Activatable head and neck cancer diagnosis and treatment integrated nanoprobe as well as preparation method and application thereof

Abstract

The invention provides an activatable head and neck cancer diagnosis and treatment integrated nanoprobe and a preparation method and application thereof.The preparation method includes the steps that according to an original data matrix, an adaptive spectrum correction algorithm is applied to eliminate background noise of an instrument, and when the relative fluctuation of spectral intensity exceeds a preset threshold value by 5%, a sliding window filtering technology is adopted, the window size is 5 nanometers, and the spectrum intensity is obtained; generating a de-noised spectral data set; measuring the pH value of the head and neck cancer tissue microenvironment by using a pH value sensor, recording the emission peak position change of the near-infrared quantum dots when the pH value is less than 6.5 in combination with the de-noising spectral data set, analyzing the frequency band displacement by adopting a Gaussian fitting algorithm, and generating a mapping data set of the pH value and the frequency band displacement; and aiming at the activation spectrum fingerprint data set, applying a dynamic spectrum matching algorithm, comparing with the spectrum response characteristic data set, calculating a cosine similarity coefficient, and when the similarity exceeds 0.85, determining the activation state and the tissue penetration depth of the nanoprobe, and generating a diagnosis and treatment parameter adjustment instruction set.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and particularly relates to an activatable head and neck cancer diagnosis and treatment integrated nanoprobes as well as a preparation method and application thereof. BACKGROUND

[0002] Problem background:

[0003] As the sixth most common malignant tumor in the world, head and neck cancer seriously threatens the life and health of patients due to its difficulty in early diagnosis and limited treatment effect. Traditional head and neck cancer diagnosis and treatment methods often need to be performed separately, which not only increases the pain of patients, but also reduces the accuracy and timeliness of treatment. The emergence of nanoprobes technology provides a new possibility for the realization of diagnosis and treatment integration, but existing nanoprobes still have significant limitations in the application of head and neck cancer.

[0004] Current nanoprobes technology mainly faces problems such as insufficient signal strength, limited tissue penetration ability, and lack of tumor-specific activation mechanism. Most probes cannot maintain stable signal output in complex physiological environments, making it difficult to effectively detect deep tumor tissues. At the same time, existing probes lack specific response ability to tumor microenvironment, resulting in low diagnostic accuracy.

[0005] In the development of head and neck cancer diagnosis and treatment integrated nanoprobes, the accurate control of spectral response characteristics constitutes the most critical technical challenge. The spectral performance of the probe under different excitation conditions directly determines the accuracy of diagnosis and effectiveness of treatment, and there is a complex mutual restraint relationship between the penetration depth and signal strength of near-infrared quantum dots. This restraint relationship further affects the emission spectrum stability of upconversion nanoparticles under different excitation powers, making it difficult for the probe to maintain consistent performance in changing physiological environments. More complex is that the probe will have a frequency band shift phenomenon before and after activation in the tumor microenvironment, and the quantitative control and prediction of this frequency band change become a bottleneck restricting the development of technology. Due to the lack of complete spectral response characteristic atlas as a reference standard, existing probes cannot achieve precise clinical application, which seriously limits the practicalization process of diagnosis and treatment integration technology.

[0006] Therefore, how to establish a complete spectral response test database of activatable head and neck cancer diagnosis and treatment integrated nanoprobes under different excitation conditions, realize the accurate quantification of the relationship between the penetration depth and signal strength of near-infrared quantum dots, master the rules of the emission spectrum of upconversion nanoparticles changing with excitation power, and establish a quantitative analysis method for the frequency band shift of the probe before and after activation in the tumor microenvironment, form a precise spectral response characteristic atlas, become the key problem to promote the development of head and neck cancer diagnosis and treatment integrated technology. SUMMARY

[0007] The application provides a head and neck cancer diagnosis and treatment integrated activatable nano probe, a preparation method and application thereof, and mainly comprises the following steps:

[0008] Raw spectral data of the nano probe in the head and neck cancer tissue sample is collected by using a spectrometer in a wavelength range of 300 nm to 1200 nm, the excitation power is set from 1 mW to 100 mW, a linear gradient with a step of 10 mW is adopted, and a raw data matrix containing the excitation power and the corresponding spectral intensity is generated;

[0009] For the raw data matrix, an adaptive spectral correction algorithm is applied to eliminate instrument background noise, when the relative fluctuation of the spectral intensity exceeds a preset threshold of 5%, a sliding window filtering technology is adopted, the window size is 5 nm, and a denoising spectral data set is generated;

[0010] According to the denoising spectral data set, a near-infrared quantum dot penetration depth model is constructed, the absorbance is calculated by using the Beer-Lambert law, the formula is A=εbc, wherein A represents the absorbance, ε represents the molar absorption coefficient, b represents the penetration depth, and c represents the quantum dot concentration, the spectral intensity is converted into the absorbance, the formula is A=-log(I / I0), wherein I represents the transmitted light intensity, I0 represents the incident light intensity, and a regression analysis is adopted to generate a penetration depth parameter set;

[0011] The pH value of the head and neck cancer tissue microenvironment is measured by using a pH sensor, the emission peak position change of the near-infrared quantum dot when the pH value is less than 6.5 is recorded in combination with the denoising spectral data set, a Gaussian fitting algorithm is adopted to analyze the frequency band displacement, and a mapping data set of the pH value and the frequency band displacement is generated;

[0012] According to the denoising spectral data set, the penetration depth parameter set and the mapping data set of the pH value and the frequency band displacement, a spectral response prediction model is constructed, a multivariate linear regression algorithm is applied, the excitation power, the penetration depth and the pH value are used as input variables, the spectral intensity is used as an output variable, the model parameters are trained until the prediction error converges to within 3%, and a spectral response feature data set is generated;

[0013] According to the spectral response feature data set, a real-time spectral monitoring system is constructed, spectral signals in a wavelength range of 300 nm to 1200 nm are continuously collected, when the emission peak red shift exceeds 10 nm is detected, it is judged that the nano probe has been activated and enters the treatment mode, and an activated spectral fingerprint data set is generated;

[0014] For the activated spectral fingerprint data set, a dynamic spectral matching algorithm is applied, compared with the spectral response feature data set, the cosine similarity coefficient is calculated, when the similarity exceeds 0.85, the activation state and the tissue penetration depth of the nano probe are determined, and a diagnosis and treatment parameter adjustment instruction set is generated;

[0015] Based on the activation spectral fingerprint dataset and the spectral response feature dataset, a spectral response quality evaluation system is constructed. The signal-to-noise ratio, spectral resolution, and frequency band stability are calculated. The signal-to-noise ratio is the ratio of the signal peak value to the background noise, the spectral resolution is the wavelength interval between adjacent peaks, and the frequency band stability is the standard deviation of the emission peak wavelength in multiple measurements. A standardized spectral response feature map is generated.

[0016] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0017] This invention discloses a head and neck cancer diagnosis and treatment system based on near-infrared quantum dots. The system acquires spectral data of nanoprobes in head and neck cancer tissue using a spectrometer, applies an adaptive correction algorithm to eliminate background noise, and constructs a quantum dot penetration depth model. Combined with pH sensor measurements of the tissue microenvironment, the system analyzes the emission peak shift of the near-infrared quantum dots and establishes a spectral response prediction model. This invention achieves real-time monitoring of the nanoprobe activation state; when a redshift of the emission peak exceeding 10 nanometers is detected, treatment mode is activated. A dynamic spectral matching algorithm determines the probe's activation state and tissue penetration depth, generating adjustment commands for diagnostic and treatment parameters. This invention also establishes a spectral response quality assessment system, calculates the signal-to-noise ratio, resolution, and stability, and generates standardized feature maps, providing a new technical means for the precise diagnosis and treatment of head and neck cancer. Attached Figure Description

[0018] Fig. 1 This is a flowchart illustrating an activatable nanoprobe for the diagnosis and treatment of head and neck cancer, its preparation method, and its application, according to the present invention.

[0019] Fig. 2 This is a schematic diagram of an activatable nanoprobe for the diagnosis and treatment of head and neck cancer, its preparation method, and its application, according to the present invention. Detailed Implementation

[0020] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0021] like Figs. 1-2 This embodiment of an activatable head and neck cancer diagnostic and therapeutic nanoprobe, its preparation method, and its application may specifically include:

[0022] S101, using a spectrometer to collect the original spectral data of the nanoprobes in the head and neck cancer tissue sample in the wavelength range of 300 nanometers to 1200 nanometers, setting the excitation power from 1 milliwatt to 100 milliwatts, using a linear gradient with a step size of 10 milliwatts, generating an original data matrix containing excitation power and corresponding spectral intensity.

[0023] The spectral data of the nanoprobes in the head and neck cancer tissue sample is obtained by collecting the original spectral signal in the preset wavelength range through the spectrometer to obtain the initial spectral data set. According to the initial spectral data set, the excitation power is adjusted to increase in a linear step, generating spectral intensity data at different power levels, obtaining the power-spectral data set. If there are noise signals in the power-spectral data set, the spectral intensity data is denoised by wavelet transform algorithm to obtain the denoised spectral data set. According to the denoised spectral data set, the spectral intensity characteristics corresponding to each wavelength are extracted to generate a feature vector set, obtaining the feature extraction data set. If the variance of the feature vectors in the feature extraction data set exceeds the preset threshold, the feature vectors are dimensionally reduced by principal component analysis algorithm to obtain the reduced feature data set. According to the reduced feature data set, the mapping relationship between spectral intensity and excitation power is constructed to generate a two-dimensional data matrix containing power and intensity, obtaining the final data matrix. The final data matrix is standardized to generate a normalized data matrix, obtaining a structured data set that can be used for subsequent analysis.

[0024] For example, the spectral data of the nanoprobes in the head and neck cancer tissue sample is collected by a spectrometer in the wavelength range of 300 nanometers to 1200 nanometers. First, the wavelength resolution of the spectrometer is configured to be 1 nanometer, the scanning range covers 300 to 1200 nanometers, and a data vector of 901 wavelength points is generated. The excitation light source uses a tunable laser, the power increases linearly from 1 milliwatt to 100 milliwatts, the step size is 10 milliwatts, and a total of 10 power levels (1, 10, 20, …, 100 milliwatts) are generated. To ensure data accuracy, the spectrometer is preheated for 30 minutes, the standard mercury lamp spectral line is calibrated, and the deviation is controlled within ±0.5 nanometers. The sample is placed on a quartz slide and fixed on the sample stage of the spectrometer. The laser beam is focused on the sample surface with a focal point diameter of 50 microns. At each power level, the spectrometer collects 10 spectral data, takes the average value to reduce noise, and sets the integration time to 100 milliseconds with a signal-to-noise ratio target of 100:1. The data processing uses a Python script, the original spectral data is stored as a 901x10 matrix, the rows correspond to 901 wavelength points, the columns correspond to 10 power levels, and the matrix elements are spectral intensity (unit: arbitrary unit a.u.).

[0025] For example, assume the intensity is 500 a.u. at 300 nm with 1 mW, 520 a.u. at 10 mW, and so on. The analysis process uses a principal component analysis (PCA) algorithm to reduce the dimensionality of the matrix, extract the first two principal components, and calculate their contribution rate (e.g., 90%) to identify the impact of power changes on spectral characteristics. The Z-score method is used for outlier detection, and data points deviating from the mean by three standard deviations are removed. The final generated data matrix is saved as a CSV file with the file name format "sample_spectrum_YYYYMMDD.csv". If further analysis is required, the matrix can be input into a convolutional neural network model to predict the distribution characteristics of the nanoprobes in cancer tissue, with the model input dimension being 901 x 10 and the output being a distribution probability map with an accuracy target of 95%. If a specific wavelength (e.g., 600 nm) is found to have significantly enhanced intensity with power in the data, it can be inferred that the nanoprobes have a specific fluorescent response at this wavelength, which needs to be further verified by Raman spectroscopy.

[0026] S102, for the original data matrix, apply adaptive spectral correction algorithm to eliminate instrument background noise, when the relative fluctuation of spectral intensity exceeds the preset threshold of 5%, use sliding window filtering technology, window size is 5 nm, generate denoising spectral data set.

[0027] Obtain the original spectral data matrix, use the fast Fourier transform algorithm for frequency domain conversion, get the frequency domain spectral data. For the frequency domain spectral data, apply the adaptive spectral correction algorithm to calculate the background noise baseline, get the corrected frequency domain data. If the intensity fluctuation of the corrected frequency domain data exceeds the preset threshold, use the sliding window filtering technology, set the fixed window width, generate the smoothed frequency domain data. Through the smoothed frequency domain data, apply the inverse Fourier transform algorithm, convert back to time domain, get the preliminary denoising spectral data set. For the preliminary denoising spectral data set, calculate the mean and variance of the signal intensity of each wavelength, get the intensity distribution characteristics. According to the intensity distribution characteristics, if the variance exceeds the preset threshold, perform local weighted regression filtering on the abnormal wavelength points to get the optimized denoising spectral data set. Through the optimized denoising spectral data set, generate the final spectral data matrix, store it as a standard format file.

[0028] For example, assume the original data matrix is a 1000 row x 500 column spectral data, each row represents a sample, and each column corresponds to a wavelength (400-900 nm, step 1 nm). To eliminate instrument background noise, apply the adaptive spectral correction algorithm. First, calculate the average intensity of each column of wavelengths to generate a background noise baseline.

[0029] For example, at wavelength 500 nm, the intensity values of 1000 samples are [1000, 1010,..., 1020], and the mean value is 1010. The background noise is estimated using a baseline fitting polynomial (3rd order) with the formula I_bg(λ) = aλ^3 + bλ^2 + cλ + d, where the coefficients a, b, c, d are solved by least squares method to obtain the fitting curve I_bg. Subtract I_bg from each column of the original data matrix to obtain the preliminary denoised data. Then, the relative fluctuation of spectral intensity is detected, defined as |I(λ) - I_bg(λ)| / I_bg(λ). If the fluctuation exceeds 5% (for example, at wavelength 600 nm, I = 1200, I_bg = 1000, the fluctuation is 20%), mark this wavelength point. For the marked points, apply sliding window filtering with a window size of 5 nm (i.e. 5 wavelength points), and perform mean filtering on the data [I_600, I_601,..., I_604] at wavelengths 600-604 nm to calculate I'_600 = (I_598 + I_599 + I_600 + I_601 + I_602) / 5, obtaining the smoothed value I'_600 = 1190. Repeat this process for all marked wavelengths to generate the denoised spectral data set. Finally, output the 1000 x 500 denoised matrix, and verify the denoising effect by comparing the signal standard deviation before and after denoising, for example, the original data standard deviation is 50, and after denoising it is reduced to 20, indicating that the noise is significantly reduced. This method realizes fully automated processing through matrix operations and filtering algorithms, and is suitable for high-throughput spectral analysis.

[0030] S103, constructing a near-infrared quantum dot penetration depth model according to the denoised spectral data set, calculating the absorbance using the Beer-Lambert law with the formula A = εbc, where A represents the absorbance, ε represents the molar absorption coefficient, b represents the penetration depth, and c represents the quantum dot concentration, converting the spectral intensity to absorbance with the formula A = -log(I / I0), where I represents the transmitted light intensity and I0 represents the incident light intensity, and generating a penetration depth parameter set using regression analysis.

[0031] The denoised spectral data set is acquired, spectral intensity data is extracted therefrom to obtain a first spectral data set. According to the first spectral data set, an absorbance value is calculated by using a formula A=-log(I / I0), wherein I is a transmitted light intensity, and I0 is an incident light intensity, to obtain a first absorbance data set. Absorbance values are extracted from the first absorbance data set, and a penetration depth is calculated by using a Beer-Lambert law A=εbc, wherein A is the absorbance value, ε is a molar absorption coefficient, b is the penetration depth, and c is a quantum dot concentration, in combination with a known quantum dot concentration and the molar absorption coefficient, to obtain a first penetration depth data set. If a value in the first penetration depth data set exceeds a preset threshold range, filtering processing is performed on an abnormal value, a mean filtering method is used, and a second penetration depth data set is generated. According to the second penetration depth data set, a near-infrared quantum dot penetration depth model is constructed, a linear regression analysis is used, model parameters are optimized, and a first model parameter set is obtained. Parameters are extracted from the first model parameter set, a model-predicted penetration depth value is calculated in combination with the second penetration depth data set, and a first predicted depth data set is obtained. If a deviation between the first predicted depth data set and the second penetration depth data set exceeds a preset threshold, linear regression model parameters are iteratively updated, a predicted depth value is recalculated, and a second predicted depth data set is obtained.

[0032] For example, based on the denoised spectral data set, a near-infrared quantum dot penetration depth model is constructed. First, the original near-infrared spectral data is denoised. It is assumed that the data set contains spectral intensity in the wavelength range of 700-1100 nm, and the number of data points is 1000. A wavelet transform denoising algorithm is used. Daubechies wavelet (db4) is selected, the number of decomposition layers is 5, high-frequency noise is filtered out by a soft threshold method, the threshold is set to 2 times the standard deviation of the signal, the denoised spectral intensity I is calculated, and the incident light intensity I0 is set to 1 (normalized). The Beer-Lambert law A=εbc is used to calculate the absorbance, wherein ε is the molar absorption coefficient of the quantum dot at 800 nm, which is set to 5000 L / (mol·cm), and c is the quantum dot concentration, which is set to 0.001 mol / L. The absorbance is converted by A=-log(I / I0). It is assumed that the transmitted light intensity I is 0.6, A=-log(0.6 / 1)=0.222 is calculated. The penetration depth b is solved by A=εbc,

[0033] b = A / (εc) = 0.222 / (5000 x 0.001) = 0.0444 cm. For multiple sets of concentration and spectral intensity data (e.g., c = 0.0005, 0.001, 0.002 mol / L, I = 0.7, 0.6, 0.5), generate absorbance data sets. Using multivariate linear regression analysis, input variables are concentration c and absorbance A, and output is penetration depth b, using least squares fitting, a regression model b = β0 + β1A + β2c is obtained, assuming the fitting result is b = 0.01 + 0.008A - 2.5c, R 2 = 0.95, indicating that the model has strong explanatory power. The parameter set is generated by regression coefficients, and the error analysis of the validation set shows that the root mean square error is 0.002 cm, indicating that the model prediction is accurate. Logically, denoising ensures data quality, absorbance conversion is based on physical laws, regression analysis quantifies parameter relationships, and finally forms a penetration depth parameter set suitable for quantum dot penetration prediction in biological tissues.

[0034] S104, measure the pH of the head and neck cancer tissue microenvironment using a pH sensor, combine the denoised spectral data set, record the emission peak position change of near-infrared quantum dots when the pH is less than 6.5, use Gaussian fitting algorithm to analyze the frequency band shift, and generate a mapping data set of pH and frequency band shift.

[0035] Collect the pH data of the head and neck cancer tissue microenvironment through the pH sensor, generate the original pH data set. Use denoising algorithm to process the original spectral data, generate denoised spectral data set. If the pH in the denoised spectral data set is lower than the preset threshold, extract the emission peak position of the near-infrared quantum dots, generate the peak position data set. Analyze the peak position data set by Gaussian fitting algorithm to get frequency band shift data. According to the frequency band shift data and the pH data, construct the mapping data set to determine the corresponding relationship between the pH and the frequency band shift. Use linear regression algorithm to fit the mapping data set to get the fitting model parameters. Predict the frequency band shift corresponding to the unknown pH through the fitting model parameters to generate the prediction data set.

[0036] For example, when measuring the pH of the microenvironment of head and neck cancer tissue using a pH sensor, a high-sensitivity micro-pH electrode sensor can be used in combination with a digital signal processing system to collect pH data from the tissue sample in real time. Assuming an experimental environment of an in vitro cultured head and neck cancer cell line (such as CAL-27), the sensor collects pH values at 1 second intervals, with a recording range of 4.0 to 8.0 and an accuracy of 0.01. The collected data is converted to digital signals by an analog-to-digital converter (ADC) with a resolution of 16 bits and stored as a time series data set. When denoising the spectral data set, the wavelet transform algorithm (Daubechies D4 basis function) is used, with a decomposition level of 5 layers and a threshold value of σ√(2ln(N)), where σ is the noise standard deviation and N is the number of data points (such as 1024). After denoising, a sub-data set with a pH less than 6.5 is extracted, and the emission spectrum of near-infrared quantum dots (CdSe / ZnS, particle size 5 nm) is analyzed. Using a spectrometer (resolution 0.1 nm), the emission peak is recorded, assuming that at pH = 6.5 the peak value is 650 nm, and for every 0.1 decrease in pH, the peak redshifts by 0.5 nm. Using the Gaussian fitting algorithm, the fitting function is f(λ) = A*exp(-(λ-μ)^2 / (2σ^2)), where A is the peak height, μ is the peak position, and σ is the half-width, and the parameters are optimized by the least squares method to obtain the fitting curve (R^2>0.95). The relationship between the frequency band displacement and the pH is analyzed by linear regression, assuming that the displacement Δλ = α(pH-6.5) + β, the regression result is α = -0.5, β = 0, and the correlation coefficient r = 0.98. Finally, a mapping data set is generated in the format (pH, Δλ), such as (6.4, 0.5), (6.3, 1.0). All data are stored as CSV files for subsequent machine learning models (such as support vector regression) to predict the relationship between pH and spectral characteristics, ensuring that the entire data processing workflow is automated, logical, and reproducible.

[0037] S105, according to the denoised spectral data set, the penetration depth parameter set, and the mapping data set of pH and frequency band displacement, a spectral response prediction model is constructed, a multivariate linear regression algorithm is applied, excitation power, penetration depth, and pH are used as input variables, and spectral intensity is used as output variable, model parameters are trained until the prediction error converges to within 3%, and a spectral response feature data set is generated.

[0038] The denoised spectral data set, the penetration depth parameter set and the pH and frequency band displacement mapping data set are acquired, a standardization processing method is adopted, and the data format and dimension are unified to obtain a preprocessed data set. According to the preprocessed data set, excitation power, penetration depth and pH are extracted as an input variable combination, a multiple linear regression model is constructed, model parameters are initialized, and an initial regression model is obtained. The spectral intensity in the preprocessed data set is defined as an output variable, the initial regression model is trained, model parameters are iteratively optimized, it is judged whether the prediction error is lower than a preset threshold, and an optimized regression model is obtained. If the prediction error of the optimized regression model is lower than the preset threshold, the input variable combination is predicted through the optimized regression model, predicted spectral intensity is generated, and a spectral response characteristic data set is obtained. According to the spectral response characteristic data set, the correlation between the input variable combination and the spectral intensity output is analyzed, a Pearson correlation coefficient calculation method is adopted, and a correlation coefficient matrix is obtained. Through the correlation coefficient matrix, the input variable combination with the highest correlation with the spectral intensity output is extracted, it is judged whether a preset correlation threshold is satisfied, and a key input variable combination is obtained. The key input variable combination and the optimized regression model are adopted to generate a final spectral response characteristic data set, which is saved in a standardized format, and a spectral response prediction result is obtained.

[0039] For example, based on the denoised spectral data set, the penetration depth parameter set, and the mapping data set of pH and frequency band shift, the process of constructing the spectral response prediction model is as follows. Assuming that the denoised spectral data set contains 1000 groups of spectral data, each group of data includes excitation power (unit: mW, range 10-100), penetration depth (unit: mm, range 0.1-5.0), pH (pH value, range 4.0-9.0), and corresponding spectral intensity (unit: a.u., range 0-1000). First, the denoised spectral data set is preprocessed, and the Z-score standardization method is used to normalize the excitation power, penetration depth, and pH to data with a mean of 0 and a standard deviation of 1, to eliminate dimensional differences. After preprocessing, the data set is divided into 80% training set (800 groups) and 20% test set (200 groups). For the mapping data set of pH and frequency band shift, it is assumed to contain 500 groups of mapping relationships, for example, pH 4.0 corresponds to frequency band shift 50 nm, and pH 9.0 corresponds to 80 nm. The linear interpolation algorithm is used to integrate the frequency band shift into additional features and supplement to the input variables to form a four-dimensional input vector (excitation power, penetration depth, pH, frequency band shift). Using the multiple linear regression algorithm, the model form is I = β0 + β1*P + β2*D + β3*pH + β4*S, where I is the spectral intensity, P, D, pH, and S are the excitation power, penetration depth, pH, and frequency band shift, respectively, and β0 to β4 are the parameters to be optimized. The least squares method is used to optimize the parameters, with an initial learning rate of 0.01, 1000 iterations, and a loss function of mean square error (MSE). During the training process, the MSE convergence is monitored, and when the test set MSE is less than 0.03 (i.e., the prediction error is within 3%), the iteration is stopped. The final model parameters are β0 = 10.5, β1 = 0.8, β2 = 1.2, β3 = 0.5, β4 = 0.3, and the MSE is 0.025. When generating the spectral response feature data set, based on the trained model, 100 new data (e.g., excitation power 50 mW, penetration depth 2.0 mm, pH 6.5, frequency band shift 65 nm) are input, the spectral intensity is predicted, and saved as a feature data set in CSV file format, containing input variables and predicted intensity. The analysis results show that the excitation power contributes the most to the spectral intensity (β1 = 0.8), indicating that it is a key influencing factor. To verify the robustness of the model, K-fold cross-validation (K = 5) is used, and the average MSE is 0.027, proving the stability of the model. The generated feature data set can be used for subsequent spectral analysis tasks, such as material property identification.

[0040] S106, according to the spectral response feature data set, construct a real-time spectral monitoring system, continuously collect spectral signals in the wavelength range of 300-1200 nm, and when the emission peak red shift exceeds 10 nm, it is judged that the nanoprobes have been activated and entered the treatment mode, and an activation spectral fingerprint data set is generated.

[0041] The spectral signal in the wavelength range of 300 nanometers to 1200 nanometers is collected by a spectrometer to obtain an original spectral data set. The original spectral data set is converted into a frequency domain spectral feature by using a fast Fourier transform algorithm. According to the frequency domain spectral feature, an emission peak position is extracted to obtain an emission peak characteristic parameter. If the emission peak characteristic parameter is red-shifted compared with a preset reference peak position and exceeds a preset threshold value, it is determined that the nanoprobes are activated, and an activation state flag is obtained. According to the activation state flag, the system is switched to a treatment mode to obtain a treatment mode signal. A spectral response feature is extracted from the treatment mode signal to generate an activation spectral fingerprint data set. The activation spectral fingerprint data set is classified by using a support vector machine algorithm to obtain a spectral fingerprint classification result.

[0042] For example, to construct a real-time spectral monitoring system, a high-resolution spectrometer (such as Ocean Optics USB4000, resolution 0.1 nanometer) is selected to collect spectral signals in the wavelength range of 300 nanometers to 1200 nanometers based on the spectral response feature data set, and the sampling frequency is 10 Hz to ensure the capture of transient changes. The system transmits the spectral data to a computer in real time through a USB interface, performs signal preprocessing by using the SciPy library in Python, removes noise by applying Savitzky-Golay filtering (window size 21, quadratic polynomial), and retains peak characteristics. Then, the spectral peak wavelength is calculated by using NumPy, and the specific algorithm is as follows: the wavelength corresponding to the maximum intensity of each sampling point is calculated, and is recorded as a time series λ_max(t). To detect the red shift of the emission peak, the system compares the current λ_max(t) with the initial reference peak value λ_ref (for example, 600 nanometers) in real time, and if |λ_max(t)-λ_ref|>10 nanometers, it is determined that the red shift occurs, indicating that the nanoprobes are activated. After activation, the system automatically enters the treatment mode, controls the laser (wavelength 532 nanometers, power 50 mW) to irradiate the sample by using LabVIEW, and triggers the photothermal effect. The activation spectral fingerprint data set is generated synchronously, contains three-dimensional data of wavelength, intensity, and timestamp, and is stored in CSV format, and the data structure is

[0043] [λ_i, I_i, t_i], where λ_i is a discrete wavelength point in the range of 300-1200 nanometers, I_i is the corresponding intensity, and t_i is the acquisition time. The fingerprint data set is reduced in dimension by principal component analysis (PCA), and the first three principal components (contribution rate >95%) are extracted for subsequent pattern recognition. The system stores the fingerprint data in a SQLite database to support real-time query and comparison, and ensures that the probe state in the treatment mode is traceable. If the red shift does not reach the threshold value, the system continues to monitor, and the above process is executed in a loop, logically forming a closed-loop control: acquisition→processing→detection→activation→recording, to ensure seamless connection between monitoring and treatment.

[0044] S107, for the activation spectrum fingerprint dataset, apply dynamic spectrum matching algorithm, compare with spectrum response feature dataset, calculate cosine similarity coefficient, when the similarity exceeds 0.85, determine the activation state and tissue penetration depth of the nano probe, and generate diagnosis and treatment parameter adjustment instruction set.

[0045] The initial spectrum signal is obtained from the spectrum fingerprint dataset, the noise is filtered through preprocessing to obtain the first spectrum feature set. The dynamic spectrum matching algorithm is used to compare the first spectrum feature set with the spectrum response feature dataset, calculate the cosine similarity coefficient, and obtain the similarity score. If the similarity score exceeds the preset threshold, the activation state of the nano probe is determined according to the score value, and the activation state identifier is obtained. The depth parameter is obtained by calculating the tissue penetration depth through the activation state identifier and combining the spectrum response feature dataset. According to the depth parameter and the activation state identifier, the diagnosis and treatment parameter adjustment instruction set is generated, and the adjustment instruction set is obtained. Key parameters are extracted from the adjustment instruction set, control signals are generated through a preset mapping rule, and device control instructions are obtained. The device control instructions are used to adjust the diagnosis and treatment device parameters to obtain the final diagnosis and treatment configuration.

[0046] For example, for the activation spectrum fingerprint dataset, first obtain the spectral data of the nanoprobe in a specific wavelength range (400-1000 nanometers) by the spectral acquisition device, generate a spectral fingerprint vector V1 = [I1, I2, …, I1024] containing 1024 wavelength points, where each Ii represents the light intensity value at the corresponding wavelength, in arbitrary units (a.u.). To ensure data quality, use the Gaussian filter algorithm (standard deviation σ = 2.0) to smooth V1, eliminate noise interference, and obtain the smoothed vector V1'. Then, apply the dynamic spectral matching algorithm (DLM) to compare V1' with the reference spectral vector Vref = [R1, R2, …, R1024] in the spectral response feature dataset D, which contains 1000 standard spectra of known tissue types. The DLM algorithm calculates the optimal matching path between V1' and Vref through dynamic time warping (DTW), sets the window width w = 10, calculates the Euclidean distance matrix, and optimizes the path to minimize the cumulative distance to obtain the matching score S. Next, calculate the cosine similarity coefficient, formula: cosθ = (V1' · Vref) / (||V1'| | |Vref||), where the dot product and norm are efficiently calculated through linear algebra libraries (such as NumPy). If cosθ > 0.85, it is determined that the nanoprobe is in an activated state, and according to the matched Vref corresponding to the tissue type, combined with the pre-set depth model (based on tissue light scattering coefficient μs = 10 cm-1 and absorption coefficient μa = 0.1 cm-1), the photon propagation path is simulated through the Monte Carlo light transmission algorithm to estimate the tissue penetration depth d, for example, d = 2.5 millimeters (error ± 0.1 millimeters). Finally, generate a set of diagnosis and treatment parameter adjustment instructions, based on the depth d and the activated state, adjust the laser power P = 50 mW (range 30-80 mW), pulse frequency f = 100 Hz, output instruction set in JSON format, for example: {“power": 50, “frequency": 100, “duration": 5}, and transmit to the diagnosis and treatment device through the API. This process is automated through a Python script to ensure logical rigor and traceable parameters.

[0047] S108, according to the activation spectrum fingerprint dataset and the spectral response feature dataset, construct a spectral response quality evaluation system, calculate the signal-to-noise ratio, spectral resolution and frequency band stability, the signal-to-noise ratio is the ratio of signal peak value to background noise, the spectral resolution is the wavelength interval of adjacent peaks, and the frequency band stability is the standard deviation of the emission peak wavelength in multiple measurements, and generate a standardized spectral response feature map.

[0048] The activation spectrum fingerprint data set and the spectrum response characteristic data set are acquired, a pretreatment algorithm is used to denoise and calibrate the data, and a first spectrum data set is obtained. Through the first spectrum data set, the ratio of signal peak value to background noise is calculated, and the signal-to-noise ratio is determined by using a fast Fourier transform algorithm. The wavelength interval of adjacent peaks is extracted from the first spectrum data set, and the spectral resolution is calculated based on a Gaussian fitting algorithm to obtain a spectral resolution value. For multiple measurement data, the distribution characteristics of the emission peak wavelength are analyzed, the standard deviation is calculated, and the frequency band stability is determined. If the standard deviation of the frequency band stability is lower than a preset threshold, the first spectrum data set is normalized to generate a second spectrum data set. According to the second spectrum data set, an interpolation algorithm is used to generate a standardized spectrum response characteristic spectrum to obtain a final spectrum. Feature parameters are extracted from the final spectrum and stored in a database to generate a spectrum response quality evaluation result.

[0049] For example, based on the activation spectrum fingerprint data set and the spectrum response characteristic data set, a spectrum response quality evaluation system is constructed. First, the signal peak value and the background noise are extracted from the activation spectrum fingerprint data set. Assuming that the data set contains 1000 wavelength points, the wavelength range is 400-1000 nm, the signal peak value at 532 nm is 1000 (arbitrary unit), and the background noise is calculated as 50 by averaging the non-peak value region from 400-450 nm. The signal-to-noise ratio algorithm is the peak value divided by the noise, i.e. 1000 / 50 = 20. Next, the spectral resolution is calculated. The adjacent peak values are 532 nm and 550 nm, the wavelength interval is 550-532 = 18 nm, and the resolution is defined. The frequency band stability is calculated by measuring the wavelength data of the 532 nm peak value 10 times (531.8, 532.1, 532.0, 531.9, 532.2, 531.7, 532.3, 531.6, 532.4, 531.9 nm), and the formula σ = √(Σ(xi-μ) 2 / N) is used, where μ is the mean value 532.0 nm, σ = 0.25 nm is calculated, and the peak wavelength stability is reflected. When generating the standardized spectrum response characteristic spectrum, the intensity values of the 1000 wavelength points are normalized to [0, 1] by the method (Ii-Imin) / (Imax-Imin), where Imax = 1000 and Imin = 0. After obtaining the normalized intensity, an interpolation algorithm (such as linear interpolation) is used to generate 1000-point spectrum data uniformly distributed in the range of 400-1000 nm. The analysis process ensures logical rigor, the signal-to-noise ratio reflects signal quality, the 18 nm resolution indicates spectral resolution, the 0.25 nm standard deviation indicates high stability, the normalized spectrum facilitates cross-dataset comparison, and the interpolation algorithm ensures spectrum smoothness, forming a complete chain from data extraction to quality evaluation and visualization.

[0050] The above embodiment is only one of the preferred embodiments of the present application, and should not be used to limit the protection scope of the present application, but any modification or polishing without substantial meaning made in the main design idea and spirit of the present application, and the technical problems solved are still consistent with the present application, and should be included in the protection scope of the present application.

Claims

1.A head and neck cancer diagnosis and treatment integrated activatable nano-probe, a preparation method and application thereof, characterized in that, The method comprises: Using a spectrometer to collect raw spectral data of the nanoprobes in the head and neck cancer tissue sample in the wavelength range of 300-1200 nm, setting the excitation power from 1-100 mW, using a linear gradient with a step size of 10 mW, generating a raw data matrix containing excitation power and corresponding spectral intensity; For the raw data matrix, apply an adaptive spectral correction algorithm to eliminate instrument background noise, when the relative fluctuation of spectral intensity exceeds the preset threshold of 5%, use a sliding window filtering technique with a window size of 5 nm to generate a denoised spectral data set; According to the denoised spectral data set, construct a near-infrared quantum dot penetration depth model, calculate the absorbance using the Beer-Lambert law, the formula is A=εbc, where A represents absorbance, ε represents molar absorption coefficient, b represents penetration depth, and c represents quantum dot concentration, convert spectral intensity to absorbance, the formula is A=-log(I / I0), where I represents transmitted light intensity and I0 represents incident light intensity, use regression analysis to generate a penetration depth parameter set; Using a pH sensor to measure the pH of the head and neck cancer tissue microenvironment, combining the denoised spectral data set, recording the emission peak position change of the near-infrared quantum dots when the pH is less than 6.5, using a Gaussian fitting algorithm to analyze the frequency band shift, generating a mapping data set of pH and frequency band shift; According to the denoised spectral data set, the penetration depth parameter set, and the mapping data set of pH and frequency band shift, construct a spectral response prediction model, apply a multiple linear regression algorithm, take excitation power, penetration depth, and pH as input variables, and take spectral intensity as output variable, train model parameters until the prediction error converges to within 3%, generating a spectral response feature data set; According to the spectral response feature data set, construct a real-time spectral monitoring system, continuously collect spectral signals in the wavelength range of 300-1200 nm, when the red shift of the emission peak exceeds 10 nm, determine that the nanoprobes have been activated and entered the treatment mode, generate an activated spectral fingerprint data set; For the activated spectral fingerprint data set, apply a dynamic spectral matching algorithm, compare with the spectral response feature data set, calculate the cosine similarity coefficient, when the similarity exceeds 0.85, determine the activation state of the nanoprobes and the tissue penetration depth, generate a diagnosis and treatment parameter adjustment instruction set; According to the activated spectral fingerprint data set and the spectral response feature data set, construct a spectral response quality evaluation system, calculate the signal-to-noise ratio, spectral resolution, and frequency band stability, the signal-to-noise ratio is the ratio of signal peak value to background noise, the spectral resolution is the wavelength interval between adjacent peaks, and the frequency band stability is the standard deviation of the emission peak wavelength in multiple measurements, generate a standardized spectral response feature spectrum. 2.The activatable head and neck cancer diagnosis and treatment integrated nanoprobes according to claim 1, wherein, The use of a spectrometer to collect raw spectral data of the nanoprobes in the head and neck cancer tissue sample in the wavelength range of 300-1200 nm, setting the excitation power from 1-100 mW, using a linear gradient with a step size of 10 mW, generating a raw data matrix containing excitation power and corresponding spectral intensity, comprises: Spectrum data of the nano probe in the head and neck cancer tissue sample is acquired, and original spectrum signals are collected in a preset wavelength range by a spectrometer to obtain an initial spectrum data set; According to the initial spectrum data set, the excitation power is adjusted to increase in a linear step to generate spectrum intensity data containing different power levels, and a power-spectrum data set is obtained; If there is noise signal in the power-spectrum data set, the spectrum intensity data is denoised by a wavelet transform algorithm to obtain a denoised spectrum data set; According to the denoised spectrum data set, the spectrum intensity features corresponding to each wavelength are extracted to generate a feature vector set, and a feature extraction data set is obtained; If the variance of the feature vector in the feature extraction data set exceeds a preset threshold, the feature vector is dimensionally reduced by a principal component analysis algorithm to obtain a reduced dimension feature data set; According to the reduced dimension feature data set, a mapping relationship between the spectrum intensity and the excitation power is constructed to generate a two-dimensional data matrix containing power and intensity, and a final data matrix is obtained; The final data matrix is standardized to generate a normalized data matrix, and a structured data set that can be used for subsequent analysis is obtained. 3.The activatable head and neck cancer diagnosis and treatment integrated nanoprobes according to claim 1, wherein, The adaptive spectrum correction algorithm is applied to the original data matrix to eliminate instrument background noise, and when the relative fluctuation of the spectrum intensity exceeds 5% of the preset threshold, the sliding window filtering technology is used, the window size is 5 nanometers, and a denoised spectrum data set is generated, including: The original spectrum data matrix is acquired, and the fast Fourier transform algorithm is used for frequency domain conversion to obtain frequency domain spectrum data; The adaptive spectrum correction algorithm is applied to the frequency domain spectrum data to calculate the background noise baseline to obtain corrected frequency domain data; If the intensity fluctuation of the corrected frequency domain data exceeds the preset threshold, the sliding window filtering technology is used to set a fixed window width to generate smoothed frequency domain data; Through the smoothed frequency domain data, the inverse Fourier transform algorithm is applied to convert back to the time domain to obtain a preliminary denoised spectrum data set; For the preliminary denoised spectrum data set, the signal intensity mean and variance of each wavelength are calculated to obtain the intensity distribution characteristics; According to the intensity distribution characteristics, if the variance exceeds the preset threshold, the local weighted regression filtering is performed on the abnormal wavelength points to obtain an optimized denoised spectrum data set; Through the optimized denoised spectrum data set, a final spectrum data matrix is generated and stored as a standard format file. 4.The activatable head and neck cancer diagnosis and treatment integrated nanoprobes according to claim 1, and the preparation method and application thereof, characterized in that, According to the denoised spectrum data set, a near-infrared quantum dot penetration depth model is constructed, the absorbance is calculated using the Beer-Lambert law, the formula is A=εbc, where A represents the absorbance, ε represents the molar absorption coefficient, b represents the penetration depth, and c represents the quantum dot concentration, the spectrum intensity is converted to absorbance, the formula is A=-log(I / I0), where I represents the transmitted light intensity and I0 represents the incident light intensity, and a penetration depth parameter set is generated by regression analysis, including: The denoised spectrum data set is acquired, and the spectrum intensity data is extracted therefrom to obtain a first spectrum data set; According to the first spectrum data set, the absorbance value is calculated using the formula A=-log(I / I0), where I is the transmitted light intensity and I0 is the incident light intensity, to obtain a first absorbance data set; The absorbance value is extracted from the first absorbance data set, combined with the known quantum dot concentration and molar absorption coefficient, and the Beer-Lambert law A = εbc is applied, wherein A is the absorbance value, ε is the molar absorption coefficient, b is the penetration depth, and c is the quantum dot concentration, to calculate the penetration depth and obtain the first penetration depth data set; If the value in the first penetration depth data set exceeds the preset threshold range, the abnormal value is filtered, the mean filtering method is used, and the second penetration depth data set is generated; According to the second penetration depth data set, a near-infrared quantum dot penetration depth model is constructed, linear regression analysis is used to optimize the model parameters, and the first model parameter set is obtained; From the first model parameter set, the model predicted penetration depth value is calculated combined with the second penetration depth data set, and the first predicted depth data set is obtained; If the deviation of the first predicted depth data set and the second penetration depth data set exceeds the preset threshold, the linear regression model parameters are iteratively updated, the predicted depth value is recalculated, and the second predicted depth data set is obtained. 5.The activatable head and neck cancer diagnosis and therapy integrated nanoprobes according to claim 1, and the preparation method and application thereof, characterized in that, The use of pH sensor measurement head and neck cancer tissue microenvironment pH, combined with denoising spectral data set, record the emission peak position change of near-infrared quantum dots when the pH is less than 6.5, use Gaussian fitting algorithm to analyze the frequency band displacement, generate pH and frequency band displacement mapping data set, including: Collect the pH data of the head and neck cancer tissue microenvironment through the pH sensor to generate the original pH data set; Use denoising algorithm to process the original spectral data to generate denoising spectral data set; If the pH in the denoising spectral data set is lower than the preset threshold, the emission peak position of the near-infrared quantum dot is extracted to generate the peak position data set; Analyze the peak position data set through the Gaussian fitting algorithm to obtain the frequency band displacement data; According to the frequency band displacement data and the pH data, a mapping data set is constructed to determine the corresponding relationship between the pH and the frequency band displacement; Use linear regression algorithm to fit the mapping data set to obtain the fitting model parameters; Predict the frequency band displacement corresponding to the unknown pH through the fitting model parameters to generate the prediction data set. 6.The activatable head and neck cancer diagnosis and treatment integrated nanoprobes according to claim 1, and the preparation method and application thereof, characterized in that, According to the denoising spectral data set, the penetration depth parameter set, and the mapping data set of pH and frequency band displacement, a spectral response prediction model is constructed, a multivariate linear regression algorithm is applied, the excitation power, penetration depth, and pH are used as input variables, and the spectral intensity is used as output variable, the model parameters are trained until the prediction error converges to within 3%, and the spectral response feature data set is generated, including: Get the denoising spectral data set, the penetration depth parameter set, and the pH and frequency band displacement mapping data set, use the standardization processing method to unify the data format and dimension, and obtain the preprocessed data set; According to the preprocessed data set, the excitation power, penetration depth, and pH are extracted as input variable combinations to construct a multivariate linear regression model, the model parameters are initialized, and the initial regression model is obtained; Use the spectral intensity in the preprocessed data set as the output variable definition to train the initial regression model, iteratively optimize the model parameters, and determine whether the prediction error is lower than the preset threshold to obtain the optimized regression model; If the prediction error of the optimized regression model is lower than the preset threshold, the input variable combination is predicted by the optimized regression model to generate a predicted spectral intensity, and a spectral response feature dataset is obtained; According to the spectral response feature dataset, the correlation between the input variable combination and the spectral intensity output is analyzed, and a correlation coefficient matrix is obtained by using a Pearson correlation coefficient calculation method; Through the correlation coefficient matrix, the input variable combination with the highest correlation with the spectral intensity output is extracted, and it is judged whether it meets a preset correlation threshold to obtain a key input variable combination; The key input variable combination and the optimized regression model are used to generate a final spectral response feature dataset, which is saved in a standardized format to obtain a spectral response prediction result. 7.The activatable head and neck cancer diagnosis and therapy integrated nanoprobes according to claim 1, and the preparation method and application thereof, characterized in that, According to the spectral response feature dataset, a real-time spectral monitoring system is constructed, and spectral signals in a wavelength range of 300 nanometers to 1200 nanometers are continuously collected. When a red shift of an emission peak exceeding 10 nanometers is detected, it is judged that the nanoprobes have been activated and entered a treatment mode, and an activated spectral fingerprint dataset is generated, including: Spectral signals in a wavelength range of 300 nanometers to 1200 nanometers are collected by a spectrometer to obtain an original spectral dataset; A fast Fourier transform algorithm is used to perform frequency domain conversion on the original spectral dataset to obtain a frequency domain spectral feature; According to the frequency domain spectral feature, an emission peak position is extracted to obtain an emission peak characteristic parameter; If the emission peak characteristic parameter has a red shift compared with a preset reference peak position and exceeds a preset threshold, it is determined that the nanoprobes are activated, and an activation state flag is obtained; According to the activation state flag, the system is switched to a treatment mode to obtain a treatment mode signal; Spectral response features are extracted from the treatment mode signal to generate the activated spectral fingerprint dataset; The activated spectral fingerprint dataset is classified by a support vector machine algorithm to obtain a spectral fingerprint classification result. 8.The activatable head and neck cancer diagnosis and treatment integrated nanoprobes according to claim 1, and the preparation method and application thereof, characterized in that, For the activated spectral fingerprint dataset, a dynamic spectral matching algorithm is applied to compare with the spectral response feature dataset to calculate a cosine similarity coefficient. When the similarity exceeds 0.85, the activation state and the tissue penetration depth of the nanoprobes are determined, and a diagnosis and treatment parameter adjustment instruction set is generated, including: An initial spectral signal is obtained from the spectral fingerprint dataset, and noise is filtered by preprocessing to obtain a first spectral feature set; A dynamic spectral matching algorithm is used to compare the first spectral feature set with the spectral response feature dataset to calculate a cosine similarity coefficient, and a similarity score is obtained; If the similarity score exceeds a preset threshold, the activation state of the nanoprobes is determined according to the score value, and an activation state identifier is obtained; The tissue penetration depth is calculated by combining the activation state identifier and the spectral response feature dataset to obtain a depth parameter; According to the depth parameter and the activation state identifier, a diagnosis and treatment parameter adjustment instruction set is generated to obtain an adjustment instruction set; Key parameters are extracted from the adjustment instruction set, and a control signal is generated by a preset mapping rule to obtain a device control instruction; The device control instruction is used to adjust the diagnosis and treatment device parameters to obtain a final diagnosis and treatment configuration. 9.The activatable head and neck cancer diagnosis and treatment integrated nanoprobes according to claim 1, and the preparation method and application thereof, characterized in that, The spectrum response quality evaluation system is constructed according to the activation spectrum fingerprint data set and the spectrum response characteristic data set, the signal-to-noise ratio, the spectrum resolution and the frequency band stability are calculated, the signal-to-noise ratio is the ratio of the signal peak value to the background noise, the spectrum resolution is the wavelength interval of adjacent peaks, the frequency band stability is the standard deviation of the emission peak wavelength in multiple measurements, and the standardized spectrum response characteristic spectrum is generated, including: An activation spectrum fingerprint data set and a spectrum response characteristic data set are acquired, a pretreatment algorithm is used to denoise and calibrate the data to obtain a first spectrum data set; The ratio of the signal peak value to the background noise is calculated through the first spectrum data set, and a fast Fourier transform algorithm is used to determine the signal-to-noise ratio; The wavelength interval of adjacent peaks is extracted from the first spectrum data set, the spectrum resolution is calculated based on a Gaussian fitting algorithm, and the spectrum resolution value is obtained; For multiple measurement data, the distribution characteristics of the emission peak wavelength are analyzed, the standard deviation is calculated, and the frequency band stability is determined; If the standard deviation of the frequency band stability is lower than a preset threshold, the first spectrum data set is normalized to generate a second spectrum data set; According to the second spectrum data set, an interpolation algorithm is used to generate a standardized spectrum response characteristic spectrum to obtain a final spectrum; Feature parameters are extracted from the final spectrum and stored in a database to generate a spectrum response quality evaluation result.

Images