Organism Raman spectrum cross-tissue migration processing method based on scalar pair function regression

Abstract

The invention discloses a scalar pair function regression-based organism Raman spectrum cross-tissue migration processing method, which comprises the following steps: constructing a cross-tissue spectrum database which comprises Raman spectrum data and target component concentration data; carrying out pretreatment on the Raman spectrum data; carrying out B-spline primary function fitting on the preprocessed Raman spectrum data, and establishing a function relationship between the Raman spectrum wave number and the light intensity; constructing and training a scalar pair function regression model; and inputting the preprocessed Raman spectrum data of the to-be-detected organism source tissue into the trained scalar pair function regression model to obtain a predicted target component concentration. According to the method disclosed by the invention, the mapping model from the easily-measured source tissue spectrum to the difficultly-measured target tissue component concentration is established, so that nondestructive detection for predicting the internal tissue components of the organism without dissection is realized; an effective solution is provided for solving the long-standing technical problem of noninvasive and dynamic monitoring of internal tissue components of living organisms.

Description

Scalar-pair function regression method for cross-tissue migration processing of biological Raman spectra Technical Field

[0001] This invention relates to the field of biotechnology, and more particularly to a method for processing trans-tissue migration of Raman spectra in organisms using scalar pair function regression. Background Technology

[0002] A spectrum is the pattern formed when monochromatic light is arranged according to wavelength or frequency after polychromatic light is dispersed by a dispersive system (such as a prism or grating). It is also called the optical spectrum. Light waves are electromagnetic radiation produced by electrons in the motion of atoms. Different substances emit or absorb light at different wavelengths because of the different electron movements within their atoms. Therefore, studying the spectral characteristics of different substances has become a specialized discipline—spectroscopy.

[0003] Spectroscopy, intersecting with physics, chemistry, and biology, studies the interaction between electromagnetic waves and matter, enabling the analysis of material composition and structure at the molecular level. In biology, spectroscopy is a powerful analytical tool. By studying the characteristic "fingerprint" spectra produced when biological tissues interact with light, we can reveal their intrinsic chemical composition, analyze the composition, molecular structure, and dynamic changes of samples, and achieve qualitative and even quantitative analysis of the chemical components of living or ex vivo tissues without complex preprocessing. It is widely used in disease diagnosis, drug development, and basic life science research, providing a crucial technological foundation for understanding the molecular basis of life processes.

[0004] Raman spectroscopy, a non-invasive analytical technique based on the Raman scattering effect, has shown great potential in biomedical research, clinical diagnosis, drug development, and quality trait detection due to its unique advantages of being non-destructive, highly sensitive, highly specific, and capable of real-time dynamic monitoring. It can reveal the conformational changes, distribution, and interactions of biological macromolecules (such as proteins, nucleic acids, and lipids) in near-in-situ conditions, providing crucial technical support for understanding the molecular basis of life processes.

[0005] However, despite its superior performance in detecting samples from the body surface or ex vivo, Raman spectroscopy still faces significant technical bottlenecks in the in vivo and in situ detection of internal tissues. Current conventional Raman detection methods typically only acquire spectral information from the body surface or dissected ex vivo samples. To probe internal tissues, two main approaches are currently employed: one is invasive dissection or sectioning, which completely negates the non-destructive advantage of this technique and may alter the original physiological state of the tissue; the other is point-by-point measurement using in situ probes, a method that is not only complex to operate and has low throughput, but also struggles to achieve efficient, global scanning of deep tissues or large areas.

[0006] In summary, existing Raman spectroscopy techniques for analyzing the internal structures of organisms still suffer from problems such as significant operational damage, low detection efficiency, and difficulty in cross-tissue migration. Therefore, there is an urgent need to develop a low-destructive, efficient method for processing Raman spectral information from organisms that enables cross-tissue migration. Summary of the Invention

[0007] The purpose of this invention is to provide a method for cross-tissue migration processing of biological Raman spectra using scalar pair function regression. By fully utilizing the morphological information of the complete spectrum through the scalar pair function regression model, rather than relying on limited characteristic peaks, the prediction accuracy and data utilization are significantly improved. The introduction of coarse penalty and basis function optimization techniques effectively enhances the stability and generalization ability of the model, ensuring reliable predictions across different individuals. A complete statistical validation system provides a solid guarantee for the reliability of the analysis results.

[0008] To achieve the above objectives, this invention provides a method for cross-tissue migration processing of biological Raman spectra using scalar-pair function regression, comprising the following steps: S1, constructing a cross-tissue spectral database, the database including Raman spectral data and target component concentration data; and preprocessing the Raman spectral data; S2, fitting the preprocessed Raman spectral data using B-spline basis functions to establish a functional relationship between Raman spectral wavenumber and light intensity; S3, constructing and training a scalar-pair function regression model; S4, inputting the preprocessed Raman spectral data of the biological source tissue to the trained scalar-pair function regression model to obtain the predicted target component concentration, thus completing the cross-tissue spectral information migration.

[0009] Preferably, S1 is as follows: S11, select the same organism as the sampling subject, and divide the organism's tissues into easily measurable source tissues and difficult-to-measurable target tissues; (1) source tissues include body wall, skin or mucous membrane; (2) target tissues include internal organs such as muscles or gonads; S12, collect paired data by dissecting and sampling after in vivo detection of the organism; paired data includes Raman spectral data and component concentration data; S13, preprocess the Raman spectral data collected in S12 to form a training dataset.

[0010] Preferably, the preprocessing in S13 includes filtering and noise reduction, removal of cosmic rays, normalization, baseline correction, and area normalization.

[0011] Preferably, S2 specifically includes: S21, determining the optimal number of basis functions based on B-spline theory; S22, based on the determined optimal number of basis functions, performing B-spline basis function fitting on the Raman spectral data after S1 preprocessing, transforming the discrete Raman spectral data into a smooth continuous function, and establishing a continuous functional relationship between spectral wavenumber and light intensity.

[0012] Preferably, S3 specifically comprises: S31, constructing a scalar-pair regression model; the input of the scalar-pair regression model is the functional relationship between the Raman spectral wavenumber and light intensity obtained in S2, and the output is the scalar concentration of the target component corresponding to the functional relationship; S32, during the regression training process, the scalar-pair regression model introduces a regularized curvature coarsening penalty, and the coefficient of the regularized curvature coarsening penalty term is automatically determined through generalized cross-validation to complete the model training and obtain the coefficient function of the trained scalar-pair regression model; S33, calculating the residual sum of squares, F-statistic, and corresponding p-value of the scalar-pair regression model, and judging the predictive significance of the model based on the p-value.

[0013] Preferably, S4 specifically involves: S41, inputting the preprocessed Raman spectral data of the biological tissue to be tested into the trained scalar-pair function regression model to obtain the predicted target component concentration, thus successfully completing the transfer of component information from easily accessible tissue to difficult-to-obtain tissue; S42, evaluating the accuracy of the predicted value through statistical verification methods.

[0014] Preferably, S42 specifically involves: performing a grouped linear regression between the predicted target component concentration output from the scalar-pair function regression model and the actual measured target component concentration, and then using the coefficient of determination... Indicators validate the accuracy of predictions.

[0015] Therefore, the present invention adopts the above-mentioned scalar pair function regression biological Raman spectrum cross-tissue migration processing method, which has the following beneficial effects: (1) The present invention establishes a mathematical mapping relationship between source tissue and target tissue through functional data processing and statistical learning model, realizes accurate and non-destructive prediction from easily measurable source tissue spectrum to difficult-to-measurable target tissue component concentration, realizes non-invasive component prediction from "from the surface to the inside", avoids damage to organisms by dissection sampling, and breaks through the bottleneck of non-destructive detection.

[0016] (2) The present invention adopts a functional data analysis method, which makes full use of the morphological information of the complete spectral curve, rather than the limited characteristic peaks in the traditional method, which significantly improves the information utilization rate and model accuracy, and enhances the depth of data analysis.

[0017] (3) This invention effectively prevents overfitting by using rough penalty and basis function optimization techniques, ensuring the stability and generalization ability of the model among different samples and enhancing the robustness of the model; it introduces a complete statistical testing system to perform multi-dimensional verification from model significance to prediction accuracy, providing a solid statistical reliability guarantee for the analysis results.

[0018] (4) This method has the advantages of low damage and high efficiency, and provides an effective solution to the long-standing technical problem of non-invasive and dynamic monitoring of the internal tissue components of living organisms. It has broad application prospects in life science research.

[0019] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0020] Figure 1 is an overall experimental flowchart of the Raman spectroscopy cross-tissue migration processing method of the present invention; Figure 2 is a box plot of the measured muscle glycogen content in paired samples of the cross-tissue spectrum and composition database of *Scallop spp.* in the embodiment of the present invention; Figure 3 is a representative gill Raman spectrum curve of the cross-tissue spectrum and composition database of *Scallop spp.* after smoothing and baseline correction in the embodiment of the present invention; Figure 4 is a schematic diagram of the B-spline basis function fitting continuous spectral function curve and the original discrete spectral data points in the embodiment of the present invention; Figure 5 is a coefficient function image obtained after training the scalar pair function regression model in the embodiment of the present invention; Figure 6 is a comparison of the predicted values ​​and actual measured values ​​grouped by the method of the present invention and the traditional prediction method based on characteristic peaks, wherein (A) is a grouped linear regression diagram of the predicted values ​​and actual measured values ​​of the method of the present invention, and (B) is a grouped linear regression diagram of the predicted values ​​and actual measured values ​​of the traditional prediction method based on characteristic peaks. Detailed Implementation

[0021] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0022] This invention discloses a scalar-pair function regression method for cross-tissue Raman spectroscopy transfer in organisms, comprising the following steps: S1, constructing a cross-tissue spectral database, the database including Raman spectral data and target component concentration data; and preprocessing the Raman spectral data; S2, fitting the preprocessed Raman spectral data with B-spline basis functions to establish a functional relationship between Raman spectral wavenumber and light intensity; S3, constructing and training a scalar-pair function regression model; S4, inputting the preprocessed Raman spectral data of the target organism's source tissue into the trained scalar-pair function regression model to obtain the predicted target component concentration, thus completing the cross-tissue spectral information transfer.

[0023] Example: This example uses the scallop (Chlamys farreri) as an example for research. The cross-tissue migration processing flow of Raman spectral information of gill filaments of the scallop is shown in Figure 1.

[0024] A method for cross-tissue migration processing of biological Raman spectra using scalar-pair function regression includes the following steps: S1, constructing a cross-tissue spectral database, the database including Raman spectral data and target component concentration data; and preprocessing the Raman spectral data.

[0025] S11. Select the same organism as the sampling subject, and divide the organism's tissues into easily measurable source tissues and difficult-to-measurable target tissues.

[0026] (1) The source tissue is the body wall, skin or mucous membrane, etc.

[0027] (2) The target tissue is an internal organ such as muscle or gonad.

[0028] S12. After in vivo detection, the organism is dissected and sampled to collect paired data; the paired data includes Raman spectroscopy and target component concentration data.

[0029] S13. Preprocess the Raman spectral data collected in S12 to form a training dataset.

[0030] In this embodiment, firstly, 100 healthy scallops of uniform size were selected; all individuals were dissected, and easily obtainable gill filament tissue and difficult-to-obtain adductor muscle (hereinafter referred to as "muscle") tissue were extracted respectively.

[0031] Secondly, the paired data were acquired: a) Source tissue spectra were collected: gill filament tissue was cut into cubes with a side length of 0.5 cm, embedded in cryo-embedding medium (OCT), and then flash-frozen in liquid nitrogen; at -20°C, it was cut into 50 μm thin sections using a cryostat and attached to aluminized glass slides; spectral acquisition was performed using a Raman spectrometer equipped with a 532 nm laser, with parameters set to: spectral range 400-3400 cm⁻¹. -1 .

[0032] b) Determination of target component concentration: Take muscle tissue less than or equal to 100 mg and determine its glycogen content using the anthrone-sulfuric acid method. The specific steps are as follows: (1) Prepare double-distilled water, alkaline solution, freshly prepared 0.01 mg / mL glucose standard solution, and concentrated sulfuric acid glycogen colorimetric solution.

[0033] (2) The muscle tissue was dried with filter paper and weighed m. The alkaline solution and tissue were added to a test tube at a ratio of 1:3 (sample weight (mg): alkaline solution volume (μL)). The test tube was boiled in a water bath for 20 minutes and cooled under running water. Double distilled water was added at a ratio of 1:16 (sample weight (mg): double distilled water volume (μL)).

[0034] (3) After digestion with alkaline solution, the tissue was reacted with concentrated sulfuric acid colorimetric solution along with freshly prepared glucose standard solution. The absorbance (OD value) was measured at a wavelength of 620 nm using a UV spectrophotometer. The glycogen content (mg / g tissue wet weight) was calculated using a standard curve and used as the true value of the target component concentration. The box plot of the measured muscle glycogen content is shown in Figure 2. It can be seen that it basically conforms to a normal distribution, and the data is highly representative.

[0035] Through the above operations, 100 sets of complete paired data were successfully obtained: gill filament Raman spectrum ↔ muscle glycogen content.

[0036] Finally, the collected gill filament Raman spectral data were preprocessed to eliminate noise and background interference. The specific steps are as follows: a) Abnormal spectrum screening: Abnormal spectra with excessively low signal-to-noise ratio or severe fluorescence saturation were removed.

[0037] b) Cosmic ray removal: A peak filtering algorithm (filter_size=4, dynamic_factor=6) is used to identify and remove cosmic ray spikes.

[0038] c) Wavenumber axis standardization: uniformly truncated to 400-3400cm -1 The effective spectral range, and with 1cm -1 Wavenumber alignment is performed for the interval.

[0039] d) Baseline correction: Asymmetric least squares (ASLS) is used to subtract the fluorescence background to eliminate intensity differences caused by factors such as laser power fluctuations.

[0040] e) Noise removal: Spectral smoothing was performed using the Savitzky-Golay algorithm (window width: 10, polynomial order: 3).

[0041] f) Area normalization: The total area of ​​the intensity signal of each spectrum is normalized to eliminate the effects of concentration and laser fluctuations.

[0042] g) Data consistency guarantee: Under the condition of internal consistency of the sample, ensure that each scallop individual ultimately retains 10 high-quality effective spectra.

[0043] After preprocessing, all Raman spectral data and their corresponding true values ​​of gonadal glycogen content are combined to construct a training dataset. The representative preprocessed gill filament Raman spectra are shown in Figure 3, exhibiting no cosmic ray peaks, a high signal-to-noise ratio, and distinct characteristic peaks.

[0044] S2. Perform B-spline basis function fitting on the preprocessed spectral data obtained in S1 to establish the functional relationship between spectral wavenumber and light intensity.

[0045] S21. Determine the optimal number of basis functions based on B-spline theory.

[0046] S22. Based on a determined number of optimal basis functions, B-spline basis function fitting is performed on the Raman spectral data after S1 preprocessing to transform the discrete Raman spectral data into a smooth continuous function and establish the functional relationship between spectral wavenumber and light intensity.

[0047] In this embodiment, based on the preprocessed spectral dataset, the optimal number of B-spline basis functions for function fitting is determined by B-spline theory to be 14. This number ensures sufficient fitting accuracy for the details of the spectral curve while effectively preventing overfitting.

[0048] Using the 14 cubic B-spline basis functions identified above, a smoothing fit is performed on each preprocessed discrete spectral data point, transforming it from a discrete (wavenumber-intensity) data point sequence into a continuous, smooth spectral function. This step is crucial for realizing functional data analysis, providing continuous mathematical input for the subsequent scalar-to-function regression model. Figure 4 illustrates the fitting effect of the B-spline basis functions, showing the correspondence between the continuous spectral function curve and the original discrete spectral data points. The blue dots represent the original discrete spectral data points, while the red curve represents the continuous spectral function curve obtained after fitting with the 14 cubic B-spline basis functions.

[0049] S3. Construct and train a scalar pair regression model.

[0050] S31. Construct a scalar-pair function regression model; the input of this model is the continuous functional relationship between spectral wavenumber and light intensity obtained in S2, and the output is the scalar concentration of the target component corresponding to the continuous functional relationship.

[0051] S32. During the regression training process, the scalar pair regression model introduces a regularized curvature coarsening penalty. The coefficient λ of the regularized curvature coarsening penalty term is automatically determined through generalized cross-validation to complete the model training and obtain the coefficient function of the trained scalar pair regression model.

[0052] S33. Calculate the sum of squared residuals, F-statistic, and corresponding p-values ​​of the scalar regression model, and determine the predictive significance of the model based on the p-values.

[0053] In this embodiment, the constructed scalar pair function regression model uses the gill filament spectral function obtained by S2 as the functional independent variable and the muscle glycogen content measured by spectrophotometry as the scalar dependent variable.

[0054] During regression training, a regularized curvature coarsening penalty is introduced into the scalar pair regression model to smooth out fluctuations in the coefficient function and suppress interference from spectral noise. Generalized cross-validation automatically determines the optimal penalty term coefficient λ to be 0.0126, thus achieving an optimal balance between goodness of fit and model complexity, significantly improving the model's generalization ability.

[0055] After the regression was completed, the coefficient function graph obtained after model training is shown in Figure 5. This graph reflects the contribution weight of different wavenumbers to the prediction of the target component.

[0056] After training, the model's significance was tested by calculating the residual sum of squares, F-statistic, and corresponding p-value. The results showed that the F-statistic was 8488.6, and the corresponding p-value was less than 0.001. This indicates that the spectral function of the source tissue (gill filaments) has a highly significant predictive ability for the target component (muscle glycogen content), and the model has solid statistical significance.

[0057] S4. Input the preprocessed Raman spectrum of the biological tissue to be tested into the trained scalar logarithmic regression model to complete the cross-tissue spectral information transfer.

[0058] S41. The preprocessed Raman spectral data of the biological tissue to be tested is input into the trained scalar logarithmic regression model to obtain the predicted concentration of the target component, thus successfully completing the transfer of component information from easily accessible tissues to difficult-to-obtain tissues.

[0059] S42. Evaluate the accuracy of the predicted values ​​using statistical validation methods.

[0060] The predicted values ​​of the target component concentration output by the scalar logarithmic function regression model are grouped and linearly regressed with the actual measured values ​​of the target component concentration. The accuracy of the model prediction is verified based on indicators such as the coefficient of determination.

[0061] In this embodiment, the Raman spectrum of the gill filaments of the scallop is input, and after preprocessing and functionalization, it is input into a pre-trained scalar-pair function regression model. The model directly outputs the predicted value of glycogen content in the muscle of the scallop, successfully completing the transfer of component information from easily obtainable gill filament tissue to difficult-to-obtain muscle tissue.

[0062] Using the actual measured values ​​of muscle glycogen in the test set as the gold standard, a grouped linear regression analysis was performed on the model's predicted values ​​and actual measured values ​​for all test samples. Validation results showed that the coefficient of determination between the predicted and actual values ​​was [value missing]. The accuracy is as high as 0.77. Figure 6 shows a comparison of the prediction accuracy between the method of this invention and the traditional prediction method based on characteristic peaks. The figure demonstrates the performance advantage through the grouped linear regression results of the predicted values ​​and the actual measured values. The indicators show that the model has high prediction accuracy, reliable results, and excellent practical value.

[0063] Therefore, the present invention adopts the above-mentioned scalar pair function regression biological Raman spectral information cross-tissue migration processing method, which has the following beneficial effects: (1) This method establishes a component prediction channel from easily obtainable tissues to difficult-to-obtain tissues by constructing an accurate mathematical mapping model of "gill filament spectrum-muscle glycogen", and realizes the purpose of accurately and quickly predicting the content of key nutrients in internal muscles by measuring gill filament tissue in a low-loss or non-destructive manner.

[0064] (2) This method completely avoids the destructive dissection required by traditional methods to obtain component data, overcomes the bottleneck that existing Raman spectroscopy technology is difficult to apply to the detection of internal tissues of organisms, and provides a new technical approach for long-term, dynamic physiological monitoring of the same living individual.

[0065] (3) This method only requires conventional Raman equipment and does not require expensive special instruments, which greatly reduces the implementation threshold; the operation process is simple, eliminating the complex sample pretreatment and time-consuming dissection operation, which significantly improves the detection efficiency; at the same time, this method has good versatility and can be extended to the prediction and analysis of other aquatic species or other components.

[0066] In aquatic animal breeding, this invention provides key technical support for the early selection of superior traits; in aquaculture management, it enables real-time, non-destructive assessment of nutritional status, providing data support for precise feeding; and in quality control, it provides new technical means for the quality grading of aquatic products. Therefore, this invention not only has significant scientific research value but also broad commercial application prospects, providing strong technical support for promoting the refined and intelligent development of aquaculture.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for processing trans-tissue migration of biological Raman spectra using scalar-pair function regression, characterized in that, Includes the following steps: S1. Construct a cross-tissue spectral database, which includes Raman spectral data and target component concentration data; The process involves: S1) Preprocessing the Raman spectral data; S2) Fitting the preprocessed Raman spectral data using B-spline basis functions to establish the functional relationship between Raman spectral wavenumber and light intensity; S3) Constructing and training a scalar-pair function regression model; and S4) Inputting the preprocessed Raman spectral data of the biological tissue to be tested into the trained scalar-pair function regression model to obtain the predicted concentration of the target component, thus completing the cross-tissue spectral information transfer.

2. The method for processing trans-tissue migration of biological Raman spectra using scalar-pair function regression according to claim 1, characterized in that, S1 specifically includes: S11, selecting the same organism as the sampling subject, and dividing the organism's tissues into easily measurable source tissues and difficult-to-measurable target tissues; (1) source tissues include body wall, skin or mucous membrane; (2) target tissues include internal organs such as muscles or gonads; S12, collecting paired data by dissecting and sampling after in vivo detection of the organism; paired data includes Raman spectral data and component concentration data; S13, preprocessing the Raman spectral data collected in S12 to form a training dataset.

3. The method for processing trans-tissue migration of biological Raman spectra using scalar-pair function regression according to claim 2, characterized in that, Preprocessing in S13 includes filtering and noise reduction, removal of cosmic rays, normalization, baseline correction, and area normalization.

4. The method for processing trans-tissue migration of biological Raman spectra using scalar-pair function regression according to claim 1, characterized in that, S2 specifically consists of: S21, determining the optimal number of basis functions based on B-spline theory; S22, based on the determined optimal number of basis functions, performing B-spline basis function fitting on the Raman spectral data after S1 preprocessing, transforming the discrete Raman spectral data into a smooth continuous function, and establishing a continuous functional relationship between spectral wavenumber and light intensity.

5. The method for processing trans-tissue migration of biological Raman spectra using scalar-pair function regression according to claim 1, characterized in that, S3 specifically comprises: S31, constructing a scalar-pair regression model; the input of the scalar-pair regression model is the functional relationship between the Raman spectral wavenumber and light intensity obtained in S2, and the output is the scalar concentration of the target component corresponding to the functional relationship; S32, during the regression training process, the scalar-pair regression model introduces a regularized curvature coarsening penalty, and the coefficients of the regularized curvature coarsening penalty term are automatically determined through generalized cross-validation to complete the model training and obtain the coefficient function of the trained scalar-pair regression model; S33, calculating the residual sum of squares, F-statistic, and their corresponding p-values ​​of the scalar-pair regression model, and judging the predictive significance of the model based on the p-values.

6. The method for processing trans-tissue migration of biological Raman spectra using scalar-pair function regression according to claim 1, characterized in that, S4 specifically involves: S41, inputting the preprocessed Raman spectral data of the biological tissue to be tested into the trained scalar-pair function regression model to obtain the predicted concentration of the target component, thus successfully completing the transfer of component information from easily accessible tissues to difficult-to-obtain tissues; S42, evaluating the accuracy of the predicted values ​​through statistical validation methods.

7. The method for processing trans-tissue migration of biological Raman spectra using scalar-pair function regression according to claim 6, characterized in that, S42 specifically involves: performing a grouped linear regression between the predicted target component concentration output from the scalar-pair function regression model and the actual measured target component concentration, and then using the coefficient of determination... Indicators validate the accuracy of predictions.

Images