Chronic pancreatitis pathological typing prediction method
Through deep learning technology, the pancreatic tissue is automatically segmented and quantified and latent category analysis is carried out, and combined with CT scan period images, a prediction model of pancreatitis pathological typing is constructed, which solves the problems of poor consistency and insufficient quantification in the existing technology, and realizes the Non-invasive and accurate diagnosis of pathological typing of chronic pancreatitis.
Patent Information
- Application Number
- CN202411028065.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-07-29
AI Technical Summary
The existing pathological evaluation methods for chronic pancreatitis have poor diagnosis consistency, low credibility, and have failed to effectively quantify the pathological severity, so they cannot accurately classify and monitor the progress of the disease.
Deep learning technology is used to automatically segment and quantify the acinars, interstitial, pancreatic islets, nerves, fat and lymphocytes of the pancreas, and pathological typing is determined through latent category analysis. Automatic quantization segmentation method is developed based on pancreatic CT scan period images, and the construction of A prediction model for pathotyping of pancreatitis, achieving non-invasive and iodine-free pathotyping diagnosis.
Improve the simplicity and universality of pathological evaluation, and realize the accuracy and reliability of pancreatitis pathological typing. There is no need to obtain stained HE pathological images and CT enhanced images. It only needs to obtain plain scanning CT images, which is suitable for liver and kidney dysfunction. crowd.
Smart Images

Figure CN118967615A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical image processing, and in particular to a pancreatitis pathology image processing method, a classification method, a typing prediction model construction method, a prediction method, a device, a medium and a program product. Background Art
[0002] Chronic pancreatitis is a chronic, progressive, irreversible fibro-inflammatory disease. Histopathology is the best way to visually demonstrate the severity of the disease and is the gold standard for diagnosing the disease. Due to the scarcity of pathological specimens of chronic pancreatitis, related studies are rare. The clinical pathological characteristics of chronic pancreatitis are progressive atrophy of the pancreatic parenchyma, fibrosis, pain response, pancreatic duct torsion, stenosis or dilatation, stones, and pancreatic endocrine and exocrine dysfunction. This process gradually worsens as the disease progresses. Accurate grading of chronic pancreatitis is the key to understanding its severity, monitoring the condition, and guiding treatment.
[0003] Among the existing pathological assessment methods for chronic pancreatitis, routine pathological reports often use general mild, moderate, and severe grades to evaluate the degree of histological changes in chronic pancreatitis. Doctors have poor diagnostic consistency, and the evaluation object is vague and unreliable. A few pathological studies on the severity of chronic pancreatitis are mostly semi-quantitative assessments of single tissue components. For example, the PF grading method proposed by Kloppel et al. is relatively simple and has a certain correlation with pancreatic function, but it cannot overcome subjective factors and ignores the correlation value between various tissue components. It has not been widely recognized and applied in clinical practice. The severity grading systems such as the M-ANNHEIM classification system are mostly based on clinical and imaging, with many evaluation indicators and cumbersome processes, making them difficult to popularize and apply in clinical practice.
[0004] A pathological severity grading based on accurate quantification of chronic pancreatitis pathology has not yet been proposed. Summary of the invention
[0005] The present invention provides a pancreatitis pathological image processing method, a classification method, a typing prediction model construction method, a prediction model, a device, a medium and a program product, which can be applied to the prognosis classification diagnosis of chronic pancreatitis.
[0006] An embodiment of the present invention discloses a pancreatic pathology image processing method, the method comprising:
[0007] Obtain pancreatic HE pathological images of patients with chronic pancreatitis;
[0008] Segmenting the pancreatic HE pathology image to obtain a mask associated with a plurality of tissues of the pancreas;
[0009] extracting an image block based on the mask;
[0010] The image blocks are used to train an image segmentation model, and parameters of a plurality of predetermined features are obtained.
[0011] Optionally, extracting the image block based on the mask includes performing flipping, rotation, Gaussian blur and color enhancement processing on the image block, and standardizing the image block.
[0012] Optionally, using the image block to train the image segmentation model includes binarizing the image block, deriving segmentation areas of the multiple tissues by traversing pixel points of the image block, and obtaining parameters of the multiple predetermined features based on the segmentation areas.
[0013] An embodiment of the present invention discloses a pancreatic pathology image classification method, the method comprising:
[0014] Obtain pancreatic HE pathological images of patients with chronic pancreatitis;
[0015] Using the method as claimed in claim 1 to obtain parameters of a plurality of predetermined characteristics,
[0016] Based on the parameters of the multiple predetermined features, latent class analysis is used to determine the pathological classification of pancreatitis, and the pathological classification is related to the endocrine function, exocrine function and prognosis of the chronic pancreatitis patient.
[0017] Optionally, the determining the pathological classification of pancreatitis using latent class analysis includes best fitting classification results based on Akaike's Information Criterion (AIC).
[0018] An embodiment of the present invention discloses a method for constructing a pancreatitis pathological typing prediction model, the model construction method comprising:
[0019] Acquiring the first plain scan phase image and portal venous phase image of the pancreatic CT image of a patient with chronic pancreatitis;
[0020] segmenting the portal phase image to obtain a mask associated with a plurality of tissues of the pancreas;
[0021] Using the mask and the first plain scan period image to train an image segmentation model to obtain a trained image segmentation model;
[0022] Inputting the second plain scan image of the patient to be tested into the trained image segmentation model to obtain a pancreatic parenchymal region image of the second plain scan image, and extracting a plurality of radiomics features based on the pancreatic parenchymal region image;
[0023] Predetermined features from the multiple imaging omics features are selected, omics scores are calculated, and a pathology classification prediction model is constructed according to the omics scores.
[0024] Optionally, the using the mask and the first plain scan phase image to train the image segmentation model also includes aligning the first plain scan phase image and the portal venous phase image, generating a mask corresponding to the first plain scan phase image based on the mask, and training the image segmentation model using the mask corresponding to the first plain scan phase image.
[0025] Optionally, extracting the radiomics feature based on the pancreatic parenchymal region image includes performing dimensionality reduction processing on the radiomics feature to obtain the predetermined feature.
[0026] The embodiments of the present invention disclose a method for predicting pancreatitis pathological typing, the method comprising:
[0027] Obtain pancreatic CT images of patients with chronic pancreatitis;
[0028] The prediction results of pancreatitis pathological classification were obtained based on the pathological classification prediction model, among which:
[0029] The pathological classification prediction model is obtained according to the above-mentioned model construction method.
[0030] An embodiment of the present invention discloses an electronic device, characterized in that the device includes a processor and a memory storing computer-executable instructions, and the processor is configured to execute the instructions to implement the above-mentioned pancreatic pathology image processing method, pancreatic pathology image classification method, pancreatitis pathology classification prediction model construction method or pancreatitis pathology classification prediction method.
[0031] An embodiment of the present invention discloses a computer-readable storage medium, characterized in that at least one computer instruction is stored in the computer-readable storage medium, and the at least one instruction is loaded and executed by a processor to implement the above-mentioned pancreatic pathology image processing method, pancreatic pathology image classification method, pancreatitis pathology classification prediction model construction method or pancreatitis pathology classification prediction method.
[0032] An embodiment of the present invention discloses a computer program product, characterized in that the computer program product includes computer instructions, which, when executed, implement the above-mentioned pancreatic pathology image processing method, pancreatic pathology image classification method, pancreatitis pathology classification prediction model construction method or pancreatitis pathology classification prediction method.
[0033] The pancreatitis pathological image processing method, classification method, typing prediction model construction method, prediction method, device, medium and program product provided by the present invention automatically segment and quantify the pancreatic acinus, stroma, islets, nerves, fat and lymphocytes through deep learning. Tissue delineation is performed on the HE pathological image, the segmentation mask of the delineation is collected, the image block is extracted based on the segmentation mask, and the training and verification of the segmentation model are performed at the image block level, and finally the quantification of indicators is achieved, including the automatic quantitative extraction of the acinar proportion, stroma proportion, islet proportion, nerve proportion, fat proportion, lymphatic structure proportion, acinar stroma ratio, islet stroma ratio, acinar islet ratio and functional cell stroma ratio. Furthermore, the similarity between the quantitative indicator groups is evaluated by latent class analysis, and finally a two-class pathological typing is determined. Since the establishment of the above two classifications depends on pathological results, the inherent limitation of chronic pancreatitis pathological research, namely the scarcity of pathological specimens, has not been broken. Therefore, a model for non-invasive prediction of pathological classification was constructed. The image conversion of the above pathological classification was realized by applying deep learning imaging omics. In the process, an automatic quantitative segmentation method for chronic pancreatitis was developed based on pancreatic CT plain scan images, multiple quantitative features were extracted, the most relevant main features were screened, the omics scores were calculated, and a pancreatitis pathological classification prediction model was constructed.
[0034] Compared with the prior art, the main difference and effect of the embodiments of the present invention are that the HE pathological image is used as the input of the algorithm without generating additional costs, which fundamentally improves the simplicity and universality of the segmentation. In addition, based on the pancreatitis pathological classification prediction model of the present invention, when making a pathological classification diagnosis of chronic pancreatitis, it is not necessary to obtain the patient's stained HE pathological image and CT enhanced image, but only the patient's plain scan CT image, which has the advantages of being non-invasive, without the risk of iodine allergy, being cheap and friendly to people with poor liver and kidney function. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 is a flow chart of a pancreatic pathology image processing method according to an embodiment of the present invention;
[0036] Figure 2 is a schematic diagram of a model for establishing pathological typing of pancreatitis according to an embodiment of the present invention;
[0037] Figure 3 is a schematic diagram of two types of pathological typing according to an embodiment of the present invention;
[0038] Figure 4 is a flow chart of a method for constructing a pancreatitis pathological typing prediction model according to an embodiment of the present invention;
[0039] Figure 5 is a schematic diagram of a pathological classification prediction model according to an embodiment of the present invention;
[0040] Figure 6 is a hardware structure block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0041] The present invention will be further described below in conjunction with specific embodiments and accompanying drawings. It will be appreciated that the specific embodiments described herein are intended only to explain the present invention, rather than to limit the present invention. In addition, for ease of description, only some, but not all, structures or processes related to the present invention are shown in the accompanying drawings. It should be noted that in this specification, similar reference numerals and letters represent similar items in the following drawings.
[0042] It should be understood that although the terms "first", "second", etc. may be used herein to describe various features, these features should not be limited by these terms. These terms are used only to distinguish and should not be understood as indicating or implying relative importance. For example, a first feature may be referred to as a second feature, and similarly a second feature may be referred to as a first feature without departing from the scope of the exemplary embodiments.
[0043] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.
[0044] The clinical pathological course of chronic pancreatitis gradually worsens as the disease progresses, and may eventually deteriorate into pancreatic cancer, a malignant tumor with a low survival rate and high lethality. Accurate grading of chronic pancreatitis is the key to understanding its severity, monitoring the condition, and guiding treatment. However, a quantitative pathological classification of chronic pancreatitis has not yet been proposed, and histopathology-related severity assessment methods are very lacking.
[0045] In order to solve the above problems, an embodiment of the present invention provides a pancreatic pathology image processing method and classification method, which automatically segment and quantify the pancreatic alveoli, stroma, islets, nerves, fat and lymphocytes through deep learning. Tissue delineation is performed on the HE pathology image, the segmentation mask of the delineation is collected, the image block is extracted based on the segmentation mask, and the segmentation model is trained and verified at the image block level, and finally the quantification of indicators is achieved, including the automatic quantitative extraction of the acinar proportion, stroma proportion, islet proportion, nerve proportion, fat proportion, lymphatic structure proportion, acinar stroma ratio, islet stroma ratio, acinar islet ratio and functional cell stroma ratio. Furthermore, the similarity between the quantitative indicator groups is evaluated by latent class analysis, and finally a two-class pathological classification is determined. The evaluation efficiency and stability are improved by automatic quantification, and a new pathological classification is established based on the combined similarity between quantitative indicators. The scheme of the present invention is an optimized chronic pancreatitis pathological evaluation method.
[0046] Figure 1The figure is a flow chart of a pancreatic pathology image processing method according to an embodiment of the present invention.
[0047] S110: Obtaining a pancreatic HE pathological image of a patient with chronic pancreatitis.
[0048] Pathological HE staining images are digitized through a slice scanner. In the diagnosis of pancreatitis, pathological diagnosis is considered to be the "gold standard" for tumor diagnosis.
[0049] The embodiment of the present invention obtains a pancreatic HE pathological image by obtaining a pancreatic pathological section of a patient with chronic pancreatitis. The following steps are explained and illustrated based on the obtained pancreatic HE pathological image.
[0050] S120: Segment the pancreatic HE pathological image to obtain a mask associated with multiple tissues of the pancreas.
[0051] To avoid unnecessary calculations on non-target areas of pancreatic HE images, pancreatic HE pathological images were preprocessed. The pancreatic region of interest (ROI) was determined by manual delineation, and the stroma, acinar, islet, nerve, fat, and lymphatic structures were delineated, and the corresponding tissue masks were generated.
[0052] S130: Extracting image blocks based on the mask.
[0053] As an implementation method, multiple tissues of the pancreas are outlined to obtain a mask of the outlined area. Combined with the tissue labels given by the pathologist, for example, stroma, acinar, islet, nerve, fat vacuoles and lymphocytes, image blocks are randomly extracted on the mask, and the obtained image blocks are used as the training set and validation set of the image classification model. The proportion of the training set and the validation set is set to a predetermined ratio according to the specific experimental requirements. The amount of data of the target image block and the non-target image block is kept the same to prevent the problem of class imbalance. When at least one pixel in the image block is classified as the target class, the image block is considered to be the target image.
[0054] In order to better adapt to various staining and acquisition methods, data enhancement can be used. Mask-based image block extraction includes flipping, rotating, Gaussian blurring and color enhancement of image blocks, and standardizing image blocks, such as "horizontal or vertical flipping", "90-degree rotation", "Gaussian blurring" and "color enhancement". Among them, color enhancement includes random changes in brightness, contrast, hue and saturation. After data enhancement processing, the image is almost standardized.
[0055] S140: Using the image blocks to train an image segmentation model, and obtaining parameters of a plurality of predetermined features.
[0056] The hybrid loss function for building the image segmentation model consists of a cross entropy loss and an intersection-over-union loss function. During training, the DeeplabV3+ model was used. The construction of DeepLabV3+ uses dilated convolutions at different rates in order to obtain multi-scale context. DeeplabV3+ extends this by transferring low-level features from the encoder to the decoder. For the first 50 iterations, only the decoder weights are trained. For the remaining iterations, both the encoder and decoder parts are trained. The learning rate decays in a deterministic manner every few iterations to allow the model to gradually converge.
[0057] refer to Figure 2 The HE image is used to train the image segmentation model 210. The HE image is divided into a training set and a validation set. Both the training set and the validation set use image blocks to train the image segmentation model 210, including a binary image block, traversing the pixel points of the image block to obtain the segmentation area of multiple tissues, and obtaining parameters of multiple predetermined features based on the segmentation area.
[0058] In one example, the image segmentation model 210 is trained based on the image blocks as the training set, and the image blocks as the verification set are input into the trained image segmentation model 210 to obtain the segmentation results. Specifically, the image blocks corresponding to multiple tissues are binarized, and the image blocks are binarized into black and white areas of 0 and 255. Each pixel point in the white area is traversed, and the total number of pixels is calculated to obtain the segmentation area of the tissue component. The white area is the segmentation target with a tissue label, and the black area is the remaining components. For example, by accurately segmenting multiple tissues of chronic pancreatitis, the quantized areas of the stroma, acinar, islet, nerve, fat, and lymphatic structures under the pancreatic HE pathological section are obtained. Taking into account the differences in the size of different sections and the content of tissues, in order to enhance comparability, the six types of target components are compared with the pancreatic tissue area of the pancreatic HE pathological image to obtain the respective proportion values of the six types of components, including the proportion of acinar, the proportion of stroma, the proportion of islet, the proportion of fat, the proportion of nerve, and the proportion of lymphatic structure. Furthermore, in order to explore the relationship between acinar, stroma and islets, four additional correlation ratios were included, including the acinar-stroma ratio, islet-stroma ratio, acinar-islet ratio and functional cell (acinar + islet)-stroma ratio, and 10 types of pathological quantitative indicators were automatically extracted.
[0059] Figure 2 Schematic diagram of a model for establishing pathological classification of pancreatitis according to an embodiment of the present invention.
[0060] The pancreatic pathological image classification method of the present invention obtains pancreatic HE pathological images of patients with chronic pancreatitis, and classifies the pancreatic pathological images based on multiple predetermined feature parameters obtained by the pancreatic pathological image processing method, and determines the pathological classification of pancreatitis by latent class analysis. The classification basis can be verified as follows, and the pathological classification is related to the endocrine function and exocrine function and prognosis of patients with chronic pancreatitis.
[0061] The image classification system 200 obtained based on the pancreatic pathology image processing method and classification method of the present invention includes an image segmentation model 210 and a latent class analysis module, and the system 200 is used to obtain the pathological classification of pancreatitis. Among them, the image segmentation model 210 can receive the pancreatic HE pathological image of a patient with chronic pancreatitis with pathology, output the segmentation result, and determine the pathological classification of pancreatitis by latent class analysis based on the parameters of multiple predetermined features of the segmentation result.
[0062] Latent class analysis is a statistical analysis technique that combines latent variable theory with categorical variables. Through the latent class model (LCM), the internal latent categorical variables are used to explain the relationship between the explicit categorical variables, so that the relationship between the explicit variables can maintain its local independence after the latent categorical variables are estimated. Through latent class analysis, the similarity between the parameters of multiple predetermined features is compared to establish a classification, so that multiple categorical variables generate a new categorical variable, and the new categorical variable divides the population into several groups.
[0063] As an implementation method, determining the pathological classification of pancreatitis using latent class analysis includes the best fitting classification result based on the Akaike information criterion (AIC). In addition, the Bayesian information criterion (BIC) can also be used as a discriminant indicator of classification fit.
[0064] In one example, the segmentation results of the pathological HE images of each chronic pancreatitis patient are selected for latent class analysis. To determine the number of categories for latent class classification, classification attempts are made from one category to several categories. The number of categories corresponding to the minimum AIC and BIC values is selected to establish the pathological classification of pancreatitis. Based on the results of the latent class analysis, the population is grouped.
[0065] Figure 3 Schematic diagram of two types of pathological classification according to an embodiment of the present invention.
[0066] The parameters of multiple predetermined features are obtained by the image segmentation model 210. The pathological classification of pancreatitis is determined based on the parameters of multiple predetermined features by latent class analysis. To verify its clinical relevance, the pathological classification is explored to be related to the endocrine function, exocrine function and prognosis of patients with chronic pancreatitis. First, the parameters of multiple predetermined features are obtained by the image segmentation model 210. As an example, pancreatitis patients include patients with chronic pancreatitis and patients with recurrent acute pancreatitis. 181 patients with chronic pancreatitis are included as sample data for this study. The pancreatic HE pathological images of 19 of them are repeatedly selected for segmentation training. The training set and the validation set are image blocks taken from the segmentation mask of the pancreatic HE pathological image. The DeeplabV3+ network is used to realize the automatic segmentation of multiple tissue components of the pancreas in the pancreatic HE pathological image. The basic component proportions of the six types of pancreatic tissue components are extracted respectively. In order to better explore the correlation between the characteristic parameters and the pancreatic function, four kinds of correlation ratios between acinar, stroma and islet are obtained, totaling 10 types of pathological parameters. The 10 types of characteristic parameters obtained by referring to the segmentation results in Table 1.
[0067] Table 1 Image segmentation results
[0068]
[0069]
[0070] Combined with the data in Table 1, three pathological classifications related to prognosis were identified, as shown in Table 2. Furthermore, the similarity between multiple characteristic parameters was evaluated by latent class analysis, and a two-category pathological classification was determined with seven parameters. Figure 3 There are two pathological types: type I and type II.
[0071] Table 2 Number of classifications
[0072] Number of categories Log-likelihood function BIC 1 -1918.84(df=12) 3900.05 2 -1806.97(df=25) 3743.91 3 -1777.25(df=38) 3752.04
[0073] Among them, the pathological characteristics of type I patients are high acinar ratio, low stroma ratio, low tertiary lymphoid structure ratio, high acinar stroma ratio, high islet stroma ratio, high acinar islet ratio and high functional cell (acinar + islet) stroma ratio; the pathological characteristics of type II patients are low acinar ratio, high stroma ratio, high tertiary lymphoid structure ratio, low acinar stroma ratio, low islet stroma ratio, low acinar islet ratio and low functional cell stroma ratio. The regression model showed that the pathological classification was significantly correlated with the endocrine and exocrine function of the pancreas. The risk of endocrine or exocrine dysfunction in type II patients was 4.2 times and 3.8 times that of type I patients, respectively.
[0074] Multivariate regression analysis was used to evaluate the correlation between the classification of chronic pancreatitis and the pancreatic exocrine function and endocrine function. The pancreatic exocrine function was significantly correlated with the latent pathological classification of chronic pancreatitis, and the pancreatic endocrine function was significantly correlated with the latent pathological classification of chronic pancreatitis.
[0075] It is worth noting that those skilled in the art can understand that although the above-mentioned pathological classification is established based on pathological results, in view of the scarcity of pathological specimens, in order to popularize the pathological classification in clinical practice and improve the severity evaluation system of chronic pancreatitis, the following embodiments will combine pancreatic CT cross-sectional imaging and deep learning imaging genomics to non-invasively predict the pathological classification of pancreatitis as a supplementary scheme for the pathological classification of pancreatitis established based on pathological HE images, which can verify and predict the pathological classification. For details, see below Figure 4 and Figure 5 The embodiments in the context are complete inventive concepts formed around pathological typing and should not be viewed separately.
[0076] Figure 4 The present invention is a flowchart of a method for constructing a pancreatitis pathological classification prediction model according to an embodiment of the present invention.
[0077] Step S410: Acquire the first plain scan phase image and portal venous phase image of the pancreatic CT image of the patient with chronic pancreatitis.
[0078] In an embodiment of the present invention, chronic pancreatitis patients include chronic pancreatitis patients and recurrent acute pancreatitis patients, among which acute pancreatitis patients are regarded as mild suspected chronic pancreatitis patients and are no longer distinguished here, and are collectively referred to as chronic pancreatitis patients. Chronic pancreatitis patients for whom pancreatic CT images are obtained are patients without pathology.
[0079] The disadvantages of enhanced CT are the high X-ray dose, the risk of iodine allergy, the fact that people with poor liver and kidney function are not suitable for enhanced CT examinations, and the high cost. These factors limit the application of enhanced CT in the diagnosis of chronic pancreatitis. Plain scan CT does not require the injection of contrast agents and is non-invasive, convenient, fast, and economical.
[0080] In enhanced CT scanning, contrast agents are usually injected intravenously into blood vessels, and then flow to various organs and lesions with the flow of blood. Plain scan images refer to CT scan images obtained before contrast agent development. Contrast agent development consists of three time periods, namely the arterial phase, portal phase, and delayed phase. Among them, the arterial phase refers to the period of arterial vascular filling and development, which can show the characteristics of contrast agents flowing into the arteries; the portal phase refers to the period of portal vein vascular filling and development, which can show the characteristics of blood passing through the portal vein after the arterial phase; the delayed phase is the characteristics after the arterial phase and portal phase.
[0081] Step S420: Segment the portal venous phase image to obtain a mask associated with multiple tissues of the pancreas.
[0082] Combination Figure 5 The image classification system uses the image segmentation model 510 to segment the outlined portal phase image to obtain multiple tissue-related segmentation masks, such as masks of pancreatic parenchyma, pancreatic duct, stones, stents, pseudocysts, and pancreatic head, body, and tail partitions.
[0083] In the preprocessing stage of image segmentation, the delineated tissue mask is extracted, the image block size is adjusted, and normalization and enhancement are performed. The enhancement methods include rotation, scaling, Gaussian noise, and Gaussian blur.
[0084] Step S430: train the image segmentation model using the mask and the first plain scan period image to obtain a trained image segmentation model.
[0085] As an embodiment, using the mask and the first plain scan period image to train the image segmentation model 510 also includes aligning the first plain scan period image and the portal venous phase image, generating a mask corresponding to the first plain scan period image based on the mask, and training the image segmentation model 510 using the mask corresponding to the first plain scan period image.
[0086] Because some structures in the CT plain phase image are not clear and difficult to distinguish, the pancreatic parenchyma, pancreatic duct, stones, stents, pseudocysts and pancreatic head, body and tail partitions must be outlined on the portal phase image first, a mask is generated based on the outlined annotations, the plain phase image and the portal phase image are aligned, the mask outlined on the portal phase image is transferred to the plain phase image, and the image segmentation model 510 is trained based on the mask corresponding to the plain phase image.
[0087] In the above-mentioned implementation manner, the first plain scan phase image and the portal venous phase image are subjected to image registration, and the registration principle is based on the portal venous phase image. In some other implementation manners, the plain scan phase image, the arterial phase image, the portal venous phase image and the delayed phase image can also be calibrated based on the portal venous phase image.
[0088] Because the first plain scan phase image has been registered with the portal venous phase image, a mask corresponding to the first plain scan phase image can be generated according to the annotation of the segmented mask. The image segmentation model is trained according to the first plain scan phase image and the pancreas, stone and pancreatic duct mask therein to obtain a trained image segmentation model 510. In this way, the trained image segmentation model 510 is a segmentation model based on the plain scan phase image.
[0089] In some embodiments, the image segmentation model 210 and the image segmentation model 510 are obtained based on uuU-Unet deep learning. The Unet network mainly includes two parts, the first half is feature extraction, and the second half is used for upsampling. Using the mask and the plain scan images of the pancreatic CT images of patients with chronic pancreatitis to train the image segmentation model 510, the trained image segmentation model 510 also includes obtaining the plain scan images of the pancreatic CT images of patients with chronic pancreatitis, and training the image segmentation model 210 according to the mask and the plain scan images of the pancreatic CT images of patients with chronic pancreatitis, that is, using the mask and the plain scan images of the pancreatic CT images of patients with chronic pancreatitis as training sets, training the image segmentation model 510 through the training set, and separately finding a batch of plain scan images of the pancreatic CT images of patients with chronic pancreatitis as verification sets, and verifying the image segmentation model 210 with the verification set, specifically refer to step S440.
[0090] Step S440: input the second plain scan image of the patient to be tested into the trained image segmentation model to obtain the pancreatic parenchymal area image of the second plain scan image, and extract multiple imaging omics features based on the pancreatic parenchymal area image.
[0091] As an embodiment, extracting radiomics features based on the pancreatic parenchymal region image includes performing dimensionality reduction processing on the radiomics features to obtain predetermined features.
[0092] Parenchyma and pancreatic duct are key areas for feature extraction. Segmentation is helpful to eliminate other interference factors in the area and improve the stability and reliability of feature extraction. Through image segmentation, stones, stents, pancreatic ducts and pseudocysts in the pancreas are removed to obtain the pancreatic parenchyma area. Based on the image of the pancreatic parenchyma area, radiomics features are extracted and the radiomics features are screened by dimensionality reduction to obtain the predetermined features, which are the main features.
[0093] In one example, features with correlation coefficients > 0.9 were screened out through correlation analysis (Spearman), and then the least absolute shrinkage and selection operator logistic regression methods were used to screen out the most relevant features.
[0094] Step S450: Select predetermined features from multiple imaging omics features, calculate omics scores, and construct a pathological classification prediction model based on the omics scores.
[0095] A deep learning radiomics score calculation formula was established based on predetermined features, and the typing threshold was determined by maximizing the Youden index, and then a regression model was formed.
[0096] In one example, since there is no significant difference between the extracted automatic quantitative indicators in the pathological classification, a deep learning radiomics score calculation formula is established based only on the features derived from the pancreatic parenchyma and pancreatic duct. The value of the analyte corresponding to the maximum Youden index is selected as the classification threshold. Pancreatitis patients with values below the threshold are predicted to be type I, and pancreatitis patients with values above the threshold are predicted to be type II.
[0097] Figure 5 Schematic diagram of a pathological classification prediction model according to an embodiment of the present invention.
[0098] The pancreatitis case classification prediction method of the present invention obtains a pancreatic CT image of a patient with chronic pancreatitis, and obtains a pancreatitis pathological classification prediction result based on a pathological classification prediction model obtained by a pancreatitis pathological classification prediction model construction method.
[0099] The image classification system 500 obtained based on the pancreatitis pathological classification prediction model construction method and the pancreatitis pathological classification prediction method of the present invention includes an image segmentation model 510 and a pathological classification prediction model 520. The image segmentation model 510 can receive the CT plain scan image of the patient to be tested and output the pancreatic mask. The pathological classification prediction model 520 performs pathological classification judgment on the CT plain scan image according to the pancreatic mask.
[0100] Figure 5 Provided is an end-to-end image classification system 500, in which the CT plain scan images of the patient to be tested are input into an image segmentation model, and the output of the image segmentation model is input into a pathological classification prediction model. There is no need to obtain the patient's pancreatic pathological section images or pancreatic enhanced CT images. Only the patient's CT plain scan images need to be obtained, and the existing non-pathological enhanced CT images are used in combination with deep learning to achieve pancreatitis pathological classification prediction based on non-enhanced CT images. The system has the advantages of being non-invasive, having no risk of iodine allergy, being inexpensive, and being friendly to people with poor liver and kidney function.
[0101] The image classification model 510 is obtained by the following method.
[0102] First, the first plain scan phase image and portal venous phase image of the pancreatic CT image of the patient with chronic pancreatitis are obtained;
[0103] Second, the portal phase images were segmented to obtain masks associated with multiple tissues of the pancreas;
[0104] Again, the mask and the first plain scan period image are used to train the image segmentation model to obtain a trained image segmentation model.
[0105] Based on the parameters of multiple predetermined features output by the image segmentation model 510, with the assistance of the pathology classification prediction model 520, the segmentation and classification results can be combined with the pathology, clinical, and CT features to provide a final pathology classification prediction result.
[0106] The pathological classification prediction model 520 classifies the plain scan image according to the pancreatic mask. The pathological classification prediction model 520 is trained by the following method.
[0107] First, the second plain scan image of the patient to be tested is input into the trained image segmentation model 510 to obtain the pancreatic parenchymal area image of the second plain scan image, and multiple imaging omics features are extracted based on the pancreatic parenchymal area image.
[0108] In addition, predetermined features are selected from a plurality of imaging omics features, omics scores are calculated, and a pathology classification prediction model 520 is constructed based on the omics scores.
[0109] Using CT images without pathology as a test set, the clinical applicability of the pathological classification prediction model was verified. Through correlation analysis, it was found that both type I and type II showed significant differences between pancreatic endocrine function and exocrine function. The risk of endocrine and exocrine dysfunction in patients with type II was 2.5 times that of patients with type I. The pathological classification prediction model can correctly reflect the severity of the disease and has a significant correlation with pancreatic endocrine and exocrine functions, which has practical clinical significance.
[0110] Figure 6 is a hardware structure block diagram of an electronic device according to an embodiment of the present invention.
[0111] like Figure 6 As shown, the electronic device 600 may include one or more processors 602, a system control logic 608 connected to at least one of the processors 602, a system memory 604 connected to the system control logic 608, a non-volatile memory (NVM) 606 connected to the system control logic 608, and a network interface 610 connected to the system control logic 608.
[0112] The processor 602 may include one or more single-core or multi-core processors. The processor 602 may include any combination of a general-purpose processor and a special-purpose processor (e.g., a graphics processor, an application processor, a baseband processor, etc.). In an embodiment of the present invention, the processor 602 may be configured to execute the following operations: Figure 1 The pancreatic pathology image processing method and Figure 4 The method for constructing a predictive model for pancreatitis pathological classification is shown.
[0113] In some embodiments, system control logic 608 may include any suitable interface controller to provide any suitable interface to at least one of processors 602 and / or any suitable device or component in communication with system control logic 608 .
[0114] In some embodiments, the system control logic 608 may include one or more memory controllers to provide an interface to the system memory 604. The system memory 604 may be used to load and store data and / or instructions. In some embodiments, the system memory 604 of the electronic device 600 may include any suitable volatile memory, such as a suitable dynamic random access memory (DRAM).
[0115] The non-volatile memory 606 may include one or more tangible, non-transitory computer-readable media for storing data and / or instructions. In some embodiments, the non-volatile memory 606 may include any suitable non-volatile memory such as flash memory and / or any suitable non-volatile storage device, such as at least one of a HDD (Hard Disk Drive, hard disk drive), a CD (Compact Disc, optical disk) drive, and a DVD (Digital Versatile Disc, digital versatile disc) drive.
[0116] The non-volatile memory 606 may include a portion of storage resources installed on a device of the electronic device 600, or it may be accessible by the device but not necessarily a portion of the device. For example, the non-volatile memory 606 may be accessed over a network via the network interface 610.
[0117] In particular, the system memory 604 and the non-volatile memory 606 may respectively include: a temporary copy and a permanent copy of the instruction 620. The instruction 620 may include: when executed by at least one of the processors 602, causing the electronic device 600 to implement the following Figure 1 The pancreatic pathology image processing method and Figure 4 Instructions of the processing method of the method for building a pancreatitis pathological classification prediction model are shown. In some embodiments, the instructions 620, hardware, firmware and / or software components thereof may be additionally / alternatively placed in the system control logic 608, the network interface 610 and / or the processor 602.
[0118] The network interface 610 may include a transceiver for providing a radio interface for the electronic device 600, thereby communicating with any other suitable device (e.g., a front-end module, an antenna, etc.) through one or more networks. In some embodiments, the network interface 610 may be integrated with other components of the electronic device 600. For example, the network interface 610 may be integrated with at least one of the processor 602, the system memory 604, the NVM 606, and a firmware device (not shown) having instructions. When at least one of the processors 602 executes the instructions, the electronic device 600 implements Figure 1 and Figure 4 One or more of the various embodiments shown.
[0119] The network interface 610 may further include any suitable hardware and / or firmware to provide a multiple-input multiple-output radio interface. For example, the network interface 610 may be a network adapter, a wireless network adapter, a telephone modem, and / or a wireless modem.
[0120] In one embodiment, at least one of the processors 602 may be packaged together with one or more controllers for the system control logic 608 to form a system in package (SiP). In one embodiment, at least one of the processors 602 may be integrated on the same die with one or more controllers for the system control logic 608 to form a system on a chip (SoC).
[0121] The electronic device 600 may further include an input / output (I / O) device 612 connected to the system control logic 608. The I / O device 612 may include a user interface to enable a user to interact with the electronic device 600; the design of the peripheral component interface enables the peripheral components to interact with the electronic device 600. In some embodiments, the electronic device 600 further includes a sensor for determining at least one of an environmental condition and location information related to the electronic device 600.
[0122] In some embodiments, input / output (I / O) devices 612 may include, but are not limited to, a display (e.g., a liquid crystal display, a touch screen display, etc.), speakers, a microphone, one or more cameras (e.g., a still image camera and / or a video camera), a flashlight (e.g., an LED flash), and a keyboard.
[0123] In some embodiments, the peripheral component interface may include, but is not limited to, a non-volatile memory port, an audio jack, and a power interface.
[0124] It is to be understood that the structure illustrated in the embodiment of the present invention does not constitute a specific limitation on the electronic device 600. In other embodiments of the present application, the electronic device 600 may include more or fewer components than shown in the figure, or combine some components, or split some components, or arrange the components differently. The components shown in the figure may be implemented in hardware, software, or a combination of software and hardware.
[0125] Program code can be applied to input instructions to perform the functions described in the present invention and generate output information. The output information can be applied to one or more output devices in a known manner. For the purposes of this application, a system for processing instructions including processor 602 includes any system having a processor such as a digital signal processor (DSP), a microcontroller, an application specific integrated circuit (ASIC), or a microprocessor.
[0126] Program code can be implemented with high-level programming language or object-oriented programming language to communicate with the processing system. When necessary, program code can also be implemented with assembly language or machine language. In fact, the mechanism described in the present invention is not limited to the scope of any specific programming language. In either case, the language can be a compiled language or an interpreted language.
[0127] According to one embodiment of the present invention, a computer-readable storage medium is further proposed, wherein at least one computer instruction is stored in the computer-readable storage medium, and the at least one instruction is loaded and executed by a processor to implement the aforementioned pancreatic pathology image processing method, pancreatic pathology image classification method, pancreatitis pathology classification prediction model construction method or pancreatitis pathology classification prediction method.
[0128] According to one embodiment of the present invention, a computer program product is also proposed, which includes computer instructions. When the computer instructions are executed, the aforementioned pancreatic pathology image processing method, pancreatic pathology image classification method, pancreatitis pathology classification prediction model construction method or pancreatitis pathology classification prediction method are implemented.
[0129] Illustrative embodiments of the present invention include, but are not limited to, a pancreatitis pathology image processing method, a classification method, a typing prediction model construction method, a prediction method, an apparatus, a medium, and a program product.
[0130] The various aspects of the illustrative embodiments will be described using terms commonly used by those skilled in the art to convey the essence of their work to other persons skilled in the art. However, it is apparent to those skilled in the art that some alternative embodiments may be implemented using some of the features described. For purposes of explanation, specific numbers and configurations are set forth so that the illustrative embodiments are more thoroughly understood. However, it is apparent to those skilled in the art that alternative embodiments may be implemented without the specific details. In some other cases, some well-known features are omitted or simplified herein to avoid obscuring the illustrative embodiments of the present invention.
[0131] In addition, various operations will be described as multiple separate operations in a manner that is most helpful for understanding the illustrative embodiments; however, the order of description should not be interpreted as implying that these operations must rely on the order of description, and many of the operations can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can also be rearranged. When the described operations are completed, the process can be terminated, but can also have additional steps not included in the drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, etc.
[0132] References in the specification to "an example", "an example", "an embodiment", "an implementation", etc. indicate that the described embodiment may include a particular feature, structure, or property, but each embodiment may or may not necessarily include the particular feature, structure, or property. Moreover, these phrases are not necessarily directed to the same embodiment. In addition, when particular features are described in conjunction with a specific embodiment, the knowledge of those skilled in the art can affect the combination of these features with other embodiments, whether or not these embodiments are explicitly described.
[0133] Unless the context dictates otherwise, the terms "comprising," "having," and "including" are synonymous. The phrase "A and / or B" means "(A), (B), or (A and B)."
[0134] As used herein, the term "module" may refer to, be part of, or include: memory (shared, dedicated, or group) for running one or more software or firmware programs, application-specific integrated circuits (ASICs), electronic circuits and / or processors (shared, dedicated, or group), combinational logic circuits, and / or other suitable components that provide the described functionality.
[0135] In the accompanying drawings, some structural or method features may be shown in a specific arrangement and / or order. However, it should be understood that such a specific arrangement and / or order is not necessary. Instead, in some embodiments, these features may be described in a manner and / or order different from that shown in the illustrative drawings. In addition, the structural or method features included in a specific figure do not mean that all embodiments need to include such features. In some embodiments, these features may not be included or may be combined with other features.
[0136] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.
[0137] Similarly, it should be understood that in order to streamline the present invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those expressly recited in each claim. Rather, as reflected in the claims, inventive aspects lie in less than all of the features of the individual embodiments previously disclosed. Therefore, the claims that follow the detailed description are hereby expressly incorporated into the detailed description, with each claim itself serving as a separate embodiment of the present invention.
[0138] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.
[0139] In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims, any one of the claimed embodiments may be used in any combination.
Claims
1. A pancreatic pathology image processing method, used in electronic equipment, characterized in that: The method comprises: Obtain pancreatic HE pathological images of patients with chronic pancreatitis; Segmenting the pancreatic HE pathology image to obtain a mask associated with a plurality of tissues of the pancreas; extracting an image block based on the mask; The image blocks are used to train an image segmentation model, and parameters of a plurality of predetermined features are obtained.
2. The method according to claim 1, characterized in that Extracting the image block based on the mask includes flipping, rotating, Gaussian blurring and color enhancement processing on the image block, and standardizing the image block.
3. The method according to claim 1, characterized in that: Using the image block to train the image segmentation model includes binarizing the image block, deriving segmentation areas of the multiple tissues by traversing pixel points of the image block, and obtaining parameters of the multiple predetermined features based on the segmentation areas.
4. A pancreatic pathology image classification method for electronic equipment, characterized in that: The method comprises: Obtain pancreatic HE pathological images of patients with chronic pancreatitis; Using the method as claimed in claim 1 to obtain parameters of a plurality of predetermined characteristics, Based on the parameters of the multiple predetermined features, latent class analysis is used to determine the pathological classification of pancreatitis, and the pathological classification is related to the endocrine function, exocrine function and prognosis of the chronic pancreatitis patient.
5. The method according to claim 4, characterized in that The method of determining the pathological classification of pancreatitis by latent class analysis includes the best fitting classification result based on Akaike's information criterion (AIC).
6. A method for constructing a prediction model for pancreatitis pathological typing, characterized in that: The model building method comprises: Acquiring the first plain scan phase image and portal venous phase image of the pancreatic CT image of a patient with chronic pancreatitis; segmenting the portal phase image to obtain a mask associated with a plurality of tissues of the pancreas; Using the mask and the first plain scan period image to train an image segmentation model to obtain a trained image segmentation model; Inputting the second plain scan image of the patient to be tested into the trained image segmentation model to obtain a pancreatic parenchymal region image of the second plain scan image, and extracting a plurality of radiomics features based on the pancreatic parenchymal region image; Predetermined features from the multiple imaging omics features are selected, omics scores are calculated, and a pathology classification prediction model is constructed according to the omics scores.
7. The method according to claim 6, characterized in that The method of training the image segmentation model using the mask and the first plain scan period image also includes aligning the first plain scan period image and the portal venous phase image, generating a mask corresponding to the first plain scan period image based on the mask, and training the image segmentation model using the mask corresponding to the first plain scan period image.
8. The method according to claim 6, characterized in that Extracting the radiomics feature based on the pancreatic parenchymal region image includes performing dimensionality reduction processing on the radiomics feature to obtain the predetermined feature.
9. A method for predicting pancreatitis pathological typing, characterized in that: The method comprises: Obtain pancreatic CT images of patients with chronic pancreatitis; The prediction results of pancreatitis pathological classification were obtained based on the pathological classification prediction model, among which: The pathological classification prediction model is obtained according to the model construction method according to claim 6.
10. An electronic device, characterized in that: The device includes a processor and a memory storing computer-executable instructions, and the processor is configured to execute the instructions to implement the pancreatic pathology image processing method according to claim 1, the pancreatic pathology image classification method according to claim 2, the pancreatitis pathology classification prediction model construction method according to claim 6, or the pancreatitis pathology classification prediction method according to claim 9.
11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one computer instruction, and the at least one instruction is loaded and executed by the processor to implement the pancreatic pathology image processing method according to claim 1, the pancreatic pathology image classification method according to claim 2, the pancreatitis pathology classification prediction model construction method according to claim 6, or the pancreatitis pathology classification prediction method according to claim 9.
12. A computer program product, characterized in that The computer program product includes computer instructions, which, when executed, implement the pancreatic pathology image processing method according to claim 1, the pancreatic pathology image classification method according to claim 2, the pancreatitis pathology classification prediction model construction method according to claim 6, or the pancreatitis pathology classification prediction method according to claim 9.
Citation Information
Patent Citations
Pancreas CT image classification method, image classification model, electronic equipment and medium
CN116416239A
Cited By
Operation area inflammation degree grading method based on pancreatic peripheral fat image features
CN120747051A
Interstitial lung disease HRCT image analysis method and system based on artificial intelligence
CN121544605A
