learning random noise. Figure 12 The predicted malignancy probability of truly normal samples (blue) is highly concentrated in the 0 range, while the predicted malignancy probability of truly malignant samples (red) is highly concentrated in the 1 range, with very few samples in the intermediate ambiguous region (0.2-0.8). These experimental results demonstrate that the model of this invention can provide a high-confidence, clear score for the benign and malignant nature of pulmonary nodules, exhibiting excellent discrimination and significantly reducing uncertainty in clinical diagnosis.

[0079] After the lesion severity assessment is completed, since the embodiments of the present invention can integrate the SERS fiber into the ablation needle, the ablation needle can be easily advanced into the nodule after the diagnosis is completed without the need to insert another ablation needle. This allows for detection and treatment to be completed with only a single insertion into the lung nodule, and has good broad applicability and high compatibility.

[0080] The second embodiment of the present invention discloses a method for using a SERS fiber optic detection device for the diagnosis of pulmonary nodules, comprising the following steps: A Raman scattering enhancement material is modified at the signal acquisition end of the SERS fiber, and the signal output end of the SERS fiber is coupled to a Raman spectrometer to form a SERS fiber-spectrometer system. The signal acquisition end of the SERS optical fiber is integrated with the needle-shaped detection element; By inserting a needle-shaped detection element into the lung nodule tissue, the lung nodule tissue comes into contact with the signal acquisition end of the SERS fiber, and the Raman scattering signal is transmitted to the spectrometer. The diagnosis of lung nodules was obtained by using a host computer for classification.

[0081] Those skilled in the art will understand that modules in the devices of the embodiments of the present invention can be adaptively modified and placed in one or more devices different from those embodiments. Modules, units, or components in the embodiments of the present invention can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the corresponding claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the corresponding claims, abstract, and drawings) may be replaced by an alternative feature serving the same, equivalent, or similar purpose.

[0082] Through the above description of the embodiments, those skilled in the art will understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0083] Furthermore, the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. In particular, for embodiments such as apparatus and devices, since they are basically similar to the method embodiments, the relevant parts can be referred to the description of the method embodiments. The apparatus, devices, and other embodiments described above are merely illustrative, and the modules, units, etc., described as separate components may or may not be physically separate, that is, they may be located in one place or distributed in multiple places, such as nodes in a system network. Specifically, some or all of the modules and units can be selected according to actual needs to achieve the purpose of the above-described embodiment solutions. Those skilled in the art can understand and implement this without creative effort.

[0084] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0085] Furthermore, the terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying relative importance, or implicitly specifying the number of technical features indicated in this embodiment. Therefore, features defined with terms such as "first" and "second" in the embodiments of this invention can explicitly or implicitly indicate that the embodiment includes at least one of those features. In the description of this invention, the word "multiple" means at least two or more, such as two, three, four, etc., unless otherwise explicitly specified in the embodiments.

[0086] In embodiments of the present invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, components, features, and elements with the same names in different embodiments of the present invention may have the same meaning or different meanings, the specific meaning of which must be determined by its interpretation in that specific embodiment or further in conjunction with the context of that specific embodiment.

[0087] Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention. Other embodiments of the present invention will readily conceive of by considering the specification and practicing the invention. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

Claims

1. A SERS fiber optic detection device for diagnosing pulmonary nodules, characterized in that, Includes SERS fiber optic cable, needle-shaped detection element, Raman spectrometer, and host computer; The signal acquisition end of the SERS fiber is modified with Raman scattering enhancement material, and the signal output end is coupled to the Raman spectrometer to form a SERS fiber-spectrometer system. The signal acquisition end of the SERS optical fiber is integrated with the needle-shaped detection element; by inserting the needle-shaped detection element into the lung nodule tissue, the lung nodule tissue comes into contact with the signal acquisition end of the SERS optical fiber, and the Raman scattering signal is transmitted to the spectrometer. The spectrometer converts the Raman scattering signal into digital spectral data, and then transmits the digital spectral data to the host computer for lung nodule classification and diagnosis to obtain the diagnostic results.

2. The SERS fiber optic detection device for diagnosing pulmonary nodules according to claim 1, characterized in that, The SERS fiber is a multimode fiber, which includes glass fiber and quartz fiber; the Raman scattering enhancement material is gold-coated silver nanocubes.

3. The SERS fiber optic detection device for diagnosing pulmonary nodules according to claim 1, characterized in that, The Raman spectrometer includes a laser, a CCD detector, and an optical fiber coupler; the optical fiber coupler is used to couple with the SERS optical fiber; the laser is used to generate laser light of a specific wavelength and emit it into the SERS optical fiber; the CCD detector is used to collect the Raman scattering signal transmitted through the SERS optical fiber and convert the Raman scattering into digital spectral data.

4. The SERS fiber optic detection device for diagnosing pulmonary nodules according to claim 1, characterized in that, The needle-shaped detection element includes a clamp-type needle body cavity, a channel, and a needle tip; the channel is covered by the clamp-type needle body cavity; the SERS fiber-optic spectrometer system is placed in the channel, and one end of the SERS fiber modified with Raman scattering enhancement material is located inside the needle tip; By inserting a needle-shaped detection element into the lung nodule tissue, the lung nodule tissue comes into contact with the detection end of the SERS optical fiber, thereby acquiring the Raman scattering signal of the lung nodule tissue.

5. A SERS fiber optic detection device for diagnosing pulmonary nodules according to claim 1, characterized in that, The needle-shaped detection element is an ablation needle; the SERS optical fiber is disposed in the channel of the ablation needle; the ablation needle is used to perform ablation treatment on lung nodules according to the diagnostic results of lung nodule classification.

6. The SERS fiber optic detection device for diagnosing pulmonary nodules according to claim 1, characterized in that, The lung nodule classification diagnosis is performed in the host computer, including the following steps: Receive digital spectral data, which is a hybrid Raman spectral data containing fiber background signal and lung nodule tissue Raman signal; Background signal identification and subtraction are performed on the mixed spectral data to obtain spectral data of lung nodule tissue, i.e., sample Raman spectral data; Raman spectral denoising and identification are performed on the Raman spectral data of the sample to obtain Raman spectral feature vectors; Based on the Raman spectral feature vectors, lung nodules are classified and the degree of lesion is assessed to obtain diagnostic results.

7. A SERS fiber optic detection device for diagnosing pulmonary nodules according to claim 6, characterized in that, Background signal identification and subtraction are performed on the mixed spectral data, specifically including the following steps: Acquire fiber background Raman spectral data, which is obtained by converting the Raman scattering signal generated when the SERS fiber is not in contact with lung nodule tissue; A background signal subtraction model is trained based on the aforementioned fiber background Raman spectral data; The mixed spectral data is input into the background signal subtraction model to identify the characteristics of the background signal, and the background signal is subtracted from the mixed spectral data according to the characteristics of the background signal to obtain the sample spectral data.

8. A SERS fiber optic detection device for diagnosing pulmonary nodules according to claim 6, characterized in that, Raman spectral identification of the sample Raman spectral data specifically includes the following steps: Preprocessing of sample spectral data improves the data quality. Training a Raman spectroscopy recognition model; The sample Raman spectral data is input into the Raman spectral recognition model, and the high-dimensional feature data of the hidden layer is extracted through the Raman spectral recognition model to output the Raman spectral feature vector.

9. A SERS fiber optic detection device for diagnosing pulmonary nodules according to claim 6, characterized in that, The process of classifying lung nodules and assessing lesion severity based on the Raman spectral feature vectors specifically includes the following steps: The Raman spectral feature vectors are input into the developed and trained lung nodule assessment model; The lung nodule assessment model is used to classify lung nodules and assess the severity of lesions, and outputs the classification probability and severity score for each classification. The classification with the highest probability is determined as the diagnostic result of the pulmonary nodule classification, and the diagnostic result of the degree of pulmonary nodule lesion is determined according to the preset scoring rules.

10. A method of using a SERS fiber optic detection device for diagnosing pulmonary nodules, characterized in that, Includes the following steps: A Raman scattering enhancement material is modified at the signal acquisition end of the SERS fiber, and the signal output end of the SERS fiber is coupled to a Raman spectrometer to form a SERS fiber-spectrometer system. The signal acquisition end of the SERS fiber optic spectrometer system is integrated with the needle-shaped detection element; By inserting a needle-shaped detection element into the lung nodule tissue, the lung nodule tissue comes into contact with the signal acquisition end of the SERS fiber, and the Raman scattering signal is transmitted to the spectrometer. The diagnosis of pulmonary nodules was obtained by using a host computer for classification.

Images