Relative blood flow estimation method and system based on pancreas plain-scan CT image

Abstract

The invention relates to the technical field of medical image processing, in particular to a relative blood flow estimation method and system based on a pancreas plain-scan CT image, and the method comprises the steps: constructing a pancreas enhanced CT estimation model, and carrying out the pre-training based on the pancreas enhanced CT estimation model to obtain a brightness estimation module; obtaining a target plain-scan CT image; predicted arterial phase CT image brightness codes, portal vein phase CT image brightness codes and vein phase CT image brightness codes are obtained based on the target plain-scan CT image and a brightness estimation module; based on CT image brightness coding in different phases, a target tetrahedral structure is constructed in a high-dimensional space; obtaining a target relative blood flow volume based on the ratio of the volume of the target tetrahedral structure to the volume of the reference tetrahedral structure; whether the target relative blood flow exceeds a threshold value or not is judged, and the subject is classified based on the judgment result. According to the method, effective assistance is provided for early screening of pancreatic cancer on the basis of processing of the plain-scan CT image.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a relative blood flow estimation method and system based on pancreatic plain scan CT images. Background Art

[0002] Pancreatic cancer is a highly malignant tumor whose early symptoms are often not obvious, resulting in most patients being in the late stages of the disease by the time of diagnosis. This greatly increases the difficulty of treatment and the patient's mortality rate. Therefore, improving the early detection rate of pancreatic cancer is of great significance for improving patient prognosis. Among current medical diagnostic methods, CT enhanced angiography provides important anatomical and functional information for the diagnosis of pancreatic cancer. CT scans performed after intravenous injection of contrast agent can clearly show the blood flow of the pancreas and surrounding tissues, helping doctors to more accurately determine the presence and nature of the tumor. This technology not only increases the possibility of early detection of pancreatic cancer, but also provides a valuable reference for clinical decision-making.

[0003] However, while CT contrast enhancement technology has important application value in the diagnosis of pancreatic cancer, it still has some shortcomings and limitations. First, this technology generally requires a certain level of prognosis for pancreatic cancer to allow for necessary preparation and planning before the scan. However, because pancreatic cancer's early symptoms are not obvious, doctors and patients often find it difficult to make an accurate prognosis in the early stages of the disease, which limits the widespread application of CT contrast enhancement technology. CT contrast enhancement technology also requires patients to undergo certain preparations, such as fasting and injection of contrast agent, before the CT scan. This relatively cumbersome process not only increases patient discomfort but may also prevent some patients from undergoing the scan without preparation, thus missing the opportunity to detect the disease early. CT contrast enhancement technology also requires a large amount of medical resources, including CT equipment and medical staff. In the context of limited medical resources, the widespread use of this technology may have a certain impact on the diagnosis and treatment of other patients.

[0004] At the same time, there is currently no tool to estimate pancreatic relative blood flow, forcing doctors to rely on image recognition technology to indirectly determine pancreatic blood flow. This method is not only time-consuming and labor-intensive, but also relatively inaccurate, making it difficult to meet clinical needs.

[0005] Therefore, developing a new technology that can use plain scan CT results to estimate contrast intensity through machine learning methods and further provide estimated blood flow is of great significance for improving the early detection rate and treatment effect of pancreatic cancer. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and system for estimating relative blood flow based on pancreatic plain scan CT images in order to solve at least one of the above technical problems.

[0007] The present invention achieves the above-mentioned purpose through the following technical solutions: A relative blood flow estimation method based on pancreatic plain scan CT images comprises the following steps: Constructing a pancreatic enhanced CT estimation model, and pre-training a brightness estimation module based on the pancreatic enhanced CT estimation model; Acquiring a plain CT image of the pancreas of the subject, that is, a target plain CT image; Acquire predicted brightness codes of the arterial phase CT image, the portal venous phase CT image, and the venous phase CT image based on the target plain scan CT image and the brightness estimation module; The target plain scan CT image brightness code is defined as the reference point, and based on the CT image brightness code at different phases, the target tetrahedron structure is constructed in high-dimensional space. A target relative blood flow is obtained based on the ratio of the target tetrahedral structure volume to the reference tetrahedral structure volume; wherein the reference tetrahedral structure volume is an average value of tetrahedral structure volumes formed in a high-dimensional space by brightness encoding of CT images of a healthy pancreas at different phases; It is determined whether the target relative blood flow exceeds a threshold, and the subject is classified based on the determination result.

[0008] Furthermore, the pancreatic enhanced CT estimation model includes: a first training model and a second training model; The first training model includes: an image encoding module, a splicing module, and a first image decoding module; the plain scan CT image data is input into the image encoding module, the input of the splicing module is the output of the image encoding module and the brightness code of the plain scan CT image, and the data obtained by processing the splicing module is input into the first image decoding module to finally obtain an enhanced CT image; wherein the brightness code of the plain scan CT image is used as a reference point; The second training model includes: an alignment module, an image encoding module, a brightness estimation module, a splicing module, and a second image decoding module; plain scan CT image data are respectively input into the alignment module and the brightness estimation module, the output of the alignment module is processed by the image encoding module, the output of the image encoding module and the output of the brightness estimation module are jointly input into the splicing module, and the output of the splicing module is processed by the second image decoding module to obtain a corresponding enhanced CT image; Among them, the image encoding module and the splicing module in the second training model are obtained through training with the first training model and remain unchanged in the second training model; the initial parameters of the second image decoding module adopt the parameters finally obtained through training with the first image decoding module; and the parameters of the second image decoding module are updated through training with the second training model.

[0009] Furthermore, the brightness estimation module includes: an arterial phase brightness estimation unit, a portal venous phase brightness estimation unit and a venous phase brightness estimation unit; Each unit is used to predict the brightness relationship between the plain scan CT image and the enhanced CT image of each phase.

[0010] Furthermore, the threshold is set in the following manner: Obtaining relative blood flow values ​​of all subjects based on the data set used to train the pancreatic enhanced CT estimation model; The relative blood flow values ​​of all subjects were fitted with Poisson distribution to obtain the standard deviation of the relative blood flow Poisson distribution. , threshold = 1+ .

[0011] Furthermore, the subjects are classified based on the judgment results, including: When the target relative blood flow does not exceed the threshold, classifying the subject as a low-risk group for pancreatic cancer; When the target relative blood flow exceeds the threshold, the subject is classified as a high-risk group for pancreatic cancer, and further pancreatic examination is performed on the high-risk group for pancreatic cancer.

[0012] A relative blood flow estimation system based on pancreatic plain scan CT images, comprising: A construction module is used to construct a pancreatic enhanced CT estimation model and obtain a brightness estimation module based on pre-training of the pancreatic enhanced CT estimation model; an acquisition module, configured to acquire a plain CT image of the pancreas of the subject, that is, a target plain CT image; A prediction module, configured to obtain, based on the target plain scan CT image and the brightness estimation module, predicted: brightness coding of the arterial phase CT image, brightness coding of the portal venous phase CT image, and brightness coding of the venous phase CT image; A calculation module is used to define the brightness code of the target plain scan CT image as a reference point, and to construct the target tetrahedron structure in a high-dimensional space based on the brightness code of the CT image at different phases; The calculation module is further configured to obtain a target relative blood flow based on a ratio of the target tetrahedral structure volume to a reference tetrahedral structure volume; wherein the reference tetrahedral structure volume is an average value of tetrahedral structure volumes formed by brightness encoding of CT images of a healthy pancreas at different phases in a high-dimensional space; The judgment module is used to judge whether the target relative blood flow exceeds a threshold value and classify the subject based on the judgment result.

[0013] Furthermore, in the construction module, the pancreatic enhanced CT estimation model includes: a first training model and a second training model; The first training model includes: an image encoding module, a splicing module, and a first image decoding module; the plain scan CT image data is input into the image encoding module, the input of the splicing module is the output of the image encoding module and the brightness code of the plain scan CT image, and the data obtained by processing the splicing module is input into the first image decoding module to finally obtain an enhanced CT image; wherein the brightness code of the plain scan CT image is used as a reference point; The second training model includes: an alignment module, an image encoding module, a brightness estimation module, a splicing module, and a second image decoding module; plain scan CT image data are respectively input into the alignment module and the brightness estimation module, the output of the alignment module is processed by the image encoding module, the output of the image encoding module and the output of the brightness estimation module are jointly input into the splicing module, and the output of the splicing module is processed by the second image decoding module to obtain a corresponding enhanced CT image; Among them, the image encoding module and the splicing module in the second training model are obtained through training with the first training model and remain unchanged in the second training model; the initial parameters of the second image decoding module adopt the parameters finally obtained through training with the first image decoding module; and the parameters of the second image decoding module are updated through training with the second training model.

[0014] Further, in the construction module, the brightness estimation module includes: an arterial phase brightness estimation unit, a portal venous phase brightness estimation unit and a venous phase brightness estimation unit; Each unit is used to predict the brightness relationship between the plain scan CT image and the enhanced CT image of each phase.

[0015] Furthermore, in the judgment module, the threshold is set in the following manner: Obtaining relative blood flow values ​​of all subjects based on the data set used to train the pancreatic enhanced CT estimation model; The relative blood flow values ​​of all subjects were fitted with Poisson distribution to obtain the standard deviation of the relative blood flow Poisson distribution. , threshold = 1+ .

[0016] Furthermore, in the judgment module, classifying the examinee based on the judgment result includes: When the target relative blood flow does not exceed the threshold, classifying the subject as a low-risk group for pancreatic cancer; When the target relative blood flow exceeds the threshold, the subject is classified as a high-risk group for pancreatic cancer, and further pancreatic examination is performed on the high-risk group for pancreatic cancer.

[0017] The beneficial effects of the present invention are: First, this invention addresses the challenge of early pancreatic cancer detection by proposing an innovative solution. Using plain scan CT images, it can predict the enhanced CT effect and further assess the patient's relative blood flow, providing a new and convenient method for early detection of pancreatic cancer. This invention not only fills a gap in pancreatic cancer indication methods but also significantly improves the detection rate of pancreatic cancer.

[0018] Secondly, this method is performed simultaneously with a plain CT scan, eliminating the need for additional testing procedures or patient preparation, significantly reducing patient discomfort and conserving medical resources. Doctors can receive pancreatic cancer risk alerts while their patients undergo a conventional CT scan (plain CT), enabling them to take further diagnostic or treatment measures promptly.

[0019] Furthermore, the present invention assists doctors in quickly and directly predicting pancreatic cancer using pancreatic CT scan images, making the use of enhanced contrast imaging more targeted. This not only improves diagnostic accuracy but also avoids unnecessary use of contrast agents, reducing patient risks.

[0020] Finally, the model used in this invention is streamlined and efficient, running on standard computers or mobile phones without the need for large-scale computing. This makes the invention more widely applicable, making it easier to promote its use in medical institutions at all levels and bring benefits to more patients.

[0021] In summary, the present invention not only improves the early detection rate of pancreatic cancer, but also provides doctors with a convenient and accurate diagnostic tool, while saving medical resources and reducing the burden on patients, with significant social benefits and clinical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is an image showing the entire process of the brightness of a subject's pancreas enhanced CT image changing from low to high; Figure 2 This is a flow chart of a relative blood flow estimation method based on pancreatic plain scan CT images according to one embodiment of the present invention; Figure 3A schematic diagram of constructing a target tetrahedral structure in a high-dimensional space according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the first training model structure according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a second training model according to an embodiment of the present invention; Figure 6 A schematic diagram of the workflow of an image encoding module according to one embodiment of the present invention; Figure 7 A schematic diagram of the workflow of an image decoding module according to an embodiment of the present invention; Figure 8 A schematic diagram of the workflow of a brightness estimation module according to an embodiment of the present invention; Figure 9 This is a structural diagram of a relative blood flow estimation system based on pancreatic plain scan CT images according to one embodiment of the present invention. DETAILED DESCRIPTION

[0023] The present invention will now be discussed with reference to exemplary embodiments. It should be understood that the embodiments discussed are only intended to enable those skilled in the art to better understand and implement the present invention, rather than to imply any limitation on the scope of the present invention.

[0024] As used herein, the term "including" and variations thereof are to be interpreted as open-ended terms meaning "including, but not limited to." The term "based on" is to be interpreted as "based, at least in part, on." The terms "one embodiment" and "an embodiment" are to be interpreted as "at least one embodiment."

[0025] During the pancreatic enhanced CT scan, contrast agent needs to be injected. According to the stage of contrast agent metabolism in the blood vessels, it can be divided into plain scan phase, arterial phase, portal vein phase, and venous phase. The enhanced CT images of different phases have different characteristics in brightness. Figure 1 The image shows the whole process of the brightness change of the pancreas enhanced CT image of a certain subject from low to high, where (1) to (16) are images collected in chronological order. Figure 1As shown in Figure 2, the brightness of enhanced CT images reflects contrast agent concentration, which is positively correlated with local blood flow. For example, during the arterial phase, the brightness of enhanced CT images primarily reflects arterial blood supply, and cancerous areas may have lower brightness than normal due to vascular abnormalities. During the portal venous / venous phase, the brightness of enhanced CT images reflects capillary permeability and interstitial enhancement, and high pressure in the cancerous interstitial space may suppress enhancement. Using the brightness of CT images during the plain scan phase as a baseline eliminates the influence of individual tissue density differences. Compared to the blood flow characteristics of a healthy pancreas, which exhibit uniform enhancement and temporal brightness changes consistent with physiological hemodynamics, CT images of pancreatic cancer patients exhibit characteristics of poor blood supply, resulting in insufficient enhancement during the arterial phase and delayed enhancement during the portal venous phase.

[0026] Based on the above theory, the present invention utilizes an innovative model construction method to first predict the subject's four-phase contrast-enhanced CT images using plain CT images of the pancreas. Subsequently, a high-dimensional geometric model (specifically, a tetrahedron structure) is constructed based on the brightness-encoded information at different phases. By calculating the volume changes of these models under different blood flow conditions, the present invention can quantify the blood flow differences between cancerous and normal pancreatic tissue. Based on this quantification, subjects can be effectively categorized, identifying those whose relative blood flow exceeds a preset threshold as high-risk groups, allowing for more detailed pancreatic cancer risk assessment and appropriate medical examinations.

[0027] Example 1 Figure 2 This is a flow chart of a relative blood flow estimation method based on pancreatic plain scan CT images according to one embodiment of the present invention. Figure 3 Schematic diagram of constructing a target tetrahedron structure in a high-dimensional space according to an embodiment of the present invention. Figure 2-3 As shown, according to one embodiment of the present invention, a relative blood flow estimation method based on pancreatic plain scan CT images includes the following steps: Step S102: constructing a pancreatic enhanced CT estimation model, and obtaining a brightness estimation module based on pre-training of the pancreatic enhanced CT estimation model; Step S104, obtaining a plain CT image of the subject's pancreas, that is, a target plain CT image; Step S106, based on the target plain scan CT image and the brightness estimation module, obtain the predicted brightness coding of the arterial phase CT image, the brightness coding of the portal venous phase CT image, and the brightness coding of the venous phase CT image; Step S108: Define the target plain scan CT image brightness code as a reference point. In this embodiment, the reference point is "00000000". Based on the CT image brightness codes at different phases, construct a target tetrahedron structure in a high-dimensional space. Step S110, obtaining a target relative blood flow based on a ratio of a target tetrahedral structure volume to a reference tetrahedral structure volume; wherein the reference tetrahedral structure volume is an average value of tetrahedral structure volumes formed by brightness encoding of CT images of a healthy pancreas at different phases in a high-dimensional space; Step S112: determine whether the target relative blood flow exceeds a threshold, and classify the subject based on the determination result.

[0028] In this embodiment, a relative blood flow estimation method based on pancreatic plain CT images is proposed. The detailed implementation steps are as follows: First, in step S102, a pancreatic enhanced CT estimation model is constructed, and a brightness estimation module is pre-trained based on this model. This module, the core of subsequent analysis, predicts the brightness characteristics of enhanced CT images at different phases (arterial phase t1, portal venous phase t2, and venous phase t3) based on pancreatic plain CT images. During the training of the pancreatic enhanced CT estimation model, a large amount of pancreatic CT image data is pre-collected, including both plain and enhanced CT images of the pancreas from the same subject, to construct a pancreatic CT image dataset. The size of the dataset is determined based on actual needs; for example, CT images of 500 subjects are used to construct the pancreatic CT image dataset. The pancreatic enhanced CT estimation model is trained based on the pre-constructed pancreatic CT image dataset to obtain the brightness estimation module.

[0029] Next, in step S104, a plain CT image of the subject's pancreas, ie, a target plain CT image, is acquired. This is the starting point of the entire analysis process and provides basic image information of the pancreas in an unenhanced state.

[0030] Then, in step S106, a pre-trained brightness estimation module is used to predict the CT image brightness codes for the arterial, portal, and venous phases based on the target plain scan CT image. This brightness estimation module captures the impact of enhanced CT contrast agents on image brightness during different time periods of their in vivo travel, thereby simulating the brightness characteristics of enhanced CT and providing basic data for subsequent tetrahedron structure construction.

[0031] In step S108, the target unenhanced CT image brightness code t0 is defined as the reference point "00000000". Combined with the predicted CT image brightness codes at different phases, a target tetrahedron structure is constructed in high-dimensional space. This tetrahedron structure, with the brightness codes of the unenhanced phase and different enhancement phases as vertices, reflects the dynamic process of pancreatic blood flow changes over time. Each vertex of the tetrahedron represents the pancreatic brightness feature at a specific time point, and the volume of the tetrahedron can be regarded as a comprehensive measure of these brightness features in high-dimensional space.

[0032] For example, the brightness codes of the CT images corresponding to each phase are mapped into N-dimensional space, and the non-coplanar coordinate points P0, P1, P2, and P3 are obtained. Based on the coordinates of the brightness codes of each CT image in N-dimensional space, an N-dimensional tetrahedron is constructed. Where N is the number of bits encoded by the encoder. The volume calculation process of the tetrahedron is as follows: Step 1: Take point P0 (t0, "00000000") as the reference point and construct a vector starting from P0; Step 2: construct the matrix; Use the vectors obtained in step 1 as column vectors (or row vectors) to form a 3×3 square matrix A; Step 3, calculate the volume V of the tetrahedron in N-dimensional space; Here, det represents the determinant.

[0033] Subsequently, in step S110, the target relative blood flow is determined by comparing the volume ratio of the target tetrahedron structure with the reference tetrahedron structure. The reference tetrahedron volume is the average volume of the tetrahedron structure formed by the intensity encoding of CT images of a healthy pancreas at different phases in high-dimensional space. It represents a normal blood flow range. By comparing the volume ratio of the target tetrahedron to the reference tetrahedron, we can indirectly reflect the relative changes in blood flow and thus assess the blood flow status of the pancreas.

[0034] Furthermore, in step S112, a determination is made as to whether the target relative blood flow exceeds a preset threshold. This threshold, derived from extensive pancreatic data statistics, distinguishes between normal and abnormal blood flow ranges. If the target relative blood flow exceeds the threshold, it indicates an abnormality in the subject's pancreatic blood flow, suggesting a risk of pancreatic cancer. The system will alert the physician to this risk and recommend further testing for a definitive diagnosis.

[0035] The four images of interest in this implementation (plain scan, arterial scan, portal venous scan, and venous scan) are similar in shape, but their brightness varies due to subtle movements of the body during CT scanning and the varying duration of CT contrast agents in the body. These four brightness levels form a sequential scanning cycle, fully describing the blood flow process: t0 plain scan without enhancement → t1 arterial scan → t2 portal venous scan → t3 venous scan → t0 plain scan without enhancement.

[0036] Based on this time series scanning cycle, the algorithm extracts two time series evolution sequences in the image feature space: one is the image feature sequence , reflecting the changes in pancreatic morphology; the other is the brightness feature sequence , which reflects the changes in blood flow. By further analyzing the brightness feature sequence, an accurate estimation of relative blood flow can be achieved.

[0037] The present invention combines plain scan CT images with a brightness estimation module, and utilizes the tetrahedron structure volume ratio method to implement a blood flow estimation method based on non-enhanced pancreatic CT scan images. This method not only provides an auxiliary means for early screening of pancreatic cancer, but also has the advantages of simple operation and non-invasiveness, and has broad clinical application prospects.

[0038] Figure 4 This is a schematic diagram of the first training model structure according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a second training model according to an embodiment of the present invention; Figure 6 A schematic diagram of the workflow of an image encoding module according to one embodiment of the present invention; Figure 7 A schematic diagram of the workflow of an image decoding module according to an embodiment of the present invention; Figure 8 FIG. 1 is a schematic diagram of the workflow of a brightness estimation module according to an embodiment of the present invention. Figure 4-8 As shown, according to one embodiment of the present invention, in step S102, the pancreatic enhanced CT estimation model includes: a first training model and a second training model; The first training model includes: an image encoding module (Encoder1), a splicing module (Concatenate), and a first image decoding module (Decoder1); The plain scan CT image data is input into the image encoding module. The input of the splicing module is the output of the image encoding module and the brightness code of the plain scan CT image. The data obtained by the splicing module is input into the first image decoding module to finally obtain the enhanced CT image.

[0039] The brightness code of the plain scan CT image is defined as a reference point. In this embodiment, the reference point is "00000000". The second training model includes: an alignment module (Align), an image encoding module (Encoder1), a brightness estimation module (Encoder2), a splicing module (Concatenate), and a second image decoding module (Decoder2); The plain scan CT image data are input into the alignment module and the brightness estimation module respectively. The output of the alignment module is processed by the image encoding module. The output of the image encoding module and the output of the brightness estimation module are input into the splicing module together. The output of the splicing module is processed by the second image decoding module to obtain the corresponding enhanced CT image. The image encoding module and the splicing module in the second training model are obtained through training with the first training model and remain unchanged in the second training model; The initial parameters of the second image decoding module adopt the parameters obtained by the final training of the first image decoding module; the parameters of the second image decoding module are updated through the training of the second training model.

[0040] In this embodiment, the pancreatic enhanced CT estimation model is designed to predict and generate corresponding enhanced CT images from plain CT image data. This model is based on deep learning techniques, specifically a combination of convolutional neural networks (CNNs) and residual networks (ResNets), to extract, refine, and map image features. By training on a large amount of pancreatic CT image data, the model learns the mapping relationship between plain CT images and enhanced CT images, thereby enabling the acquisition of enhanced CT images without the need for actual enhanced scanning. This technology not only improves diagnostic efficiency but also significantly reduces the risk of radiation exposure for the patient. The pancreatic enhanced CT estimation model includes a first training model and a second training model. During the training process of the pancreatic enhanced CT estimation model, a large amount of pancreatic CT image data is collected, including both plain and enhanced CT images of the pancreas from the same patient, to construct a pancreatic CT image dataset. The subjects selected in the pancreatic CT image dataset are randomly sampled from the natural population. That is, the pancreatic CT image dataset includes both subjects with healthy pancreas and those with pancreatic cancer, with the proportion of these two types of subjects close to that of the natural population. The dataset is divided into a training set and a test set for model training and evaluation. The first and second training models are trained using the training set. After 400 rounds of training, the training set is evaluated on the test set to achieve the expected accuracy. Training is then stopped, resulting in a basic model that can be applied.

[0041] The first training model forms the foundational framework of the pancreatic enhanced CT estimation model and consists of an image encoding module, a splicing module, and a first image decoding module. The image encoding module takes plain CT image data as input. Through a series of downsampling and residual units, it gradually reduces the image size while extracting and refining effective image features. The downsampling unit reduces image resolution, reducing computational effort; the residual unit alleviates the vanishing gradient problem in deep neural networks, preserving feature integrity and detail. The splicing module takes the output of the image encoding module and the brightness code of the plain CT image. In this implementation, the brightness code of the plain CT image is defined as "00000000." The splicing module combines the two, fusing the image brightness information with the extracted features, providing richer information for subsequent image decoding and improving the accuracy of the reconstructed image. The first image decoding module takes the output of the splicing module as input. Through a series of upsampling and residual units, it gradually increases the feature size while refining and enhancing the features. Finally, a convolutional unit maps the features back to the image space to obtain a reconstructed enhanced CT image. The upsampling unit restores the image size, the residual unit further enhances the features, and the convolution unit converts the features into image pixels.

[0042] The second training model expands and optimizes on the first training model. It consists of an alignment module, an image encoding module (shared with the first training model), a brightness estimation module, a stitching module (similar to the first training model), and a second image decoding module. The alignment module takes unenhanced CT image data as input and uses Voxelmorph registration technology to align the images, ensuring spatial alignment between the input and the images used during model training, thereby improving model prediction accuracy. The brightness estimation module takes unenhanced CT image data as input and estimates image brightness through a series of convolution, activation, pooling, and fully connected operations, capturing image brightness information and providing stronger support for subsequent image stitching and decoding. The stitching module takes the outputs of the image encoding module and the brightness estimation module as input and stitches them together to fuse the image's feature and brightness information, providing a more comprehensive input for image decoding and improving the quality of the reconstructed image. The second image decoding module takes the output of the stitching module as input and uses a series of upsampling units and residual units to gradually increase the feature size, refine, and enhance the features, ultimately producing a reconstructed enhanced CT image. The initial parameters of the second image decoding module are the parameters obtained from the final training of the first image decoding module. The parameters are updated in the second training model to adapt to the new network structure and input data. The second training model is trained on the basis of the first training model, leveraging the feature extraction and mapping capabilities already learned by the first training model to accelerate the convergence of the model. By sharing the image encoding module and the splicing module, the training efficiency and stability of the model are improved. The parameters of the second image decoding module are updated to adapt to the new network structure and task requirements. The second training model combines multiple technologies such as image alignment, brightness estimation, and feature extraction to more accurately predict and generate enhanced CT images. At the same time, by sharing and updating network modules, the training efficiency and prediction accuracy of the model are improved.

[0043] The pancreatic enhanced CT estimation model of the present invention combines a first training model with a second training model, utilizing deep learning techniques (specifically, a combination of convolutional neural networks and residual networks) to successfully predict and generate corresponding enhanced CT images from plain CT image data. By training on a large amount of pancreatic CT image data, the model learns the mapping relationship between plain and enhanced CT images, significantly improving diagnostic efficiency while significantly reducing the risk of radiation exposure to patients. The second training model expands and optimizes the first training model, incorporating multiple techniques such as image alignment, brightness estimation, and feature extraction to further improve the accuracy of predicting and generating enhanced CT images. Furthermore, by sharing and updating network modules, the model's training efficiency and stability are enhanced.

[0044] According to an embodiment of the present invention, the brightness estimation module obtained through training in step S102 includes: an arterial phase brightness estimation unit, a portal venous phase brightness estimation unit, and a venous phase brightness estimation unit; Each unit is used to predict the brightness relationship between the plain scan CT image and the enhanced CT image of each phase.

[0045] In this embodiment, the brightness estimation module is meticulously divided into three independent units: an arterial phase brightness estimation unit, a portal venous phase brightness estimation unit, and a venous phase brightness estimation unit. Each unit has a specific task: predicting the brightness relationship between the plain CT image and the enhanced CT image of the corresponding phase (arterial phase, portal venous phase, and venous phase). During the training of the pancreatic enhanced CT estimation model, a separate training strategy is adopted for these three different brightness estimation units. By independently training each unit, the model can more accurately learn the complex brightness variation patterns between plain CT images and enhanced CT images of each phase, thereby improving the prediction accuracy and reliability of the entire pancreatic enhanced CT estimation model. Specifically, the arterial phase brightness estimation unit is trained using the brightness relationship between the plain CT image and the arterial phase CT image; the portal venous phase brightness estimation unit is trained using the brightness relationship between the plain CT image and the portal venous phase CT image; and the venous phase brightness estimation unit is trained using the brightness relationship between the plain CT image and the venous phase brightness estimation unit.

[0046] The present invention finely divides the brightness estimation module into three independent units: arterial phase, portal venous phase, and venous phase, and adopts a separate training strategy, so that the model can more accurately learn the brightness change pattern between plain CT and enhanced CT in each phase, thereby improving the prediction accuracy and reliability of the pancreatic enhanced CT estimation model and providing more accurate and personalized auxiliary information for clinical practice.

[0047] According to one embodiment of the present invention, in step S112, the threshold is set in the following manner: Based on the data set used to train the pancreatic enhanced CT estimation model, the relative blood flow values ​​of all subjects were obtained; All relative blood flow values ​​are fitted with Poisson distribution to obtain the standard deviation of the relative blood flow Poisson distribution , threshold = 1+ .

[0048] In this embodiment, the threshold setting method is clearly and reasonably defined: first, based on the data set used to train the pancreatic enhanced CT estimation model, the relative blood flow values ​​of all subjects are extracted. In the process of training the pancreatic enhanced CT estimation model, a pancreatic CT image data set is pre-constructed, in which a large number of pancreatic image data of subjects are collected, including pancreatic plain CT images and pancreatic enhanced CT image sets of the same subject. In order to make the final result closer to the actual situation of the natural population, in the process of constructing the pancreatic CT image data set, the subjects selected are randomly sampled from the natural population, that is, the subjects in the pancreatic CT image data set include both subjects with healthy pancreas and subjects with pancreatic cancer, and the ratio of the number of subjects of the two types is close to that of the natural population. As in the method provided in step S108, when calculating the relative blood flow value for each subject, the brightness code of the subject's plain CT image is defined as "00000000". A first method can predict and obtain the brightness codes of the subject's arterial phase CT image, portal venous phase CT image, and venous phase CT image using the method described in step S106. A second method can obtain the brightness codes of the subject's arterial phase CT image, portal venous phase CT image, and venous phase CT image by processing the subject's pancreatic enhanced CT image set. To ensure that the relative blood flow value is closer to real data from a natural population, we use the second method.

[0049] The tetrahedral structure volumes formed in high-dimensional space by CT image brightness encoding at different phases of the pancreatic CT image dataset for subjects judged to have healthy pancreas were averaged to obtain the benchmark tetrahedral structure volume. The tetrahedral structure volumes formed in high-dimensional space for all subjects were compared with the benchmark tetrahedral structure volume to obtain the relative blood flow value of each subject.

[0050] Next, we fit these relative blood flow values ​​using the Poisson distribution and calculate the standard deviation σ of the Poisson distribution of relative blood flow. On this basis, the threshold is set to 1 plus this standard deviation, that is, threshold = 1 + σ. The relative blood flow estimates exhibit the characteristics of the Poisson distribution. By using the calculated relative blood flow data, we can fit the parameter lambda of the Poisson distribution and then calculate the standard deviation When the relative blood flow estimation value of a subject deviates from the normal range and reaches or exceeds this threshold, it is considered to be suspected of blood flow abnormality.

[0051] The present invention can more scientifically define the standard of abnormal blood flow through the fitting of Poisson distribution, thereby improving the accuracy and reliability of the model in detecting abnormal blood flow.

[0052] According to an embodiment of the present invention, in step S112, classifying the subject based on the judgment result includes: When the target relative blood flow does not exceed the threshold, the subject is classified as a low-risk group for pancreatic cancer; When the target relative blood flow exceeds a threshold, the subject is classified as a high-risk group for pancreatic cancer, and further pancreatic examination is performed on the high-risk group for pancreatic cancer.

[0053] In this implementation, the subject's target relative blood flow is compared with a preset threshold to achieve accurate classification. Specifically, if the target relative blood flow does not exceed the threshold, the subject is classified as low-risk for pancreatic cancer and no further in-depth examination is required. If the target relative blood flow exceeds the threshold, the subject is classified as high-risk for pancreatic cancer and is recommended to undergo a more detailed pancreatic examination.

[0054] The present invention is based on the quantitative assessment of relative blood flow, which not only effectively utilizes medical resources and avoids unnecessary examinations, but also provides doctors with a scientific diagnostic reference, helps to detect pancreatic cancer risks at an early stage, and improves the accuracy and efficiency of diagnosis.

[0055] Example 2 like Figure 2-8 As shown, according to one embodiment of the present invention, a relative blood flow estimation method based on pancreatic plain scan CT images includes the following steps: Step S202: constructing a pancreatic enhanced CT estimation model, and obtaining a brightness estimation module based on pre-training of the pancreatic enhanced CT estimation model; The pancreatic enhanced CT estimation model includes: a first training model and a second training model; The first training model includes: an image encoding module, a splicing module and a first image decoding module; The plain scan CT image data is input into the image encoding module. The input of the splicing module is the output of the image encoding module and the brightness code of the plain scan CT image. The data obtained by the splicing module is input into the first image decoding module to finally obtain the enhanced CT image.

[0056] Among them, the brightness code of the plain scan CT image is "00000000"; The second training model includes: an alignment module, an image encoding module, a brightness estimation module, a splicing module and a second image decoding module; The plain scan CT image data are input into the alignment module and the brightness estimation module respectively. The output of the alignment module is processed by the image encoding module. The output of the image encoding module and the output of the brightness estimation module are input into the splicing module together. The output of the splicing module is processed by the second image decoding module to obtain the corresponding enhanced CT image. The image encoding module and the splicing module in the second training model are obtained through training with the first training model and remain unchanged in the second training model; The initial parameters of the second image decoding module adopt the parameters obtained by the final training of the first image decoding module; the parameters of the second image decoding module are updated through the training of the second training model.

[0057] During the training of the pancreatic enhanced CT estimation model, a large amount of pancreatic CT image data was collected, including pancreatic plain CT images and pancreatic enhanced CT images of the same subject, to construct a pancreatic CT image dataset. In order to make the final result closer to the actual situation of the natural population, the subjects selected in the construction of the pancreatic CT image dataset were randomly sampled from the natural population. That is, the subjects in the pancreatic CT image dataset included both subjects with healthy pancreas and subjects with pancreatic cancer, and the proportion of the two types of subjects was close to that of the natural population. The dataset was divided into a training set and a test set for model training and evaluation. The first training model and the second training model were trained using the training set. After 400 rounds of training, the test set evaluated that the training set could achieve the expected accuracy, and the training was stopped to obtain a basic model that can be applied.

[0058] During training, t1, t2, and t3 are selected in a time sequence (fixed time intervals). A decoding module is required to decode the features and calculate the mean square error (MSE) loss between the predicted and true highlight images. However, the decoding module is not used in training the encoding module; it is only used to train the encoding module. Specifically, the decoding module reads the concatenation of the two sets of features and decodes them to restore the enhanced image. In actual use, new enhanced CT images are collected and fed into the model to obtain pancreatic cancer location information. This model can be deployed on standard computing devices.

[0059] Step S204, obtaining a plain CT image of the subject's pancreas, that is, a target plain CT image; Step S206, based on the target plain scan CT image and the brightness estimation module, obtain the predicted: arterial phase CT image brightness coding, portal venous phase CT image brightness coding, and venous phase CT image brightness coding; The brightness estimation module includes: an arterial phase brightness estimation unit, a portal venous phase brightness estimation unit and a venous phase brightness estimation unit; Each unit is used to predict the brightness relationship between the plain scan CT image and the enhanced CT image of each phase.

[0060] Based on the target plain scan CT image and the arterial phase brightness estimation unit, the predicted arterial phase CT image brightness code is obtained. Based on the target plain scan CT image and the portal venous phase brightness estimation unit, the predicted portal venous phase CT image brightness code is obtained. Based on the target plain scan CT image and the venous phase brightness estimation unit, the predicted venous phase CT image brightness code is obtained.

[0061] Step S208, defining the brightness code of the target plain scan CT image as "00000000", and constructing the target tetrahedron structure in a high-dimensional space based on the brightness codes of the CT images at different phases; For unenhanced CT (i.e., plain scan CT), the brightness code is "00000000." The three enhanced images (t1 arterial phase, t2 portal venous phase, and t3 venous phase) are brightness-encoded by the brightness estimation module. After the brightness code is obtained, it is formed into a tetrahedron in high-dimensional space. This tetrahedron describes the blood flow and is called a brightness tetrahedron.

[0062] Step S210, obtaining a target relative blood flow based on a ratio of the target tetrahedral structure volume to the reference tetrahedral structure volume; wherein the reference tetrahedral structure volume is an average value of tetrahedral structure volumes formed by brightness encoding of CT images of a healthy pancreas at different phases in a high-dimensional space; For subjects with negative pancreatic cancer, their brightness tetrahedron conforms to the universal law, and the corresponding tetrahedron is the negative general tetrahedron V - , the average of the tetrahedral structure volumes corresponding to the pancreatic cancer negative subjects (i.e., healthy pancreas) is calculated (the data used in this process is based on the data of the pancreatic cancer negative subjects in the pancreatic CT image dataset), and the reference tetrahedral structure volume S is obtained. - For patients with positive pancreatic cancer, their brightness tetrahedron will be distorted, corresponding to their own positive tetrahedron V + , its volume will also change. Therefore, when obtaining the target tetrahedral structure volume S of the subject + , the target relative blood flow is obtained by dividing the target tetrahedral structure volume by the reference tetrahedral structure volume .

[0063] Whether the subject is suspected of being a pancreatic cancer-positive subject is determined by the target relative blood flow.

[0064] Step S212: determine whether the target relative blood flow exceeds a threshold, and classify the subject based on the determination result.

[0065] The threshold is set in the following way: Based on a data set used to train a pancreatic enhanced CT estimation model, relative blood flow values ​​of all subjects are obtained; specifically, the method comprises: averaging the tetrahedral structure volumes formed in a high-dimensional space by brightness encoding of CT images of subjects judged to have healthy pancreas at different phases in the pancreatic CT image data set to obtain a reference tetrahedral structure volume; and comparing the tetrahedral structure volumes formed in the high-dimensional space of all subjects with the reference tetrahedral structure volume to obtain the relative blood flow value of each subject.

[0066] All relative blood flow values ​​are fitted with Poisson distribution to obtain the standard deviation of the relative blood flow Poisson distribution , threshold = 1+ .

[0067] The examinees are classified based on the judgment results, including: When the target relative blood flow does not exceed the threshold, the subject is classified as a low-risk group for pancreatic cancer; When the target relative blood flow exceeds a threshold, the subject is classified as a high-risk group for pancreatic cancer, and further pancreatic examination is performed on the high-risk group for pancreatic cancer.

[0068] The present invention is not limited to simply making a positive or negative judgment on pancreatic CT scan images. It greatly improves the accuracy of detection by estimating relative blood flow, allowing doctors to understand the patient's pancreatic condition more clearly and providing strong support for subsequent treatment.

[0069] The present invention predicts enhanced CT images from plain scan CT images and evaluates the patient's relative blood flow accordingly. It provides doctors with richer diagnostic information, helps doctors understand the patient's pancreatic blood flow status more comprehensively, and thus make more accurate diagnoses.

[0070] The present invention also demonstrates significant advantages in enhancing pancreatic angiography. Compared with existing technologies, it can provide more complete and clearer pancreatic angiography images, allowing doctors to observe the detailed structure of the pancreas more clearly, thereby improving the accuracy of diagnosis.

[0071] This method does not impose additional burdens on patients or medical resources. It utilizes existing CT scan images for analysis and prediction, saving both time and costs. Furthermore, this solution can assist doctors in quickly and directly predicting pancreatic cancer using pancreatic CT scan images, making subsequent enhanced contrast imaging more targeted and further improving diagnostic efficiency.

[0072] The judgment model used in the present invention is extremely flexible. It can not only run on high-performance computer platforms, but also be adapted to ordinary computers or mobile phone platforms. This enables doctors to use the model for diagnosis anytime and anywhere, greatly improving the availability and popularity of the model.

[0073] Example 3 Figure 9 FIG1 is a structural diagram of a relative blood flow estimation system based on pancreatic plain scan CT images according to an embodiment of the present invention. Figure 9 As shown, a relative blood flow estimation system based on pancreatic plain scan CT images includes: A construction module 10 is used to construct a pancreatic enhanced CT estimation model and obtain a brightness estimation module based on the pre-training of the pancreatic enhanced CT estimation model; An acquisition module 20 is configured to acquire a plain CT image of the pancreas of the subject, that is, a target plain CT image; A prediction module 30 is configured to obtain, based on the target plain scan CT image and the brightness estimation module, predicted brightness coding of the arterial phase CT image, brightness coding of the portal venous phase CT image, and brightness coding of the venous phase CT image; A calculation module 40 is configured to define the target plain scan CT image brightness code as a reference point "00000000" and construct a target tetrahedron structure in a high-dimensional space based on the CT image brightness codes at different phases; The calculation module 40 is further configured to obtain a target relative blood flow based on a ratio of the target tetrahedral structure volume to the reference tetrahedral structure volume; wherein the reference tetrahedral structure volume is an average of tetrahedral structure volumes formed by brightness encoding of CT images of a healthy pancreas at different phases in a high-dimensional space; The judgment module 50 is used to judge whether the target relative blood flow exceeds a threshold value and classify the subject based on the judgment result.

[0074] According to one embodiment of the present invention, in the construction module 10, the pancreatic enhanced CT estimation model includes: a first training model and a second training model; The first training model includes: an image encoding module, a splicing module, and a first image decoding module; the plain scan CT image data is input into the image encoding module, the splicing module inputs the output of the image encoding module and the brightness code of the plain scan CT image, and the data obtained by the splicing module is input into the first image decoding module to finally obtain an enhanced CT image; Among them, the brightness code of the plain scan CT image is the reference point "00000000"; The second training model includes: an alignment module, an image encoding module, a brightness estimation module, a stitching module, and a second image decoding module; the plain scan CT image data is input into the alignment module and the brightness estimation module respectively, the output of the alignment module is processed by the image encoding module, the output of the image encoding module and the output of the brightness estimation module are input into the stitching module together, and the output of the stitching module is processed by the second image decoding module to obtain the corresponding enhanced CT image; Among them, the image encoding module and splicing module in the second training model are obtained through training with the first training model and remain unchanged in the second training model; the initial parameters of the second image decoding module adopt the parameters obtained by the final training of the first image decoding module; and the parameters of the second image decoding module are updated through training with the second training model.

[0075] According to one embodiment of the present invention, in the construction module 10, the brightness estimation module includes: an arterial phase brightness estimation unit, a portal venous phase brightness estimation unit, and a venous phase brightness estimation unit; Each unit is used to predict the brightness relationship between the plain scan CT image and the enhanced CT image of each phase.

[0076] According to one embodiment of the present invention, in the judgment module 50, the threshold is set in the following manner: Based on the data set used to train the pancreatic enhanced CT estimation model, the relative blood flow values ​​of all subjects were obtained; The relative blood flow values ​​of all subjects were fitted with Poisson distribution to obtain the standard deviation of the relative blood flow Poisson distribution. , threshold = 1+ .

[0077] According to one embodiment of the present invention, in the judgment module 50, classifying the subject based on the judgment result includes: When the target relative blood flow does not exceed the threshold, the subject is classified as a low-risk group for pancreatic cancer; When the target relative blood flow exceeds a threshold, the subject is classified as a high-risk group for pancreatic cancer, and further pancreatic examination is performed on the high-risk group for pancreatic cancer.

[0078] The present invention is based on the quantitative assessment of relative blood flow, which not only effectively utilizes medical resources and avoids unnecessary examinations, but also provides doctors with a scientific diagnostic reference, helps to detect pancreatic cancer risks at an early stage, and improves the accuracy and efficiency of diagnosis.

[0079] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0080] The above description is merely a preferred embodiment of the present application and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention herein is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also encompasses other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the inventive concept. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features having similar functions disclosed in this application.

[0081] It should be understood that the size of the serial numbers of each step in the content of the invention and the embodiments of the present invention does not absolutely mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

Claims

1. A relative blood flow estimation method based on pancreatic plain scan CT images, characterized in that: The following steps are involved: Constructing a pancreatic enhanced CT estimation model, and pre-training a brightness estimation module based on the pancreatic enhanced CT estimation model; Acquiring a plain CT image of the pancreas of the subject, that is, a target plain CT image; Acquire predicted brightness codes of the arterial phase CT image, the portal venous phase CT image, and the venous phase CT image based on the target plain scan CT image and the brightness estimation module; The target plain scan CT image brightness code is defined as the reference point, and based on the CT image brightness code at different phases, the target tetrahedron structure is constructed in high-dimensional space. A target relative blood flow is obtained based on the ratio of the target tetrahedral structure volume to the reference tetrahedral structure volume; wherein the reference tetrahedral structure volume is an average value of tetrahedral structure volumes formed in a high-dimensional space by brightness encoding of CT images of a healthy pancreas at different phases; It is determined whether the target relative blood flow exceeds a threshold, and the subject is classified based on the determination result.

2. The relative blood flow estimation method based on pancreatic plain CT images according to claim 1, characterized in that: The pancreatic enhanced CT estimation model includes: a first training model and a second training model; The first training model includes: an image encoding module, a splicing module, and a first image decoding module; the plain scan CT image data is input into the image encoding module, the input of the splicing module is the output of the image encoding module and the brightness code of the plain scan CT image, and the data obtained by processing the splicing module is input into the first image decoding module to finally obtain an enhanced CT image; wherein the brightness code of the plain scan CT image is used as a reference point; The second training model includes: an alignment module, an image encoding module, a brightness estimation module, a splicing module, and a second image decoding module; plain scan CT image data are respectively input into the alignment module and the brightness estimation module, the output of the alignment module is processed by the image encoding module, the output of the image encoding module and the output of the brightness estimation module are jointly input into the splicing module, and the output of the splicing module is processed by the second image decoding module to obtain a corresponding enhanced CT image; Among them, the image encoding module and the splicing module in the second training model are obtained through training with the first training model and remain unchanged in the second training model; the initial parameters of the second image decoding module adopt the parameters finally obtained through training with the first image decoding module; and the parameters of the second image decoding module are updated through training with the second training model.

3. The relative blood flow estimation method based on pancreatic plain CT images according to claim 1, characterized in that: The brightness estimation module includes: an arterial phase brightness estimation unit, a portal vein phase brightness estimation unit and a venous phase brightness estimation unit; Each unit is used to predict the brightness relationship between the plain scan CT image and the enhanced CT image of each phase.

4. The relative blood flow estimation method based on pancreatic plain CT images according to claim 1, characterized in that: The threshold is set as follows: Obtaining relative blood flow values ​​of all subjects based on the data set used to train the pancreatic enhanced CT estimation model; The relative blood flow values ​​of all subjects were fitted with Poisson distribution to obtain the standard deviation of the relative blood flow Poisson distribution. , threshold = 1+ .

5. The relative blood flow estimation method based on pancreatic plain CT images according to claim 1, characterized in that: The examinees are classified based on the judgment results, including: When the target relative blood flow does not exceed the threshold, classifying the subject as a low-risk group for pancreatic cancer; When the target relative blood flow exceeds the threshold, the subject is classified as a high-risk group for pancreatic cancer, and further pancreatic examination is performed on the high-risk group for pancreatic cancer.

6. A relative blood flow estimation system based on pancreatic plain scan CT images, characterized in that: include: A construction module is used to construct a pancreatic enhanced CT estimation model and obtain a brightness estimation module based on pre-training of the pancreatic enhanced CT estimation model; an acquisition module, configured to acquire a plain CT image of the pancreas of the subject, that is, a target plain CT image; A prediction module, configured to obtain, based on the target plain scan CT image and the brightness estimation module, predicted: brightness coding of the arterial phase CT image, brightness coding of the portal venous phase CT image, and brightness coding of the venous phase CT image; A calculation module is used to define the brightness code of the target plain scan CT image as a reference point, and to construct the target tetrahedron structure in a high-dimensional space based on the brightness code of the CT image at different phases; The calculation module is further configured to obtain a target relative blood flow based on a ratio of the target tetrahedral structure volume to a reference tetrahedral structure volume; wherein the reference tetrahedral structure volume is an average value of tetrahedral structure volumes formed by brightness encoding of CT images of a healthy pancreas at different phases in a high-dimensional space; The judgment module is used to judge whether the target relative blood flow exceeds a threshold value and classify the subject based on the judgment result.

7. The relative blood flow estimation system based on pancreatic plain CT images according to claim 6, characterized in that: In the construction module, the pancreatic enhanced CT estimation model includes: a first training model and a second training model; The first training model includes: an image encoding module, a splicing module, and a first image decoding module; the plain scan CT image data is input into the image encoding module, the input of the splicing module is the output of the image encoding module and the brightness code of the plain scan CT image, and the data obtained by processing the splicing module is input into the first image decoding module to finally obtain an enhanced CT image; wherein the brightness code of the plain scan CT image is used as a reference point; The second training model includes: an alignment module, an image encoding module, a brightness estimation module, a splicing module, and a second image decoding module; plain scan CT image data are respectively input into the alignment module and the brightness estimation module, the output of the alignment module is processed by the image encoding module, the output of the image encoding module and the output of the brightness estimation module are jointly input into the splicing module, and the output of the splicing module is processed by the second image decoding module to obtain a corresponding enhanced CT image; Among them, the image encoding module and the splicing module in the second training model are obtained through training with the first training model and remain unchanged in the second training model; the initial parameters of the second image decoding module adopt the parameters finally obtained through training with the first image decoding module; and the parameters of the second image decoding module are updated through training with the second training model.

8. The relative blood flow estimation system based on pancreatic plain CT images according to claim 6, characterized in that: In the construction module, the brightness estimation module includes: an arterial phase brightness estimation unit, a portal venous phase brightness estimation unit and a venous phase brightness estimation unit; Each unit is used to predict the brightness relationship between the plain scan CT image and the enhanced CT image of each phase.

9. The relative blood flow estimation system based on pancreatic plain CT images according to claim 6, characterized in that: In the judgment module, the threshold is set in the following manner: Obtaining relative blood flow values ​​of all subjects based on the data set used to train the pancreatic enhanced CT estimation model; The relative blood flow values ​​of all subjects were fitted with Poisson distribution to obtain the standard deviation of the relative blood flow Poisson distribution. , threshold = 1+ .

10. The relative blood flow estimation system based on pancreatic plain CT images according to claim 6, characterized in that: In the judgment module, the subject is classified based on the judgment result, including: When the target relative blood flow does not exceed the threshold, classifying the subject as a low-risk group for pancreatic cancer; When the target relative blood flow exceeds the threshold, the subject is classified as a high-risk group for pancreatic cancer, and further pancreatic examination is performed on the high-risk group for pancreatic cancer.

Images