Prediction method and device for pH value of interstitial fluid of tumor infiltration tissue and medium

Abstract

The invention relates to an interstitial fluid pH value prediction method of a tumor infiltration tissue, computer equipment and a storage medium. The method comprises the following steps: acquiring a surface enhanced Raman spectrum corresponding to the tumor infiltration tissue; preprocessing the surface enhanced Raman spectrum to obtain one-dimensional spectrum data; converting the one-dimensional spectral data into a corresponding two-dimensional recurrence plot; and inputting the two-dimensional recursion plot into a trained two-dimensional deep learning network model to obtain an interstitial fluid pH value of the tumor infiltration tissue output by the pre-trained two-dimensional deep learning network model. According to the method, the prediction accuracy of the pH value of the interstitial fluid of the tumor infiltration tissue can be improved.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to a method for predicting the pH value of interstitial fluid of tumor-infiltrated tissue, a computer device, and a storage medium. Background Art

[0002] Surgery is the main treatment option for brain gliomas, but the recurrence rate of gliomas after surgery is currently very high, mainly due to residual tumor-infiltrated tissue. Acidification of the interstitial fluid of tumor-infiltrated tissue is a common feature of solid tumors. If the pH value of the interstitial fluid can be measured non-destructively during surgery, it can help locate the malignant tissue in the glioma-infiltrated area and provide a new means to improve surgical prognosis.

[0003] Traditionally, the ratio of Raman peaks is used to determine the pH value of the interstitial fluid of tumor-infiltrating tissue. Specifically, the peak value at 311 cm-1 (peak 1) of the Raman spectrum remains almost unchanged as the solution becomes acidic, and the double peaks at 527 and 558 cm-1 (peak 2) gradually increase as the solution becomes acidic. Therefore, by dividing the area of ​​peak 2 by the area of ​​peak 1, the pH value of the Raman spectrum can be basically determined. However, calculating the area of ​​the peak may produce errors, and the interstitial fluid pH value obtained in this way has poor accuracy. Summary of the invention

[0004] Based on this, it is necessary to provide a method, computer equipment and storage medium for predicting the pH value of the interstitial fluid of tumor-infiltrating tissue in response to the above-mentioned technical problems, which can improve the prediction accuracy of the pH value of the interstitial fluid of tumor-infiltrating tissue.

[0005] A method for predicting the pH value of interstitial fluid of tumor-infiltrating tissue comprises: obtaining a surface enhanced Raman spectrum corresponding to the tumor-infiltrating tissue; preprocessing the surface enhanced Raman spectrum to obtain one-dimensional spectrum data; converting the one-dimensional spectrum data into a corresponding two-dimensional recursive graph; inputting the two-dimensional recursive graph into a trained two-dimensional deep learning network model to obtain the pH value of interstitial fluid of the tumor-infiltrating tissue output by the pre-trained two-dimensional deep learning network model.

[0006] In one of the embodiments, a method for predicting the pH value of interstitial fluid of tumor-infiltrated tissue also includes: obtaining a surface enhanced Raman spectrum corresponding to each sample tumor-infiltrated tissue in multiple sample tumor-infiltrated tissues to obtain multiple sample surface enhanced Raman spectra; preprocessing each sample surface enhanced Raman spectrum to obtain multiple one-dimensional sample spectrum data; converting each one-dimensional sample spectrum data into a corresponding two-dimensional sample recursive graph to obtain multiple two-dimensional sample recursive graphs; obtaining a training set, a validation set and a test set from the multiple two-dimensional sample recursive graphs, and the training set, the validation set and the test set all contain multiple two-dimensional sample recursive graphs; inputting the training set and the validation set into an untrained two-dimensional deep learning network model for model training to obtain a trained two-dimensional deep learning network model; inputting the test set into the trained two-dimensional deep learning network model to obtain a pH prediction value of the test set; if the pH prediction value of the test set matches the actual pH value of the test set, the trained two-dimensional deep learning network model is used as the trained two-dimensional deep learning network model.

[0007] In one embodiment, the surface enhanced Raman spectrum is preprocessed to obtain one-dimensional spectral data, including: intercepting the Raman spectrum within a first set range in the surface enhanced Raman spectrum to obtain a intercepted Raman spectrum; using an iterative adaptive weighted penalty least squares method to calculate the baseline of the intercepted Raman spectrum; subtracting the baseline from the intercepted Raman spectrum to obtain a corrected spectrum; dividing the corrected spectrum by a maximum peak value to obtain one-dimensional spectral data, wherein the maximum peak value is the maximum spectral value within a second set range in the corrected spectrum.

[0008] In one embodiment, converting the one-dimensional spectral data into a corresponding two-dimensional recursive graph comprises: i,j =||w i -w j ||; i, j = 1, 2…N converts the one-dimensional spectral data into the corresponding two-dimensional recursive graph; where the one-dimensional spectral data is W = (w 1 ,w 2 ,,,,w N ), w i is the intensity of the i-th point in the one-dimensional spectral data W, w j is the intensity of the jth point in the one-dimensional spectral data. There are a total of N points in the one-dimensional spectral data, where N is a positive integer. R i,j is a two-dimensional recurrence graph.

[0009] In one embodiment, a two-dimensional deep learning network model includes a first convolutional layer, a maximum pooling layer, a plurality of convolutional units connected in sequence, a global average pooling layer and a fully connected layer, wherein each convolutional unit includes a plurality of second convolutional layers connected in sequence and an adder; a first input end of the adder in any convolutional unit is connected to an output end of the last second convolutional layer in any convolutional unit, a second input end of the adder in any convolutional unit is connected to an input end of any convolutional unit, and an output end of the adder in any convolutional unit is connected to a next connection structure of any convolutional unit, and the next connection structure is a corresponding next convolutional unit or a global average pooling layer.

[0010] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any of the above-mentioned embodiments are implemented.

[0011] A computer-readable storage medium stores a computer program, which implements the steps of any of the above-mentioned method embodiments when the computer program is executed by a processor.

[0012] The above-mentioned method for predicting the pH value of the interstitial fluid of tumor-infiltrating tissue obtains the surface enhanced Raman spectrum corresponding to the tumor-infiltrating tissue; pre-processes the surface enhanced Raman spectrum to obtain one-dimensional spectral data; converts the one-dimensional spectral data into a corresponding two-dimensional recursive graph; inputs the two-dimensional recursive graph into a trained two-dimensional deep learning network model to obtain the pH value of the interstitial fluid of the tumor-infiltrating tissue output by the pre-trained two-dimensional deep learning network model. Therefore, the one-dimensional spectral data is converted into a two-dimensional recursive graph, and the two-dimensional recursive graph can reflect the internal characteristics of the one-dimensional spectral data. Then, the two-dimensional deep learning network model is used to mine the internal characteristic information of the spectral data in the two-dimensional recursive graph, so that the final predicted pH value of the interstitial fluid of the tumor-infiltrating tissue is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A diagram of an application environment of a method for predicting the pH value of interstitial fluid of tumor-infiltrated tissue in an embodiment;

[0014] Figure 2 A schematic diagram of a process for predicting the pH value of interstitial fluid of tumor-infiltrated tissue in an embodiment;

[0015] Figure 3 A schematic diagram of the spectrum of the surface enhanced Raman spectrum before and after interception in one embodiment;

[0016] Figure 4 is a spectrum schematic diagram of a corrected Raman spectrum in one embodiment;

[0017] Figure 5is a spectrum schematic diagram of a normalized spectrum in one embodiment;

[0018] Figure 6 A structural block diagram of a two-dimensional deep learning network model in an embodiment;

[0019] Figure 7 It is a structural diagram of a two-dimensional deep learning network model in a specific embodiment;

[0020] Figure 8 A comparison table of the effects of predicting pH values ​​corresponding to various indicators in an embodiment;

[0021] Fig. 9 A diagram showing the effect of a two-dimensional deep learning network model in predicting the pH value of Raman spectra compared to a one-dimensional network in one embodiment;

[0022] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0024] The present application provides a method for predicting the pH value of interstitial fluid in tumor-infiltrating tissue, which is applied to Figure 1 In the application environment shown. Figure 1 As shown, the terminal 100 is used to execute a method for predicting the pH value of interstitial fluid of tumor-infiltrated tissues of the present application. Specifically, Figure 1 As shown, the terminal 100 cooperates with manual operation to obtain the surface enhanced Raman spectrum corresponding to the tumor infiltrating tissue, and pre-processes the surface enhanced Raman spectrum to obtain one-dimensional spectrum data. Figure 1 As shown, the one-dimensional spectral data is converted into a corresponding two-dimensional recursive graph, and the two-dimensional recursive graph is input into the trained two-dimensional deep learning network model to obtain the interstitial fluid pH value of the tumor-infiltrating tissue output by the pre-trained two-dimensional deep learning network model. Therefore, the terminal 100 can directly output the interstitial fluid pH value of the tumor-infiltrating tissue.

[0025] In one embodiment, the present application provides a method for predicting the pH value of interstitial fluid in tumor-infiltrated tissue, which is applied to Figure 1 The terminal 100 shown in FIG. Figure 2 As shown, a method for predicting the pH value of interstitial fluid of tumor-infiltrated tissue comprises the following steps:

[0026] S202, obtaining a surface enhanced Raman spectrum corresponding to the tumor infiltrated tissue.

[0027] Raman spectroscopy is a spectral technique that can reflect the characteristics of chemical bonds within molecules. When the target molecule is adsorbed on the surface of a noble metal, the Raman signal can be greatly enhanced. This phenomenon is called surface enhanced Raman scattering (SERS). The SERS spectra of some molecules have a ratiometric response to changes in environmental pH. The pH value of the environment in which they are located can be determined by analyzing the SERS spectra of these molecules, thus providing a reference for the complete removal of gliomas during surgery.

[0028] In this embodiment, a Raman scanner is used to scan the SERS chip to obtain a surface enhanced Raman spectrum corresponding to the tumor infiltrated tissue.

[0029] S204, preprocessing the surface enhanced Raman spectrum to obtain one-dimensional spectrum data.

[0030] In this embodiment, the surface enhanced Raman spectrum is a one-dimensional sequence of spectrum data. The one-dimensional spectrum data is obtained after preprocessing the surface enhanced Raman spectrum.

[0031] The one-dimensional spectral data obtained by preprocessing the surface enhanced Raman spectrum includes: intercepting the Raman spectrum within a first set range in the surface enhanced Raman spectrum to obtain a intercepted Raman spectrum; using an iterative adaptive weighted penalty least squares method to calculate the baseline of the intercepted Raman spectrum; subtracting the baseline from the intercepted Raman spectrum to obtain a corrected spectrum; dividing the corrected spectrum by the maximum peak value to obtain one-dimensional spectral data, wherein the maximum peak value is the maximum spectral value within a second set range in the corrected spectrum.

[0032] Specifically: First, for each surface enhanced Raman spectrum, the portion between 217 cm-1 and 1648 cm-1, i.e., the 50th point to the 336th point, is intercepted to obtain the intercepted Raman spectrum. The intercepted Raman spectrum is as follows: Figure 3 As shown. Secondly, the baseline of the intercepted Raman spectrum is calculated using the adaptive iteratively reweighted penalized least squares (airPLS) algorithm, and then the baseline is subtracted from the intercepted Raman spectrum to obtain the corrected spectrum. The corrected spectrum is shown in Figure 4 Finally, the corrected spectrum is divided by the maximum peak between 200 cm-1 and 400 cm-1 of the spectrum to obtain a normalized spectrum. The normalized spectrum is shown in Figure 5 The first setting range refers to the spectrum range between 217cm-1 and 1648cm-1, the second setting range refers to the spectrum range between 200cm-1 and 400cm-1, and the one-dimensional spectrum data refers to the final normalized spectrum.

[0033] Therefore, through the above preprocessing, the data in the Raman spectrum is reduced in dimension, unnecessary redundant parts in the Raman spectrum are removed, and the noise in the spectrum is filtered, which is conducive to improving the subsequent detection accuracy of pH value using a two-dimensional deep learning network model.

[0034] S206, converting the one-dimensional spectrum data into a corresponding two-dimensional recursive graph.

[0035] In this embodiment, the one-dimensional spectral data is a one-dimensional sequence, and the one-dimensional spectral data is converted into a corresponding two-dimensional recursive graph, and the internal structure of the one-dimensional sequence is reflected by the two-dimensional recursive graph. The recursive graph was originally a method for analyzing the periodicity, chaos and non-stationarity of a time series, and the method can reflect the internal structure of a one-dimensional sequence. Since the Raman spectrum is also a one-dimensional sequence, the internal characteristics of the Raman spectrum can be reflected by converting the one-dimensional spectral data into a two-dimensional recursive graph. Converting the one-dimensional spectral data into a corresponding two-dimensional recursive graph enables the two-dimensional deep learning network model to mine its inherent richer feature information and improve the prediction performance of the pH value.

[0036] The one-dimensional spectral data is converted into the corresponding two-dimensional recursive graph, including: according to R i,j =||w i -w j ||; i, j = 1, 2…N converts the one-dimensional spectral data into the corresponding two-dimensional recursive graph; where the one-dimensional spectral data is W = (w 1 ,w 2 ,,,,w N ), w i is the intensity of the i-th point in the one-dimensional spectral data W, w j is the intensity of the jth point in the one-dimensional spectral data. There are a total of N points in the one-dimensional spectral data, where N is a positive integer. R i,j is a two-dimensional recurrence graph.

[0037] Specifically, for the one-dimensional spectrum data of Raman spectroscopy W = (w 1 ,w 2 ,...,w N ), the corresponding two-dimensional recursive graph is:

[0038] R i,j =||w i -w j ||,i,j=1,2,...,N

[0039] Among them, w i is the intensity of the i-th point in the one-dimensional spectral data W, w j is the intensity of the jth point in the one-dimensional spectral data. There are a total of N points in the one-dimensional spectral data, where N is a positive integer. R i,jis a two-dimensional recursive graph. As can be seen from the above formula, the recursive graph R i,j The spectral data of the Raman spectrum with a size of 1*N is converted into a two-dimensional image with a size of N*N, and each point in the image represents the absolute value of the intensity difference between two points in the spectrum. Therefore, the two-dimensional recursion graph contains richer internal information in the Raman spectrum.

[0040] S208, inputting the two-dimensional recursive graph into the trained two-dimensional deep learning network model to obtain the interstitial fluid pH value of the tumor infiltrated tissue output by the pre-trained two-dimensional deep learning network model.

[0041] In this embodiment, a regression model of a two-dimensional convolutional neural network, i.e., a two-dimensional deep learning network model, is constructed. Specifically, Figure 6 As shown, the two-dimensional deep learning network model includes a first convolutional layer, a maximum pooling layer, a plurality of convolutional units connected in sequence, a global average pooling layer and a fully connected layer, wherein each convolutional unit includes a plurality of second convolutional layers connected in sequence and an adder; a first input end of the adder in any convolutional unit is connected to an output end of the last second convolutional layer in any convolutional unit, a second input end of the adder in any convolutional unit is connected to an input end of any convolutional unit, and an output end of the adder in any convolutional unit is connected to a next connection structure of any convolutional unit, and the next connection structure is a corresponding next convolutional unit or a global average pooling layer.

[0042] In one example, the plurality of sequentially connected convolution units include a first convolution unit, a second convolution unit, a third convolution unit, and a fourth convolution unit that are sequentially connected, each convolution unit includes a plurality of convolution blocks, and each convolution block includes two second convolution layers.

[0043] Specifically, Figure 7 As shown in Figure 1, the two-dimensional deep learning network model contains several convolutional layers, a maximum pooling layer, a global average pooling layer, and a fully connected layer. The specific connection order is: first, the convolutional layer is followed by the maximum pooling layer, and then the convolutional blocks of the four stages are stacked. Each convolutional block contains two convolutional layers and adds the input to the output after the two convolutional layers. The number of convolutional blocks in the four stages is 3, 4, 6, and 3 respectively. Finally, the global average pooling layer and the fully connected layer are connected to obtain the predicted pH value output.

[0044] Compared with the two-dimensional deep learning network model constructed in this embodiment, the one-dimensional network structure is simple and cannot extract rich features inside the spectrum from the one-dimensional spectral data, resulting in limited performance of directly using one-dimensional spectral data to predict pH values. The two-dimensional deep learning network model constructed in this embodiment can mine the internal features of the Raman spectrum in the two-dimensional recursive graph, and the two-dimensional deep learning network model constructed in this embodiment can predict the pH value corresponding to the Raman spectrum more accurately.

[0045] In order to prove that the two-dimensional deep learning network model constructed in this embodiment can predict the pH value corresponding to the Raman spectrum more accurately, Figure 8 As shown, the present application compares the effects of using one-dimensional network and two-dimensional deep learning network models to predict the pH value corresponding to the Raman spectrum through multiple indicators.

[0046] like Figure 8 As shown in the table, the two-dimensional deep learning network model has obvious improvements in the three indicators of different pH values, namely the mean absolute error MAE, standard deviation SD, and sum of squared errors SSE, compared with the one-dimensional network. Fig. 9 It is more intuitive that the prediction results of the two-dimensional deep learning network model are more stable and accurate, indicating that the two-dimensional deep learning network model can more effectively and accurately predict the pH value of Raman spectra than the one-dimensional network.

[0047] In one embodiment, before the above-mentioned step of inputting the two-dimensional recursive graph into the trained two-dimensional deep learning network model, it also includes: obtaining a surface enhanced Raman spectrum corresponding to each sample tumor infiltrated tissue in multiple sample tumor infiltrated tissues to obtain multiple sample surface enhanced Raman spectra; preprocessing each sample surface enhanced Raman spectrum to obtain multiple one-dimensional sample spectrum data; converting each one-dimensional sample spectrum data into a corresponding two-dimensional sample recursive graph to obtain multiple two-dimensional sample recursive graphs; obtaining a training set, a validation set and a test set from the multiple two-dimensional sample recursive graphs, and the training set, the validation set and the test set all contain multiple two-dimensional sample recursive graphs; inputting the training set and the validation set into an untrained two-dimensional deep learning network model for model training to obtain a trained two-dimensional deep learning network model; inputting the test set into the trained two-dimensional deep learning network model to obtain a pH prediction value of the test set; if the pH prediction value of the test set matches the actual pH value of the test set, the trained two-dimensional deep learning network model is used as the trained two-dimensional deep learning network model.

[0048] The above embodiments provide a method for predicting the interstitial fluid pH value of tumor-infiltrating tissue, which converts one-dimensional spectral data into a two-dimensional recursive graph. The two-dimensional recursive graph can reflect the internal characteristics of the one-dimensional spectral data, and then uses a two-dimensional deep learning network model to mine the internal characteristic information of the spectral data in the two-dimensional recursive graph, so that the final predicted interstitial fluid pH value of the tumor-infiltrating tissue is more accurate.

[0049] The above embodiments provide a method for predicting the pH value of the interstitial fluid of tumor-infiltrated tissue. In the one-dimensional spectral data, each point only has information about its neighboring points, and the features are relatively simple. In the two-dimensional recursive graph, each point itself contains information between two points, and its neighboring points are more diverse, so it can reflect richer features within the spectrum. Furthermore, a two-dimensional deep learning network model is used to mine the richer feature information within the two-dimensional recursive graph to improve the prediction performance of the pH value.

[0050] It should be understood that, although the various steps in the flow chart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flow chart may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0051] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.10 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for measuring the pH value of interstitial fluid of tumor-infiltrated tissue is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0052] Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0053] In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program: obtaining a surface enhanced Raman spectrum corresponding to tumor-infiltrating tissue; preprocessing the surface enhanced Raman spectrum to obtain one-dimensional spectral data; converting the one-dimensional spectral data into a corresponding two-dimensional recursive graph; inputting the two-dimensional recursive graph into a trained two-dimensional deep learning network model to obtain the interstitial fluid pH value of the tumor-infiltrating tissue output by the pre-trained two-dimensional deep learning network model.

[0054] In one of the embodiments, the processor implements the following steps when executing the computer program: obtaining a surface enhanced Raman spectrum corresponding to each sample tumor infiltrate tissue in multiple sample tumor infiltrate tissues to obtain multiple sample surface enhanced Raman spectra; preprocessing each sample surface enhanced Raman spectrum to obtain multiple one-dimensional sample spectrum data; converting each one-dimensional sample spectrum data into a corresponding two-dimensional sample recursive graph to obtain multiple two-dimensional sample recursive graphs; obtaining a training set, a validation set and a test set from the multiple two-dimensional sample recursive graphs, and the training set, the validation set and the test set all contain multiple two-dimensional sample recursive graphs; inputting the training set and the validation set into an untrained two-dimensional deep learning network model for model training to obtain a trained two-dimensional deep learning network model; inputting the test set into the trained two-dimensional deep learning network model to obtain a pH prediction value of the test set; if the pH prediction value of the test set matches the actual pH value of the test set, the trained two-dimensional deep learning network model is used as the trained two-dimensional deep learning network model.

[0055] In one embodiment, when the processor executes a computer program to implement the above-mentioned step of preprocessing the surface enhanced Raman spectrum to obtain one-dimensional spectral data, the following steps are specifically implemented: intercepting the Raman spectrum within a first set range in the surface enhanced Raman spectrum to obtain a intercepted Raman spectrum; using an iterative adaptive weighted penalty least squares method to calculate the baseline of the intercepted Raman spectrum; subtracting the baseline from the intercepted Raman spectrum to obtain a corrected spectrum; dividing the corrected spectrum by the maximum peak value to obtain one-dimensional spectral data, wherein the maximum peak value is the maximum spectral value within a second set range in the corrected spectrum.

[0056] In one embodiment, when the processor executes the computer program to implement the above-mentioned step of converting the one-dimensional spectral data into the corresponding two-dimensional recursive graph, the following steps are specifically implemented: according to R i,j =||w i -w j ||; i, j = 1, 2…N converts the one-dimensional spectral data into the corresponding two-dimensional recursive graph; where the one-dimensional spectral data is W = (w 1 ,w 2 ,,,,w N), w i is the intensity of the i-th point in the one-dimensional spectral data W, w j is the intensity of the jth point in the one-dimensional spectral data. There are a total of N points in the one-dimensional spectral data, where N is a positive integer. R i,j is a two-dimensional recurrence graph.

[0057] In one embodiment, a two-dimensional deep learning network model includes a first convolutional layer, a maximum pooling layer, a plurality of convolutional units connected in sequence, a global average pooling layer and a fully connected layer, wherein each convolutional unit includes a plurality of second convolutional layers connected in sequence and an adder; a first input end of the adder in any convolutional unit is connected to an output end of the last second convolutional layer in any convolutional unit, a second input end of the adder in any convolutional unit is connected to an input end of any convolutional unit, and an output end of the adder in any convolutional unit is connected to a next connection structure of any convolutional unit, and the next connection structure is a corresponding next convolutional unit or a global average pooling layer.

[0058] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: obtaining a surface enhanced Raman spectrum corresponding to tumor-infiltrating tissue; preprocessing the surface enhanced Raman spectrum to obtain one-dimensional spectral data; converting the one-dimensional spectral data into a corresponding two-dimensional recursive graph; inputting the two-dimensional recursive graph into a trained two-dimensional deep learning network model to obtain the interstitial fluid pH value of the tumor-infiltrating tissue output by the pre-trained two-dimensional deep learning network model.

[0059] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: obtaining a surface enhanced Raman spectrum corresponding to each sample tumor infiltrate tissue in multiple sample tumor infiltrate tissues to obtain multiple sample surface enhanced Raman spectra; preprocessing each sample surface enhanced Raman spectrum to obtain multiple one-dimensional sample spectrum data; converting each one-dimensional sample spectrum data into a corresponding two-dimensional sample recursive graph to obtain multiple two-dimensional sample recursive graphs; obtaining a training set, a validation set and a test set from the multiple two-dimensional sample recursive graphs, and the training set, the validation set and the test set all contain multiple two-dimensional sample recursive graphs; inputting the training set and the validation set into an untrained two-dimensional deep learning network model for model training to obtain a trained two-dimensional deep learning network model; inputting the test set into the trained two-dimensional deep learning network model to obtain a pH prediction value of the test set; if the pH prediction value of the test set matches the actual pH value of the test set, the trained two-dimensional deep learning network model is used as the trained two-dimensional deep learning network model.

[0060] In one embodiment, when a computer program is executed by a processor to implement the above-mentioned step of preprocessing the surface enhanced Raman spectrum to obtain one-dimensional spectral data, the following steps are specifically implemented: intercepting the Raman spectrum within a first set range in the surface enhanced Raman spectrum to obtain a intercepted Raman spectrum; using an iterative adaptive weighted penalty least squares method to calculate the baseline of the intercepted Raman spectrum; subtracting the baseline from the intercepted Raman spectrum to obtain a corrected spectrum; dividing the corrected spectrum by the maximum peak value to obtain one-dimensional spectral data, wherein the maximum peak value is the maximum spectral value within a second set range in the corrected spectrum.

[0061] In one embodiment, when the computer program is executed by the processor to implement the above-mentioned step of converting the one-dimensional spectral data into the corresponding two-dimensional recursive graph, the following steps are specifically implemented: according to R i,j =||w i -w j ||; i, j = 1, 2…N converts the one-dimensional spectral data into the corresponding two-dimensional recursive graph; where the one-dimensional spectral data is W = (w 1 ,w 2 ,,,,w N ), w i is the intensity of the i-th point in the one-dimensional spectral data W, w j is the intensity of the jth point in the one-dimensional spectral data. There are a total of N points in the one-dimensional spectral data, where N is a positive integer. R i,j is a two-dimensional recurrence graph.

[0062] In one embodiment, a two-dimensional deep learning network model includes a first convolutional layer, a maximum pooling layer, a plurality of convolutional units connected in sequence, a global average pooling layer and a fully connected layer, wherein each convolutional unit includes a plurality of second convolutional layers connected in sequence and an adder; a first input end of the adder in any convolutional unit is connected to an output end of the last second convolutional layer in any convolutional unit, a second input end of the adder in any convolutional unit is connected to an input end of any convolutional unit, and an output end of the adder in any convolutional unit is connected to a next connection structure of any convolutional unit, and the next connection structure is a corresponding next convolutional unit or a global average pooling layer.

[0063] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0064] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for predicting the pH value of interstitial fluid in tumor-infiltrated tissues. It is characterized in that The method comprises: Obtaining surface-enhanced Raman spectra corresponding to tumor-infiltrated tissues; Preprocessing the surface enhanced Raman spectrum to obtain one-dimensional spectrum data; Converting the one-dimensional spectral data into a corresponding two-dimensional recurrence graph; The two-dimensional recursive graph is input into a trained two-dimensional deep learning network model to obtain the interstitial fluid pH value of the tumor-infiltrated tissue output by the pre-trained two-dimensional deep learning network model.

2. The method according to claim 1, It is characterized in that The method further comprises: Acquire a surface enhanced Raman spectrum corresponding to each sample tumor infiltrating tissue in a plurality of sample tumor infiltrating tissues to obtain a plurality of sample surface enhanced Raman spectra; Preprocessing the surface enhanced Raman spectrum of each sample to obtain multiple one-dimensional sample spectrum data; Convert each one-dimensional sample spectrum data into a corresponding two-dimensional sample recursion graph to obtain a plurality of two-dimensional sample recursion graphs; Acquire a training set, a validation set, and a test set from a plurality of two-dimensional sample recursive graphs, wherein the training set, the validation set, and the test set all contain a plurality of two-dimensional sample recursive graphs; Inputting the training set and the validation set into an untrained two-dimensional deep learning network model to perform model training, thereby obtaining a trained two-dimensional deep learning network model; Inputting the test set into the trained two-dimensional deep learning network model to obtain a pH prediction value of the test set; If the predicted pH value of the test set matches the actual pH value of the test set, the trained two-dimensional deep learning network model is used as the trained two-dimensional deep learning network model.

3. The method according to claim 1, It is characterized in that The one-dimensional spectral data obtained by preprocessing the surface enhanced Raman spectrum includes: intercepting a Raman spectrum within a first set range of the surface enhanced Raman spectrum to obtain an intercepted Raman spectrum; The baseline of the intercepted Raman spectrum is calculated by using an iterative adaptive weighted penalty least square method; Subtracting the baseline from the intercepted Raman spectrum to obtain a corrected spectrum; The corrected spectrum is divided by the maximum peak value to obtain the one-dimensional spectrum data, wherein the maximum peak value is the maximum spectrum value within the second set range in the corrected spectrum.

4. The method according to claim 1, It is characterized in that The step of converting the one-dimensional spectral data into a corresponding two-dimensional recursive graph comprises: According to R i，j = ||w i - w j ||; i, j = 1, 2....N to convert the one-dimensional spectral data into a corresponding two-dimensional recurrence plot; Among them, the one-dimensional spectral data is W = (w 1 , w 2 ,,,,,w N ), w i is the intensity of the i-th point in the one-dimensional spectral data W, w j is the intensity of the jth point in the one-dimensional spectral data. There are a total of N points in the one-dimensional spectral data, where N is a positive integer. R i，j It is a two-dimensional recurrence graph.

5. The method according to claim 1, It is characterized in that The two-dimensional deep learning network model includes a first convolutional layer, a maximum pooling layer, a plurality of convolutional units connected in sequence, a global average pooling layer and a fully connected layer, wherein each convolutional unit includes a plurality of second convolutional layers and an adder connected in sequence; A first input end of the adder in any convolution unit is connected to the output end of the last second convolution layer in any convolution unit, a second input end of the adder in any convolution unit is connected to the input end of any convolution unit, and an output end of the adder in any convolution unit is connected to the next connection structure of any convolution unit, and the next connection structure is the corresponding next convolution unit or the global average pooling layer.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images