Lymph node state judgment method and system based on collagen characteristics, terminal and storage medium

Abstract

The invention discloses a collagen feature-based lymph node state judgment method and system, a terminal and a storage medium, and the method comprises the steps: obtaining a pancreatic duct gland data sample, and carrying out the collection of a multi-photon image, and obtaining a target multi-photon image; obtaining a macroscopic collagen feature in the target multi-photon image, and calculating a first score of the macroscopic collagen feature; obtaining a microscopic collagen feature in the target multi-photon image, and calculating a second score of the microscopic collagen feature; constructing a column graph model according to the first score and the second score; and acquiring current pancreatic duct gland data of the patient, and inputting the data into the column graph model to obtain a lymph node metastasis result. According to the method, the macroscopic collagen features and the microcosmic collagen features in the pancreatic duct gland data sample are extracted and combined to obtain the column graph model, and accurate judgment of lymph node metastasis can be achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to a lymph node state judgment method and system based on collagen characteristics, a terminal and a computer readable storage medium. BACKGROUND

[0002] For the determination of the lymph node state, the existing technology generally adopts various imaging technologies, including computed tomography, magnetic resonance imaging and positron emission tomography, which can help preoperative assessment of the state of the lymph node.

[0003] However, these imaging methods are based on morphology (such as size, shape and intensity changes) to evaluate the state of the lymph node, which can result in a high false positive rate, thereby leading to low accuracy of the obtained lymph node state results.

[0004] Therefore, the prior art still needs to be improved and developed. SUMMARY

[0005] The main purpose of the present application is to provide a lymph node state judgment method and system based on collagen characteristics, a terminal and a computer readable storage medium, which aims to solve the problem of high false positive rate in the prior art based on morphology to evaluate the state of the lymph node, thereby leading to low accuracy of the obtained lymph node state results.

[0006] To achieve the above purpose, the present application provides a lymph node state judgment method based on collagen characteristics, which comprises the following steps:

[0007] Obtain a pancreatic ductal gland data sample, and perform multi-photon image acquisition on the pancreatic ductal gland data sample to obtain a target multi-photon image;

[0008] Obtain macroscopic collagen characteristics in the target multi-photon image, and calculate a first score of the macroscopic collagen characteristics;

[0009] Obtain microscopic collagen characteristics in the target multi-photon image, and calculate a second score of the microscopic collagen characteristics;

[0010] Construct a nomogram model according to the first score and the second score;

[0011] Obtain current pancreatic ductal gland data of a patient, input the current pancreatic ductal gland data into the nomogram model, and obtain a lymph node metastasis result.

[0012] Optionally, the lymph node state judgment method based on collagen characteristics, wherein the pancreatic ductal gland data sample is obtained, and the multi-photon image acquisition is performed on the pancreatic ductal gland data sample to obtain a target multi-photon image, specifically comprising:

[0013] obtaining a pancreatic ductal adenocarcinoma data sample, and dividing the pancreatic ductal adenocarcinoma data sample into a training set and a validation set;

[0014] performing multi-photon image acquisition on the training set by using a multi-photon microscope to obtain a target multi-photon image.

[0015] Optionally, the collagen feature-based lymph node state judgment method, wherein the multi-photon microscope comprises a first synchronous channel and a second synchronous channel.

[0016] The multi-photon image acquisition on the training set by using a multi-photon microscope to obtain a target multi-photon image specifically comprises:

[0017] performing second-harmonic signal detection processing on the training set through the first synchronous channel to obtain a first scanning image;

[0018] performing two-photon excitation fluorescence signal detection processing on the training set through the second synchronous channel to obtain a second scanning image;

[0019] performing image splicing processing on the first scanning image and the second scanning image to obtain a target multi-photon image.

[0020] Optionally, the collagen feature-based lymph node state judgment method, wherein the obtaining of the macroscopic collagen feature in the target multi-photon image and the calculation of a first score of the macroscopic collagen feature specifically comprises:

[0021] obtaining a stained section image, and determining a region of interest in the stained section image;

[0022] performing annotation processing on the region of interest in the stained section image to obtain an annotated region;

[0023] performing image acquisition on the annotated region in the target multi-photon image to obtain a second-harmonic signal image and a two-photon excitation fluorescence signal image;

[0024] obtaining a macroscopic collagen feature in the second-harmonic signal image and the two-photon excitation fluorescence signal image, and performing score calculation on the macroscopic collagen feature by using a cross-validation ridge regression analysis method to obtain a first score.

[0025] Optionally, the collagen feature-based lymph node state judgment method, wherein the obtaining of the microscopic collagen feature in the target multi-photon image and the calculation of a second score of the microscopic collagen feature specifically comprises:

[0026] performing region of interest extraction processing on the target multi-photon image to obtain a target region of interest.

[0027] extracting micro-collagen features in the target region of interest, wherein the micro-collagen features include morphological features, histogram-based features, gray level co-occurrence matrix features, and Gabor wavelet transform features;

[0028] calculating scores of the micro-collagen features by using a LASSO regression method to obtain second scores.

[0029] Optionally, the lymph node state judgment method based on collagen features, wherein the constructing a nomogram model according to the first scores and the second scores specifically comprises:

[0030] determining a Youden index threshold, and dividing the patient types in the training set into lymph node metastasis negative and lymph node metastasis positive according to the first scores, the second scores, and the Youden index threshold;

[0031] calculating a first corresponding relationship between the first scores and the second scores and the lymph node metastasis negative, and a second corresponding relationship between the first scores and the second scores and the lymph node metastasis positive by using a single factor and multi-factor logistic regression analysis method, and constructing a nomogram model according to the first corresponding relationship and the second corresponding relationship.

[0032] Optionally, the lymph node state judgment method based on collagen features, wherein the constructing a nomogram model according to the first scores and the second scores further comprises:

[0033] performing model performance discrimination on the nomogram model by using an ROC analysis method, and performing model performance verification on the nomogram model according to the verification set to obtain a model performance result.

[0034] In addition, to achieve the above-mentioned purposes, the application further provides a lymph node state judgment system based on collagen features, wherein the lymph node state judgment system based on collagen features comprises:

[0035] a multi-photon image acquisition module, configured to acquire a pancreatic ductal adenocarcinoma data sample, and perform multi-photon image acquisition on the pancreatic ductal adenocarcinoma data sample to obtain a target multi-photon image;

[0036] a macro-feature scoring module, configured to acquire macro-collagen features in the target multi-photon image, and calculate first scores of the macro-collagen features;

[0037] a micro-feature scoring module, configured to acquire micro-collagen features in the target multi-photon image, and calculate second scores of the micro-collagen features;

[0038] a nomogram model construction module configured to construct a nomogram model according to the first score and the second score;

[0039] a lymph node metastasis result output module configured to obtain current pancreatic ductal adenocarcinoma data of a patient, input the current pancreatic ductal adenocarcinoma data into the nomogram model, and obtain a lymph node metastasis result.

[0040] In the present application, pancreatic ductal adenocarcinoma data samples are obtained, and multi-photon image acquisition is performed on the pancreatic ductal adenocarcinoma data samples to obtain a target multi-photon image. Macroscopic collagen features in the target multi-photon image are obtained, and a first score of the macroscopic collagen features is calculated. Microscopic collagen features in the target multi-photon image are obtained, and a second score of the microscopic collagen features is calculated. A nomogram model is constructed according to the first score and the second score. Current pancreatic ductal adenocarcinoma data of a patient is obtained, and the current pancreatic ductal adenocarcinoma data is input into the nomogram model to obtain a lymph node metastasis result. By extracting macroscopic collagen features and microscopic collagen features in the pancreatic ductal adenocarcinoma data samples and calculating the respective scores of the macroscopic collagen features and the microscopic collagen features, and then constructing a nomogram model according to the scores, the present application can more comprehensively understand the relationship between collagen protein features and lymph node metastasis, and effectively improves the accuracy of lymph node metastasis result output. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 is a flowchart of a preferred embodiment of the lymph node state judgment method based on collagen features of the present application;

[0042] Figure 2 is a flowchart of a preferred embodiment of the lymph node state judgment method based on collagen features of the present application;

[0043] Figure 3 is a flowchart of a preferred embodiment of the lymph node state judgment method based on collagen features of the present application;

[0044] Figure 4 is a flowchart of a preferred embodiment of the lymph node state judgment method based on collagen features of the present application;

[0045] Figure 5 is a flowchart of a preferred embodiment of the lymph node state judgment method based on collagen features of the present application;

[0046] Figure 6 is a flowchart of a preferred embodiment of the lymph node state judgment method based on collagen features of the present application;

[0047] Figure 7 FIG. 10 is a schematic diagram of ROC curves of clinical models (A, B) and ma-TACS scores (C, D) in the training and validation cohorts of the preferred embodiment of the lymph node status determination method based on collagen features of the present application;

[0048] Figure 8 FIG. 11 is a schematic diagram of ROC curves of mi-TACS score models (E, F) and ma / mi-TACS scores (G, H) in the training and validation cohorts of the preferred embodiment of the lymph node status determination method based on collagen features of the present application;

[0049] Figure 9 FIG. 12 is a schematic diagram of correlation analysis between the six collagen features selected by LASSO regression and lymph node metastasis in the training (A) and validation cohorts (B) of the preferred embodiment of the lymph node status determination method based on collagen features of the present application;

[0050] Figure 10 FIG. 13 is a schematic diagram of score differences between LN0 and LN1 groups in the training and validation cohorts of the five models of the preferred embodiment of the lymph node status determination method based on collagen features of the present application;

[0051] Figure 11 FIG. 14 is a schematic diagram of ROC curves (I, J) of the complete model (clinical + ma / mi-TACS scores) in the training and validation cohorts of the preferred embodiment of the lymph node status determination method based on collagen features of the present application;

[0052] Figure 12 FIG. 15 is a schematic diagram of prediction results of lymph node metastasis by the five models in the training and validation cohorts of the preferred embodiment of the lymph node status determination method based on collagen features of the present application;

[0053] Figure 13 FIG. 16 is a schematic diagram of nomograms for prediction by integrating ma-TACS and mi-TACS scores, and a schematic diagram of nomogram calibration curves for the training and validation cohorts of the preferred embodiment of the lymph node status determination method based on collagen features of the present application;

[0054] Figure 14 FIG. 17 is a schematic diagram of DCA for each model of the preferred embodiment of the lymph node status determination method based on collagen features of the present application;

[0055] Figure 15 FIG. 18 is a schematic diagram of the structure of the preferred embodiment of the lymph node status determination system of the present application;

[0056] Figure 16 FIG. 19 is a schematic diagram of the structure of the preferred embodiment of the terminal of the present application. DETAILED DESCRIPTION

[0057] In order to make the objectives, technical solutions, and advantages of the present application clearer and more explicit, the present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0058] In the present application, pancreatic ductal adenocarcinoma is exemplified. Pancreatic ductal adenocarcinoma (PDAC) is a highly invasive and lethal cancer with significant metastatic potential. The status of lymph node is crucial for determining the treatment plan and predicting the prognosis of pancreatic cancer patients. However, the current methods for evaluating lymph node metastasis of pancreatic cancer are not perfect.

[0059] The lymph node status of pancreatic cancer is an important predictor of survival after pancreatic cancer resection, which can determine the surgical procedure and prognosis of patients. For most patients undergoing pancreatic cancer surgery, lymph node dissection is usually a necessary step to achieve radical resection. Some patients without lymph node metastasis still need to undergo lymph node dissection and bear the related risks. Therefore, accurate assessment of the lymph node status of pancreatic cancer patients is crucial for developing personalized surgical plans.

[0060] Various imaging techniques, including computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET), can help preoperative evaluation of the LN status of pancreatic cancer. These imaging methods evaluate lymph node status based on morphology (e.g., size, shape, and intensity changes), which often results in a high false-positive rate. Conversely, normal-sized lymph nodes can contain micrometastases, leading to a high false-negative rate. These methods cannot provide effective clinical guidance, and imaging features are not sufficient as predictors. Pancreatic cancer with higher tumor grade, larger volume, and lymphatic vessel invasion is considered to have a higher risk of lymph node metastasis, suggesting the need for radical surgery, but there is currently no consensus. Various models have been developed in the prior art to assess the likelihood of lymph node metastasis in pancreatic cancer. However, these models mainly focus on clinical pathological variables and do not study the relationship between the tumor microenvironment and lymph node metastasis.

[0061] The extracellular matrix (ECM) constitutes the scaffold of the tumor microenvironment and plays a crucial role in regulating cancer behavior. Collagen fibers, glycoproteins, and proteoglycans are the main structural components of the ECM. Collagen fibers are key components of the ECM and can affect cell motility, both as a barrier to invasion and as a pathway for rapid tumor cell migration depending on their arrangement. PDAC is characterized by fibrotic tissue and abundant ECM surrounding the tumor. This extensive fibrotic stroma can account for up to 90% of the total tumor volume. Fibrous collagen constitutes a large part of the PDAC stroma and is associated with tumor cell migration, proliferation, and invasion. However, the specific role of collagen fibers in PDAC lymph node metastasis remains unclear.

[0062] The development of nonlinear microscopy, such as multiphoton microscopy (MPM), has enabled the visualization of changes in collagen fibrils with cellular resolution. By combining two-photon excited fluorescence (TPEF) with second harmonic generation (SHG), MPM can provide detailed histological information from untreated samples. SHG imaging provides a direct, label-free method for visualizing collagen structure. Collagen fibrils have a unique noncentrosymmetric structure, enabling them to act as frequency doublers when exposed to multiphoton laser light. By exploiting coherent, non-absorptive light interactions, high-resolution imaging of individual collagen fibrils can be achieved without external staining. Several groups have demonstrated that changes in collagen fibril structure and organization during tumorigenesis have important biological implications, with these changes correlating with clinical outcomes in various solid tumors. Specifically, several collagen organization patterns, termed tumor-associated collagen signatures (TACS), have been identified in breast tumors and have important implications for disease progression. Previous studies have expanded upon the original TACS1-3 by identifying additional patterns, TACS4-8, specifically at the invasive front of primary breast tumors. Notably, the present invention establishes strong correlations between various TACS patterns and breast cancer prognosis. Furthermore, these structural patterns are not unique to breast cancer. Recent studies have shown that these collagen structures are also common in pancreatic ductal adenocarcinoma (PDAC), another highly metastatic fibroproliferative disease. Despite recent studies demonstrating the presence of TACS patterns in PDAC, the relationship between TACS and PDAC progression, particularly lymph node metastasis, remains largely unexplored. Building on our previous research, we explored macroscopic TACS and liquid nitrogen metastasis. Recently, extracting detailed quantitative features from SHG images has become a rapidly developing field, helping to reveal the connection between microscopic collagen signatures and various pathological conditions. Therefore, to better understand the relationship between subtle changes in TACS and PDAC lymph node metastasis, we extracted microscopic TACS signatures (mi-TACS) and combined them with ma-TACS, enabling us to more comprehensively understand the relationship between collagen signatures and lymph node metastasis. Finally, we developed and validated a nomogram integrating ma-TACS and mi-TACS from MPM imaging to provide personalized prediction of lymph node metastasis in PDAC patients.

[0063] To solve the above problems, the present application develops and verifies a new nomogram that integrates macro- and micro-collagen features (ma-TACS and mi-TACS) in the tumor microenvironment to assess the risk of lymph node metastasis in PDAC patients. This retrospective study included 150 PDAC patients, of which 92 were in the training cohort and 58 were in the validation cohort. ma-TACS and mi-TACS were obtained by multiphoton microscopy. mi-TACS includes morphological and textural features, which are extracted from segmented regions of interest using matlab 2022a. Ridge regression and LASSO regression analysis were used to calculate ma-TACS and mi-TACS scores.

[0064] Analysis results: Both univariate and multivariate logistic regression analyses showed that ma-TACS and mi-TACS scores were significantly associated with lymph node metastasis (ma-TACS score, odds ratio (odds ratio is an index that measures the strength of association between two binary variables, for example, this value is 2.304, which is greater than 1, indicating that the higher the value, the greater the likelihood of lymph node metastasis, and the likelihood of lymph node metastasis increases 2.567 times for every unit increase in this score), 2.304; 95% CI (95% confidence interval for the value 2.034), 1.412-3.761; P=0.001; 2.934 (same meaning as 2.304, these two values represent the odds ratio obtained in univariate and multivariate analyses, respectively, indicating that both univariate and multivariate analyses are meaningful), 1.409-6.108, P=0.004 (this P value indicates whether the statistics are statistically significant, P<0.05 is statistically significant, indicating that the score is significantly associated with lymph node metastasis); mi-TACS score, odds ratio, 3.325; 95% CI, 2.296-4.814; P<0.001; 3.861, 2.488-5.993, P<0.001). The nomogram model integrating ma-TACS and mi-TACS scores successfully classified patients into lymph node-negative and positive groups, with an AUC of 0.918 in the training group and 0.831 in the validation group.

[0065] Conclusion: The results show that tumor-associated collagen features can independently predict lymph node metastasis in PDAC and may help guide treatment decisions.

[0066] The collagen feature-based lymph node status determination method according to the preferred embodiment of the present application, as shown in Figure 1 The collagen feature-based lymph node status determination method includes the following steps:

[0067] In step S10, a pancreatic ductal adenocarcinoma data sample is obtained, and multi-photon image acquisition is performed on the pancreatic ductal adenocarcinoma data sample to obtain a target multi-photon image.

[0068] For the collection (case collection) of pancreatic ductal adenocarcinoma samples (i.e., the pancreatic ductal adenocarcinoma data sample in the present application): The present application collected 246 formalin-fixed paraffin-embedded (FFPE) PDAC tissue samples from patients aged 28 to 75 years. After 96 samples were excluded according to the exclusion criteria, 150 samples passed the quality control and were included in the final analysis. These samples were then randomly divided into two groups, of which 92 were the training cohort and 58 were the validation cohort.

[0069] The inclusion criteria are: patients with histologically diagnosed PDAC; patients who have not received radiotherapy or chemotherapy in the past, and patients who have received pancreaticoduodenectomy and have pathological evidence of lymph node status. The exclusion criteria include: patients with missing relevant clinicopathological features, patients with combined malignant tumors, and specimens damaged or without tumor tissue. The detailed patient selection process is shown in Figure 2 The obtained clinical characteristics include gender, age at surgery, tumor size, differentiation grade, lymphatic vessel invasion, perineural invasion, and tumor location.

[0070] Specifically, a pancreatic ductal adenocarcinoma data sample is obtained, and the pancreatic ductal adenocarcinoma data sample is divided into a training set and a validation set; the training set is subjected to second-harmonic signal detection processing through the first synchronization channel to obtain a first scanning image; the training set is subjected to two-photon excitation fluorescence signal detection processing through the second synchronization channel to obtain a second scanning image; the first scanning image and the second scanning image are subjected to image splicing processing to obtain a target multi-photon image; and the target multi-photon image is subjected to staining processing to obtain a stained section image.

[0071] For the preparation of pancreatic ductal adenocarcinoma data samples and the process of multi-photon image acquisition: In the present application, a large number of samples and related clinical data were collected using formalin-fixed paraffin-embedded (FFPE) tissues. From the FFPE tissue samples, two consecutive 5 pm thick sections were cut: one for MPM imaging and the other for H&E staining. The present application used an upright microscope (LSM 880, Zeiss, Germany) coupled with a mode-locked femtosecond titanium sapphire laser (Chameleon Ultra, Coherent) to capture high-resolution images. Optical imaging used 810 nm linearly polarized light and a Plan-Apochromat x 20 objective lens (NA = 0.8, Zeiss, Germany) for image acquisition. Two synchronized channels were used to capture backscattered signals: one channel for detecting the second harmonic (SHG) signal (green channel) in the wavelength range of 395-415 nm; the other channel for detecting the two-photon excited fluorescence (TPEF) signal (red channel) in the wavelength range of 428-695 nm. In order to create large size images, the sample was moved using a fine focus stage, and a series of xy scan images were spliced together (representing a two-dimensional plane image obtained by scanning multiple two-dimensional plane images, and then splicing them into a large image). Each xy scan image pixel is 512 x 512, and the data depth is 12 bits.

[0072] Step S20, obtaining macroscopic collagen features in the target multi-photon image, and calculating a first score of the macroscopic collagen features.

[0073] Specifically, a stained section image is obtained, and a region of interest in the stained section image is determined; the region of interest in the stained section image is labeled to obtain a labeled region; image acquisition is performed on the target multi-photon image according to the labeled region to obtain a second harmonic signal image and a two-photon excited fluorescence signal image; macroscopic collagen features in the second harmonic signal image and the two-photon excited fluorescence signal image are obtained, and a cross-validation ridge regression analysis method is used to calculate scores of the macroscopic collagen features to obtain a first score.

[0074] There are multiple non-overlapping regions of interest (ROIs) in the entire H&E image (i.e., the stained section image in the present application), and two pathologists who are unaware of the patient's pathological results label and number the invasive edges and adjacent tumor regions (such as Figure 3 as shown in H&E in FIG. 1, Figure 3The images in the left middle panel are, from top to bottom: H&E image of the whole slide (with ROIs highlighted), MPM image of the specific ROIs (including TPEF and SHG), corresponding SHG image, and binary image converted from the SHG image. Ridge and Lasso regression were then used to quantify collagen features, and then models were developed and validated that combined these collagen scores. SHG (i.e., second harmonic signal image in the present invention) and TPEF images (i.e., two-photon excited fluorescence signal image in the present invention) of all annotated ROIs were simultaneously acquired on another unstained slide using label-free MPM.

[0075] In the study prior to the present invention, a comprehensive protocol was provided to quantify macroscopic tumor-associated collagen signatures (ma-TACS). TACS1-8 represent macroscopic patterns of collagen fiber distribution at the tumor center and tumor-stroma interface. Detailed description of TACS1-8 is shown in Table 1. Using the quantified TACS and LN status in the training cohort, ridge regression was used to determine the coefficient of each TACS with cross-validation, and all the coefficients of TACS were incorporated into a formula to calculate the ma-TACS score for each patient in both cohorts.

[0076] Table 1: Detailed description of TACS1-8

[0077]

[0078]

[0079] The scoring with ridge regression is that a program is written in R language to put the data in and run it to get the coefficients of the 8 ma-TACS, and this coefficient is used as a weight to calculate a total ma-TACS score. The formula for calculating the score of the 8 ma-TACS coefficients (i.e., ma-TACS score) is as follows: ma-TACS score = -1.4663464 + (0.9970001*TACS1) + (10.1951931*TACS2) + (-1.9484164*TACS3) + (0.4167665*TACS4) + (1.5025524*TACS5) + (2.2413038*TACS6) + (0.9657341*TACS7) + (0.2288334*TACS8), wherein -1.4663464 is a constant term calculated.

[0080] Step S30, obtaining a micro-collagen feature in the target multi-photon image, and calculating a second score of the micro-collagen feature.

[0081] Specifically, a region of interest of the target multi-photon image is intercepted to obtain a target region of interest; microscopic collagen features in the target region of interest are extracted, wherein the microscopic collagen features include morphological features, histogram-based features, gray level co-occurrence matrix features and Gabor wavelet transform features; and LASSO regression is used to calculate scores of the microscopic collagen features to obtain second scores.

[0082] Using the MPM image (i.e., the target multi-photon image in the present application), the present application identifies 8 main TACS, similar to the way of identifying histopathological subtypes through H&E images. For TACS1-8, they are mainly based on the macroscopic morphological changes of collagen in the tumor microenvironment. As shown in Figure 3 The ma-TACS score related to lymph node metastasis represents a combination of the 8 identified TACS, and the present application also provides a formula for the ma-TACS score. The present application segments the SHG image and extracts 142 mi-TACS features, and performs LASSO logistic regression analysis on the training cohort (LASSO regression uses R language code to select the most relevant features to lymph node metastasis from the 142-dimensional features, and obtains weight values, and then linearly adds the 6 feature values to obtain the mi-TACS score), to identify 6 features related to lymph node metastasis (as shown in Figure 4 Figure 4 To use LASSOCox regression analysis for Mi-TACS selection. Figure 4 Figure A in the middle shows the relationship between the binomial bias and the logarithm (lambda), and the vertical dotted line is drawn at the optimal value using the minimum criterion and one standard error (1-SE criterion). Figure 4 Figure B in the middle shows the relationship between the LASSO coefficient and the logarithm (lambda), and the vertical dotted line represents the optimal lambda result with 6 non-zero coefficients), the calculation formula of the mi-TACS score (using LASSO logistic regression analysis, the present application obtains the mi-TACS score for each patient according to the selected six microscopic features) is as follows: mi-TACS score = -0.06864118 + (-0.09851105 * Length) + (0.64171781 * Straightness) + (0.54252351 * Orientation) + (0.78679901 * Kurtosis of histograms) + (-0.46889455 * GLCM correlation_135°_5pixel) + (-0.06386891 * Gabor_variance_120°_2scale).

[0083] ​Performance of prediction of ma-TACS score and mi-TACS score: Univariate and multivariate logistic regression analysis showed significant correlation between ma-TACS score, mi-TACS score and LN metastasis (see Table 2). Histograms of ma-TACS and mi-TACS scores in LN0 and LN1 are shown in Figure 5 Figure 5 For each patient's Ma-TACS and mi-TACS score, LN0 or LN1 patients were marked with different colors, where LN0 represents no lymph node metastasis and LN1 represents lymph node metastasis, and high scores are consistently associated with LN1 and low scores are associated with LN0. High scores are mostly LN1 and low scores are LN0. The scores of LN1 (orange box) are significantly different from the scores of LN0 (green box). There are statistically significant differences in ma-TACS and mi-TACS scores (i.e. interquartile range) between LN1 group and LN0 group. (ma-TACS score: -0.406 [-0.770 to 0.060] vs. -0.079 [-0.545 to 0.570], P < 0.05; mi-TACS score: -1.341 [-1.802 to -0.462] vs. 0.952 [0.283 to 1.314], P < 0.05) as shown in Figure 6 Figure 6 A is the Ma-TACS score distribution box plot of LN0 and LN1 in B; Figure 6 B is the Mi-TACS score distribution box plot of LN0 and LN1 in C).

[0084] Table 2: Univariate and multivariate logistic regression analysis variables and lymph node metastasis

[0085]

[0086]

[0087] Step S40, constructing nomogram model according to the first score and the second score.

[0088] Specifically, a Youden index threshold is determined, and the patient types in the training set are divided into lymph node metastasis negative and lymph node metastasis positive according to the first score, the second score and the Youden index threshold; univariate and multivariate logistic regression analysis method is used to calculate the first correspondence between the first score and the second score and the lymph node metastasis negative, and the second correspondence between the first score and the second score and the lymph node metastasis positive, and a nomogram model is constructed according to the first correspondence and the second correspondence.

[0089] ​​The present application extracts the region of interest (ROI) from each large-scale MPM image (extracted along the tumor and its marginal area), each region has a field of view of 512x512 pixels, to extract mi-TACS. Four types of mi-TACS are extracted using matlab 2022a, including 8 morphological features, 6 histogram-based features, 80 gray level co-occurrence matrix (GLCM) features and 48 Gabor wavelet transform (GWT) features. Using a Gaussian mixture model segmentation algorithm, the collagen fibers in the SHG image are initially segmented from the background (the SHG image is binarized using the Gaussian mixture model algorithm, and the collagen image is changed to a binary image using matlab software, such as Figure 3 Then, the present application applies a fiber network extraction algorithm to the binary mask image of collagen fibers to track each fiber and identify cross-linking points (i.e., connecting points connecting multiple fibers), where the collagen area in the image is calculated by dividing the collagen pixel by the total pixel. The fiber density, length, width, straightness, cross-linking density and cross-linking space are calculated using the identified cross-linking points. In addition, the present application also uses Fourier transform analysis to quantify the arrangement orientation index of collagen. The SHG pixel intensity histogram feature describes the intensity distribution, such as mean, variance, skewness, kurtosis, energy and entropy. The gray level co-occurrence matrix (GLCM) features include four types of statistical features: contrast, correlation, energy and homogeneity. For GWT features, Gabor filters are used to determine the mean and variance of the convolution amplitude on the SHG image. The SHG image is convolved using Gabor filters at four scales and six different directions of 0 degrees, 30 degrees, 60 degrees, 90 degrees, 120 degrees and 150 degrees, and the mean and variance of the image convolution amplitude are calculated for each setting, but 142 mi-TACS are not all related to lymph node metastasis, and some irrelevant features may reduce the prediction performance of the model. Feature selection of high-dimensional data helps to eliminate redundant features and retain the most relevant features, thereby constructing a more effective prediction model. The least absolute shrinkage and selection operator (LASSO) is a widely used variable selection method that can retain valuable variables while preventing overfitting. In order to construct a reliable mi-TACS score, the present application applies LASSO regression to identify the most relevant and most robust features from the 142 mi-TACS. By linearly combining the selected features, a new collagen marker is formed, and each feature is weighted according to its coefficient.

[0090] Further, the ROC analysis method is used to determine the performance of the nomogram model, and the performance of the nomogram model is verified according to the verification set, and the performance result of the model is obtained.

[0091] The present application uses single factor and multi-factor logistic regression analysis to evaluate the conventional clinical risk factors, ma-TACS score and mi-TACS score, aiming to explore the relationship with lymph node metastasis (LN0, negative; LN1, positive). The receiver operating characteristic (ROC) curve is used to evaluate the discriminant performance of the developed model (one column is the independent variable, and one column is the dependent variable). In addition, the performance of the model is verified in the verification queue, the nomogram is constructed using the training queue, and the verification is carried out in the verification queue, which provides a direct, quantitative tool for clinicians to quickly predict patient prognosis. At the same time, the calibration chart is used to evaluate the accuracy of the nomogram model, which graphically depicts the correlation between observed probability and predicted probability. The alignment of the calibration curve with the diagonal line in the figure indicates the prediction accuracy of the model, and the threshold value determined according to the maximum Youden index in the training queue divides the patients into LN0 group and LN1 group. Then, the best sensitivity, specificity and cutoff value obtained from the index are applied to the verification queue. Among them, χ 2 The test compares the clinical classification variables (χ 2 The test, also known as the chi-square test, is a non-parametric statistical method mainly used to analyze the correlation or difference between classification variables (i.e. qualitative data). The core idea is to compare the deviation between the observed frequency (actual data) and the expected frequency (theoretical hypothesis) to determine whether there is a statistical correlation between variables. In the specific implementation process, it is to compare the statistical classification variables and the lymph node metastasis statistical P value to determine whether the statistical result is significant), while the mann-Whitney U test analyzes the difference of collagen characteristics; the decision curve analysis (DCA) is used to evaluate the clinical value, by calculating the net benefit of different threshold probabilities in clinical, ma-TACS, mi-TACS and nomogram model, and using R 3.5.2 and IBM SPSS Statistics 25 for statistical analysis.

[0092] The performance of the model is mainly evaluated by ROC analysis on the training queue and the verification queue, and the AUC of the ma-TACS score in predicting lymph node metastasis reaches 0.687 (95% CI, 0.587-0.779) in the training queue and 0.636 (95% CI, 0.499-0.758) in the verification queue (as shown in Figure 7 C and D in the table, Figure 7ROC curves for clinical models (A and B) and ma-TACS scores (C and D) in the training and validation cohorts). The mi-TACS score showed strong predictive ability with an AUC of 0.884 (95% CI, 0.800-0.941) in the training cohort and 0.831 (95% CI, 0.709-0.916) in the validation cohort (as shown in Figs. 1C and 1D, respectively). Figure 8 E and F in Fig. 1, Figure 8 ROC curves for mi-TACS score models (E and F) and ma / mi-TACS scores (G and H) in the training and validation cohorts). The ma-TACS and mi-TACS models outperformed the clinical model (0.642; 95% CI, 0.535-0.739; 0.572; 95% CI, 0.435-0.701) (as shown in Figs. 1E and 1F, respectively). Figure 7 The present disclosure integrates all clinical parameters to build a clinical model, which shows that the ma-TACS or mi-TACS score has good predictive performance.

[0093] In contrast, the mi-TACS score showed the best predictive effect, and the six collagen features selected by LASSO regression to build the mi-TACS score were highly correlated with lymph node metastasis (as shown in Fig. 2A). Figure 9 Figure 9 A in Fig. 2 is the correlation analysis between the six collagen features selected by LASSO regression and the training, Figure 9 in the validation cohort), and from the morphological features, the orientation angle and straightness of the collagen fibers were positively correlated with lymph node metastasis, while the length of the collagen fibers was negatively correlated with lymph node metastasis.

[0094] Performance comparison of different prediction models: To further evaluate the predictive performance of various models, five models for predicting PDAC lymph node metastasis were established (including clinical, ma-TACS, mi-TACS, ma / mi-TACS, and clinical+ma / mi-TACS). The prediction model established by clinical pathological factors (i.e., the clinical model) performed the worst among the five models. The clinical model showed a significant difference between the LN0 group and the LN1 group in the training cohort (P<0.05), but no significant difference in the validation cohort (P=0.36). The ma-TACS model was close to a significant difference between the LN0 group and the LN1 group in the validation (P=0.08, P<0.1). In both the training and validation cohorts, the mi-TACS model showed a significant difference between the two groups (P<0.0001). The TACS combination (ma / mi-TACS) model and the complete model (clinical+ma / mi-TACS) also showed significant differences in both cohorts (as shown in Figs. 3A and 3B, respectively). Figure 10 Figure 10 ​​Violin plots showing the difference in scores between LN0 and LN1 groups for the five models in the training and validation cohorts.

[0095] When the ma-TACS and mi-TACS models were combined, the AUC in the training cohort increased to 0.918 (95% CI, 0.842-0.965), however, the AUC in the validation cohort did not increase compared to the mi-TACS model alone (as shown in Figure 8 G and H in FIG. 6). This can be due to the mediocre performance of the ma-TACS model, and the combination of the three single models actually decreased the predictive performance on the validation set. In fact, this is related to the poor predictive performance of the clinical model (as shown in Figure 11 I and J in FIG. 6, Figure 11 ROC curves for the full models (clinical + ma / mi-TACS scores) in the training and validation cohorts (I, J). The corresponding sensitivity and specificity for the five models are shown in Table 3. Among the single models, the mi-TACS performed the best with the highest sensitivity and specificity. The specificity of the combined models (e.g., full models) improved, but the sensitivity decreased, especially in the validation cohort.

[0096] Table 3: Performance comparison of different predictive models for lymph node metastasis

[0097]

[0098] Patients were divided into LN0 and LN1 groups using the threshold that maximized the Youden index. This threshold was used to divide patients in the training and validation cohorts into LN0 and LN1 groups. Figure 3 The classification performance of the five models based on their respective thresholds is visually demonstrated in FIG. 6. As shown in Figure 12 the left of the dashed line are predicted to be lymph node negative (LN0) and the right are predicted to be lymph node positive (LN1). Green dots represent actual lymph node negative (LN0) patients and red dots represent actual lymph node positive (LN1) patients. The markers inside the black dashed boxes highlight patients with misclassified lymph nodes. As shown in Figure 12 Figure 12 the training and validation cohorts), the specificity of the clinical model is poor, i.e., it is more likely to misdiagnose patients without lymph node metastasis as having metastasis (as shown by the large number of green dots in the right red box of Figure 12 the ma-TACS model is poor, resulting in more true lymph node metastasis patients being misdiagnosed as non-metastatic (as shown by the large number of red dots in the left red box of Figure 12 ​The mi-TACS model outperformed the previous two single models, particularly in terms of sensitivity, with only a small number of patients with true lymph node metastases being misdiagnosed as having no metastases. Both the combined ma / mi-TACS model and the complete model demonstrated further improvements.

[0099] Regarding the construction process of the combined nomogram: The present invention develops a clinically applicable nomogram integrating ma-TACS score and mi-TACS score to predict the risk of lymph node metastasis (e.g. Figure 13 A in the figure is a nomogram that integrates ma-TACS and mi-TACS scores for prediction). Each ma-TACS or mi-TACS score is assigned a corresponding score on the scoring scale. For example, if a patient has a mi-TACS score of 1 and a ma-TACS score of 1.5, the total score is 147 points, indicating a 98.2% probability of lymph node metastasis. In addition, the calibration plots of the training cohort and the validation cohort showed good consistency (C index: 0.918 and 0.831, respectively). The calibration curve of the nomogram is shown in Figure 1. Figure 13 B and C in the figure are the calibration curves of the nomogram for the training and validation cohorts. In order to further clarify the predictive performance of the ma-TACS+mi-TACS score in different subgroups, the present invention conducted multiple subgroup analyses based on clinical factors. It was found that the score maintained a high predictive performance in all subgroups (see Table 4). The clinical decision-making utility of the prediction model was demonstrated using decision curve analysis (DCA). Figure 14 A and B in ( Figure 14 The DCA for each model is shown on the y-axis. The green, orange, red, and black lines represent the ma / mi-TACS, mi-TACS, ma-TACS, and clinical models, respectively. The clinical utility of the corresponding models in both cohorts is shown based on the area under the decision curve (AUC), which indicates an "all-treatment" strategy (gray), and a "no-treatment" strategy (blue). In both the training and validation cohorts, the nomogram model (ma / mi-TACS (green) model) had a larger area than other models (e.g., ma-TACS score, red; mi-TACS score, orange; and clinical, black), outperforming both the "all-treatment" (gray) and "no-treatment" (blue) strategies. Compared with individual models (clinical, ma-TACS, mi-TACS), the nomogram model demonstrated the highest net benefit, demonstrating its utility as a reliable clinical tool for predicting lymph node metastasis in patients undergoing surgical resection of PDAC.

[0100] Table 4: Prediction of ma-TACS+mi-TACS scores for patients classified by clinical pathology

[0101]

[0102]

[0103] Step S50, obtaining current pancreatic ductal adenocarcinoma data of the patient, inputting the current pancreatic ductal adenocarcinoma data into the nomogram model, and obtaining a lymph node metastasis result.

[0104] In summary, according to the selection criteria, a total of 150 patients were included, and divided into a training group of 92 cases and a validation group of 58 cases. The results showed that there was no significant difference in PDCA LN status or clinical characteristics (such as gender, age, tumor size, differentiation grade, lymphatic vessel invasion, perineural invasion, and tumor location) between the training group and the validation group (P>0.05). Table 5 summarizes the clinical characteristics and collagen score distribution of the LN0 and LN1 groups. In the LN0 and LN1 groups of the training group, there was a significant difference in ma-TACS score and mi-TACS score (P<0.05). In the validation, the ma-TACS score between the LN0 group and the LN1 group was close to significant difference (P=0.08).

[0105] Table 5: Characteristics of PDAC patients in LN metastasis negative (LN0) and LN metastasis positive (LN1) groups

[0106]

[0107]

[0108] The research of the present application shows that there is a correlation between macroscopic and microscopic tumor-associated collagen protein markers ma / mi-TACS and lymph node metastasis of pancreatic ductal adenocarcinoma (PDAC) patients. ROC analysis shows that the nomogram model has excellent prediction ability in both the training and validation cohorts. The AUC of the ma / mi-TACS model (0.918, 0.831) and the single mi-TACS model (0.884, 0.831) is significantly higher than that of the clinical model (0.642, 0.572). Single factor and multiple factor Logistic regression analysis shows that both ma-TACS and mi-TACS scores are independent predictors of lymph node metastasis. In addition, in the Logistic regression analysis, no clinicopathological factor is independently related to lymph node metastasis, which highlights the challenge of identifying high-risk PDAC patients by evaluating only clinical data. The mi-TACS model itself shows strong prediction performance. During the LASSO optimization process selected for mi-TACS, the selected six features have a strong correlation with lymph node metastasis (such as Figure 9Microscopic collagen features of pancreatic tumors reflect tumor progression and invasion better. Among them, the first-order histogram kurtosis, collagen fiber straightness, and collagen fiber orientation are positively correlated with lymph node metastasis, while the collagen fiber length is negatively correlated. Geovanni et al. showed that collagen fibers in tumor tissues are more aligned and shorter in length than in normal tissues. Previous studies have shown that the direction and straightness of collagen fibers are prognostic indicators for patients with invasive breast cancer. In addition, pancreatic tumor studies have found that highly aligned stromal collagen is a poor prognostic factor after PDAC resection. Histogram, GLCM, and GWT features are texture features of collagen fibers found in multiple studies, showing potential for clinical application in disease diagnosis. Quantitative analysis of collagen fibers can reveal microscopic features related to tumor metastasis that cannot be observed in standard medical imaging. The present invention uses ma-TACS and mi-TACS scores to develop and validate a nomogram for predicting lymph node metastasis in PDAC patients, which has been validated in an independent cohort and shown to have good reproducibility and reliability.

[0109] Lymph node metastasis is a poor prognostic factor for pancreatic cancer (PDAC), and strong evidence suggests that lymph node status is crucial for determining treatment options and predicting prognosis for pancreatic cancer patients. To improve long-term survival, pancreatic resection is the most effective treatment, but there is still controversy about whether it should include standard lymph node dissection or extended lymph node dissection. However, more extensive pancreatic resection can lead to a higher incidence of complications. For high-risk patients, lymph node dissection should be carefully considered during surgery. Therefore, preoperative or intraoperative prediction of lymph node metastasis in PDAC patients can help avoid overtreatment and provide information for surgical decision-making. CT, MRI, and PET are widely used for screening and diagnosis of pancreatic cancer. Although these imaging diagnoses are not entirely satisfactory, surgical decisions are largely dependent on them. While macroscopic features can provide prognostic information for PDAC patients, they are not sufficient to assess lymph node status. It is becoming increasingly common to conduct radiomics research using these images. For example, existing technologies integrate MRI and diffusion-weighted imaging (DWI) to predict lymph node metastasis in pancreatic neuroendocrine tumors. A similar MRI-based radiomics model has also been developed and validated in the prior art to predict lymph node metastasis in PDAC, which combines portal venous phase (PVP) enhanced T1-weighted imaging (T1WI) and T2-weighted imaging (T2WI). An AI model using CT imaging has also been developed in the prior art to automatically predict lymph node metastasis in PDAC patients. These findings can guide treatment decisions to some extent. However, due to objective factors such as limited resolution, these imaging methods cannot effectively assess histopathological features at the cellular and molecular levels like H&E staining, which affects the prediction accuracy and stability of the model.

[0110] According to Paget's seed and soil hypothesis, the tumor microenvironment, or the soil surrounding the tumor seed, is critical for cancer development. Collagen fibers are an important component of the tumor microenvironment, and changes in collagen structure and morphology play an important role in solid tumors such as breast and pancreatic cancer, which can increase the stiffness, invasiveness and metastasis of tumors. The mechanical properties of the collagen network are influenced by factors such as fiber length, straightness, arrangement, density and spatial orientation, which can affect the movement of the stroma and tumor cells, and the arrangement of collagen fibers is a key factor in regulating tumor cell contact guidance. Different patterns of collagen fiber arrangement in the tumor stroma have been identified and are believed to contribute to tumor cell migration and invasion.

[0111] In the research of the present application, macro- and micro-tumor-associated collagen protein markers (ma / mi-TACS) were extracted from pancreatic ductal tumor tissues. The research results show that these two collagen protein markers are closely related to lymph node metastasis of pancreatic ductal adenocarcinoma (PDAC), and the correlation of mi-TACS is particularly significant. This may indicate that changes in the microstructure of collagen fibers before the formation of macroscopic features can significantly affect tumor progression. Early stages of the disease can lead to microscopic structural changes that are not visible to the naked eye. In the future, this method can be used for differential diagnosis of early pancreatic tumors and improve the accuracy of early detection. In summary, the observations show that the ma-TACS and mi-TACS scores can more accurately predict lymph node metastasis compared to traditional clinical models. In addition, these scores provide valuable insights for PDAC analysis, thereby improving collagen fiber-based lymph node metastasis prediction. Despite these exciting findings, the present application still has some limitations. First, a collagen-based prediction model was constructed based on a single-center retrospective trial, but the small sample size limits its application, and additional external validation using a larger dataset is needed to confirm the robustness and accuracy of the model. Second, the present application only focuses on whether PDAC patients have lymph node metastasis. However, according to the latest cancer staging guidelines, the number of metastatic lymph nodes is crucial.

[0112] In summary, the preliminary results of the present application show that the ma-TACS and mi-TACS scores in MPM images can accurately distinguish the lymph node status of PDAC patients, and the nomogram model developed and validated in the present application may have value for the prognosis and treatment management of PDAC patients.

[0113] Further, as shown in Figure 15 based on the above collagen feature-based lymph node status determination method, the present application also correspondingly provides a collagen feature-based lymph node status determination system, wherein the collagen feature-based lymph node status determination system comprises:

[0114] a multi-photon image acquisition module 51, configured to acquire a pancreatic ductal gland data sample, and perform multi-photon image acquisition on the pancreatic ductal gland data sample to obtain a target multi-photon image;

[0115] a macroscopic feature scoring module 52, configured to acquire a macroscopic collagen feature in the target multi-photon image, and calculate a first score of the macroscopic collagen feature;

[0116] a microscopic feature scoring module 53, configured to acquire a microscopic collagen feature in the target multi-photon image, and calculate a second score of the microscopic collagen feature;

[0117] a nomogram model construction module 54, configured to construct a nomogram model according to the first score and the second score;

[0118] a lymph node metastasis result output module 55, configured to acquire current pancreatic ductal gland data of a patient, input the current pancreatic ductal gland data into the nomogram model, and obtain a lymph node metastasis result.

[0119] Further, as shown in the following, Figure 16 Based on the lymph node state judgment method and system based on collagen features, the application further provides a terminal, which comprises a processor 10, a memory 20 and a display 30. Figure 16 Only some components of the terminal are shown, but it should be understood that all the shown components are not required, and more or less components can be alternatively implemented.

[0120] The memory 20 can be an internal storage unit of the terminal in some embodiments, such as a hard disk or a memory of the terminal. The memory 20 can also be an external storage device of the terminal in other embodiments, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the memory 20 can include both the internal storage unit and the external storage device of the terminal. The memory 20 is used to store application software and various data installed on the terminal, such as program codes of the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In an embodiment, the memory 20 stores a lymph node state judgment program 40 based on collagen features, which can be executed by the processor 10, so as to implement the lymph node state judgment method based on collagen features in the application.

[0121] The processor 10 can be a Central Processing Unit (CPU), a microprocessor or other data processing chip in some embodiments, for running program codes stored in the memory 20 or processing data, such as executing the collagen feature-based lymph node state judgment method, etc.

[0122] The display 30 can be an LED display, a liquid crystal display, a touch liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, etc. in some embodiments. The display 30 is used to display information of the terminal and to display a visualized user interface.

[0123] In an embodiment, the steps of the collagen feature-based lymph node state judgment method are implemented when the processor 10 executes the collagen feature-based lymph node state judgment program 40 in the memory 20.

[0124] In summary, the present application provides a collagen feature-based lymph node state judgment method, system and terminal, the method comprising: acquiring a pancreatic ductal gland data sample, and performing multi-photon image acquisition on the pancreatic ductal gland data sample to obtain a target multi-photon image; acquiring a macroscopic collagen feature in the target multi-photon image and calculating a first score of the macroscopic collagen feature; acquiring a microscopic collagen feature in the target multi-photon image and calculating a second score of the microscopic collagen feature; constructing a nomogram model according to the first score and the second score; acquiring current pancreatic ductal gland data of a patient, and inputting the current pancreatic ductal gland data into the nomogram model to obtain a lymph node metastasis result. The present application can more comprehensively understand the relationship between collagen protein features and lymph node metastasis by extracting macroscopic collagen features and microscopic collagen features in the pancreatic ductal gland data sample, calculating the respective scores of the macroscopic collagen features and the microscopic collagen features, and then constructing a nomogram model according to the scores, thereby effectively improving the accuracy of the lymph node metastasis result output.

[0125] It should be noted that in this document, the terms "comprising", "including", or any other variant thereof are intended to cover non-exclusive inclusions, so that processes, methods, articles or terminals including a series of elements not only include those elements, but also include other elements not explicitly listed, or include elements inherent to such processes, methods, articles or terminals. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article or terminal including the element.

[0126] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing relevant hardware (such as a processor, a controller, etc.) through a computer program. The program can be stored in a computer-readable computer-readable storage medium, and the program can include the processes of the above-mentioned method embodiments when executed. The computer-readable storage medium can be a memory, a disk, an optical disk, etc.

[0127] It should be understood that the application is not limited to the above examples, and those skilled in the art can make improvements or changes according to the above description, and all these improvements and changes shall belong to the protection scope of the appended claims of the application.

Claims

1. A method for determining lymph node status based on collagen characteristics, characterized in that: The method for determining lymph node status based on collagen characteristics includes: Acquiring a pancreatic duct gland data sample, and performing multiphoton image acquisition on the pancreatic duct gland data sample to obtain a target multiphoton image; Acquiring a macroscopic collagen feature in the target multiphoton image and calculating a first score of the macroscopic collagen feature; acquiring microscopic collagen features in the target multiphoton image and calculating a second score of the microscopic collagen features; constructing a nomogram model according to the first score and the second score; The current pancreatic duct gland data of the patient is obtained, and the current pancreatic duct gland data is input into the nomogram model to obtain the lymph node metastasis result.

2. The method for determining lymph node status based on collagen characteristics according to claim 1, characterized in that: The step of acquiring the pancreatic ductal gland data sample and performing multiphoton image acquisition on the pancreatic ductal gland data sample to obtain a target multiphoton image specifically includes: Obtaining pancreatic ductal gland data samples, and dividing the pancreatic ductal gland data samples into a training set and a validation set; A multiphoton microscope is used to collect multiphoton images of the training set to obtain a target multiphoton image.

3. The method for determining lymph node status based on collagen characteristics according to claim 2, wherein: The multiphoton microscope includes a first synchronization channel and a second synchronization channel; The multiphoton image acquisition of the training set using a multiphoton microscope to obtain a target multiphoton image specifically includes: Performing second harmonic signal detection processing on the training set through the first synchronization channel to obtain a first scanned image; Performing two-photon excitation fluorescence signal detection processing on the training set through the second synchronization channel to obtain a second scanned image; The first scanned image and the second scanned image are subjected to image stitching processing to obtain a target multiphoton image.

4. The method for determining lymph node status based on collagen characteristics according to claim 2, wherein: The step of obtaining the macroscopic collagen feature in the target multiphoton image and calculating the first score of the macroscopic collagen feature specifically includes: Acquiring a stained section image and determining a region of interest in the stained section image; performing labeling processing on the region of interest in the stained section image to obtain a labeled region; Performing image acquisition on the target multiphoton image according to the marked area to obtain a second harmonic signal image and a two-photon excitation fluorescence signal image; Macroscopic collagen features in the second harmonic generation signal image and the two-photon excitation fluorescence signal image are acquired, and a cross-validated ridge regression analysis method is used to score and calculate the macroscopic collagen features to obtain a first score.

5. The method for determining lymph node status based on collagen characteristics according to claim 1, wherein: The acquiring of the microscopic collagen feature in the target multiphoton image and calculating the second score of the microscopic collagen feature specifically includes: Performing a region of interest interception process on the target multiphoton image to obtain a target region of interest; Extracting microscopic collagen features in the target region of interest, wherein the microscopic collagen features include morphological features, histogram-based features, gray-level co-occurrence matrix features, and Gabor wavelet transform features; The LASSO regression method is used to score and calculate the microscopic collagen characteristics to obtain a second score.

6. The method for determining lymph node status based on collagen characteristics according to claim 2, characterized in that: The constructing of a nomogram model according to the first score and the second score specifically includes: determining a Youden Index threshold, and dividing the patient types in the training set into lymph node metastasis-negative and lymph node metastasis-positive according to the first score, the second score, and the Youden Index threshold; Univariate and multivariate logistic regression analysis methods were used to calculate the first correspondence between the first score, the second score and the negative lymph node metastasis, as well as the second correspondence between the first score, the second score and the positive lymph node metastasis. A nomogram model was constructed based on the first correspondence and the second correspondence.

7. The method for determining lymph node status based on collagen characteristics according to claim 2, wherein: The step of constructing a nomogram model according to the first score and the second score further includes: The ROC analysis method is used to judge the model performance of the nomogram model, and the model performance of the nomogram model is verified according to the verification set to obtain the model performance result.

8. A lymph node status judgment system based on collagen characteristics, characterized in that: The lymph node status judgment system based on collagen characteristics includes: A multiphoton image acquisition module is used to acquire pancreatic duct gland data samples and perform multiphoton image acquisition on the pancreatic duct gland data samples to obtain a target multiphoton image; a macroscopic feature scoring module, configured to obtain macroscopic collagen features in the target multiphoton image and calculate a first score for the macroscopic collagen features; a microscopic feature scoring module, configured to obtain microscopic collagen features in the target multiphoton image and calculate a second score for the microscopic collagen features; a nomogram model construction module, configured to construct a nomogram model according to the first score and the second score; The lymph node metastasis result output module is used to obtain the patient's current pancreatic duct gland data, input the current pancreatic duct gland data into the nomogram model, and obtain the lymph node metastasis result.

9. A terminal, characterized in that: The terminal includes: a memory, a processor, and a lymph node status judgment program based on collagen features stored in the memory and runnable on the processor. When the lymph node status judgment program based on collagen features is executed by the processor, the steps of the lymph node status judgment method based on collagen features as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a lymph node status judgment program based on collagen characteristics, and when the lymph node status judgment program based on collagen characteristics is executed by a processor, the steps of the lymph node status judgment method based on collagen characteristics as described in any one of claims 1-7 are implemented.

Images