A liver CT image continuity discontinuity dynamic imaging information fusion method

Abstract

The application discloses a liver CT image continuity discontinuous dynamic imaging information fusion method, which combines a machine learning algorithm, filters and feature segments images by extracting and identifying feature points, obtains a dynamic change curve graph and a standard floating line, constructs a maximum loss difference value and a minimum loss difference value of the dynamic change curve graph, establishes a loss difference value allowable range, obtains minimum values and maximum values in a dynamic change period, judges threshold difference based on the maximum values, realizes liver CT image fusion image probability analysis, and solves the problems that in the prior liver change diagnosis and treatment process, doctors have heavy workloads, evaluation efficiency is relatively low, and analyzed lesion information is not comprehensive enough, so that key feature points are easily ignored, and diagnosis and treatment difficulty is increased.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to a method for fusing continuous and discontinuous dynamic imaging information from liver CT images. Background Technology

[0002] With the continuous development and integration of medical imaging and computer science, analyzing clinical medical data using computer technology has become a highly accurate and efficient method of auxiliary diagnosis, significantly improving disease prevention and treatment success rates. Early diagnosis of liver cancer caused by liver deterioration is difficult; in most cases, by the time a patient is diagnosed, the curable characteristics have already disappeared. Therefore, using auxiliary diagnostic methods based on the characteristics of liver deterioration to assess the progress and direction of deterioration is of significant clinical value in preventing liver cancer.

[0003] In clinical applications, medical imaging techniques such as computed tomography (CT) and magnetic resonance imaging (MR) can generate data representing information about internal organs, playing a crucial role in disease diagnosis. Taking liver CT as an example, during treatment efficacy evaluation, doctors can only judge the treatment effect based on the visual appearance of tissues in dynamically enhanced images. On the one hand, this evaluation process is relatively crude, resulting in a heavy workload for doctors and low efficiency. On the other hand, the information-assisted diagnosis process does not accurately perform three-dimensional image fusion, leading to incomplete lesion information and easily overlooking key feature points, thus increasing the difficulty of diagnosis and treatment. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] In view of the problems existing in the above-mentioned auxiliary diagnosis and treatment methods for liver lesions, this invention is proposed.

[0006] Therefore, the technical problem solved by this invention is that in the existing diagnosis and treatment of liver lesions, doctors can only judge the treatment effect based on the intuitive situation of the tissue in various dynamically enhanced images. On the one hand, this assessment process is relatively crude, the workload of doctors is large, and the assessment efficiency is low. On the other hand, the information-assisted diagnosis and treatment process does not perform accurate three-dimensional fusion of images, the analyzed lesion information is not comprehensive enough, and it is easy to overlook key feature points, thereby increasing the difficulty of diagnosis and treatment.

[0007] To address the aforementioned technical problems, this invention provides the following technical solution: a method for fusing continuous and discontinuous dynamic imaging information from liver CT images, comprising: acquiring liver CT images within a preset period; extracting identification feature points from all the liver CT images and performing feature segmentation on the liver CT images based on the extracted identification feature points; comparing the segmented liver CT images to extract the liver CT images exhibiting dynamic changes, and arranging the extracted liver CT images in a time-aligned manner; obtaining a dynamic change curve of the arranged liver CT images; obtaining a standard floating line of the dynamic change curve; constructing the maximum and minimum loss differences of the dynamic change curve based on the standard floating line, and establishing an allowable range for the loss difference; obtaining the minimum and maximum values ​​within the dynamic change period, and defining a liver change warning line when the maximum value reaches a threshold, thereby achieving probabilistic analysis of the fused images of the liver CT images.

[0008] As a preferred embodiment of the continuous discontinuous dynamic imaging information fusion method for liver CT images described in this invention, the acquisition of liver CT images within a preset period specifically includes: CT imaging; acquiring the liver CT image of the current period; analyzing the liver CT image of the current period to determine the number of image acquisition periods; acquiring all liver CT images within the preset period and performing inter-slice alignment processing; introducing liver image recognition features, constructing a recognition objective function, and extracting the recognition feature regions of all liver CT images.

[0009] As a preferred embodiment of the continuous-discontinuous dynamic imaging information fusion method for liver CT images described in this invention, the method for analyzing the current phase of the liver CT image and determining the number of phases for image acquisition specifically includes: determining the corresponding properties of the lesion, including the lesion location, size, and severity; segmenting the current phase of the liver CT image based on the corresponding properties of the lesion; acquiring a CT image of a normal liver, and using this CT image as a basis, obtaining the difference between the current phase of the liver CT image and the normal liver CT image in terms of the corresponding properties of the lesion; and determining the number of phases for acquisition as 5 when the difference reaches a threshold of 25%, as 7 when the difference reaches a threshold of 35%, and as 10 when the difference is greater than a threshold of 35%.

[0010] As a preferred embodiment of the continuous-discontinuous dynamic imaging information fusion method for liver CT images described in this invention, the liver image recognition features are introduced, and then the liver image recognition features are amplified to minimize the loss value output by the recognition target function.

[0011] As a preferred embodiment of the continuous discontinuous dynamic imaging information fusion method for liver CT images described in this invention, the extraction of identification feature points from all the liver CT images specifically includes: constructing an identification model using the XGBoost model algorithm; inputting all identification feature regions of each of the liver CT images respectively, using the liver image identification features as the identification benchmark, and initially identifying the identification feature points of all the liver CT images; inputting each identification feature point into the identification model, performing self-learning and self-optimization to obtain the optimal parameters of the identification model, and each optimal parameter is the identification feature point of each identification feature region.

[0012] As a preferred embodiment of the continuous discontinuous dynamic imaging information fusion method for liver CT images described in this invention, the method for feature segmentation of the liver CT image based on the extracted identification feature points specifically includes: constructing a dimensional information fusion planar model; identifying the identification feature region and identification feature points of the normal liver CT image according to the above steps; using the identification feature points as the feature segmentation center points, the corresponding identification feature region as the first range region, and the identification feature region of the normal liver CT image as the second range region, fusing and connecting the identification feature points of the identification feature points with those of the normal liver CT image to perform feature segmentation on the first range region and the second range region; obtaining the feature identification region after feature segmentation, and obtaining the feature identification points of the feature identification region according to the above steps.

[0013] As a preferred embodiment of the continuous discontinuous dynamic imaging information fusion method for liver CT images described in this invention, the extraction of dynamically changing liver CT images by comparing the segmented liver CT images specifically includes: constructing a coordinate information fusion plane model; inputting the feature recognition points after feature segmentation of each CT image into the coordinate information fusion plane model; connecting the input feature recognition points one by one according to the acquisition time sequence; obtaining the slope value of each segment, and defining that when the slope value does not belong to (-1,1), the previous liver CT image has dynamic changes compared to the previous liver CT image, and extracting it.

[0014] As a preferred embodiment of the continuous discontinuous dynamic imaging information fusion method for liver CT images described in this invention, wherein: in the coordinate information fusion plane model, the extracted feature recognition points are connected by a smooth curve to obtain a dynamic change curve of the arranged liver CT images; the input feature recognition points are averaged in the coordinate information fusion plane model to obtain a standard floating line of the dynamic change curve.

[0015] As a preferred embodiment of the continuous-discontinuous dynamic imaging information fusion method for liver CT images described in this invention, the maximum and minimum loss differences of the dynamic change curves are obtained using the following formulas.

[0016] Y max ′=Y max(动态变化曲线图) -Y (标准浮动线) ;

[0017] Y min ′=|Y min(动态变化曲线图) -Y (标准浮线) |;

[0018] H = (Y min ′, Y max ′);

[0019] Among them, Y max ' represents the maximum loss difference; Y min ' represents the minimum loss difference; H represents the allowable range of the loss difference; Y represents the longitudinal position coordinate value of the coordinate information fusion plane model.

[0020] As a preferred embodiment of the continuous-discontinuous dynamic imaging information fusion method for liver CT images described in this invention, the minimum and maximum values ​​during the dynamic change period are obtained using the following formulas.

[0021]

[0022]

[0023] Among them, y min The minimum value during the dynamic change period; y max Y represents the maximum value during the dynamic change period. max ' represents the maximum loss difference; Y min ' represents the minimum loss difference; Y represents the longitudinal positional coordinate value of the coordinate information fusion planar model;

[0024] The calculated minimum and maximum values ​​within the dynamic change period are then used for training. The minimum and maximum values ​​within the dynamic change period are defined as meeting the allowable range of the loss difference to be considered as qualified for training. Otherwise, the selection and calculation of the minimum and maximum values ​​within the dynamic change period are repeated.

[0025] The beneficial effects of this invention are as follows: This invention provides a method for fusing continuous and discontinuous dynamic imaging information from liver CT images. Combining machine learning algorithms, it extracts and identifies feature points to filter and segment images, obtaining dynamic change curves and standard floating lines. It constructs the maximum and minimum loss differences of the dynamic change curves, establishes an allowable range for the loss difference, and obtains the minimum and maximum values ​​within the dynamic change period. Based on the maximum value, it judges the threshold difference, achieving probabilistic analysis of the fused images of liver CT images. This solves the problem that in existing liver disease diagnosis and treatment processes, doctors can only judge the treatment effect based on the intuitive situation of tissue reaction in various dynamically enhanced images. On the one hand, this evaluation process is relatively crude, with a heavy workload for doctors and low evaluation efficiency; on the other hand, the information-assisted diagnosis and treatment process does not perform accurate three-dimensional fusion of images, resulting in insufficiently comprehensive lesion information and easy neglect of key feature points, thus increasing the difficulty of diagnosis and treatment. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0027] Figure 1 The flowchart of the method for fusing continuous and discontinuous dynamic imaging information of liver CT images provided by the present invention is shown.

[0028] Figure 2 This invention relates to a method for acquiring liver CT images during a preset period.

[0029] Figure 3 This is a flowchart of the method for determining the number of image acquisition periods involved in this invention.

[0030] Figure 4 This invention relates to a method for extracting identification feature points from all liver CT images.

[0031] Figure 5 This invention relates to a method for feature segmentation of liver CT images based on extracted identification feature points.

[0032] Figure 6 This invention relates to a method for extracting dynamically changing liver CT images.

[0033] Figure 7 The present invention provides current liver CT images.

[0034] Figure 8The magnified current liver CT image provided by this invention.

[0035] Figure 9 This is the tree-like generation diagram of the XGboost model involved in this invention.

[0036] Figure 10 This invention relates to a state curve diagram. Detailed Implementation

[0037] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0038] Taking liver CT as an example, during the efficacy evaluation process, doctors can only judge the treatment effect based on the intuitive situation of the tissue's reaction in various dynamically enhanced images. On the one hand, this evaluation process is relatively crude, the workload of doctors is large, and the evaluation efficiency is low. On the other hand, the information-assisted diagnosis and treatment process does not accurately perform three-dimensional fusion of images, and the lesion information is not comprehensive enough, making it easy to overlook key feature points, thus increasing the difficulty of diagnosis and treatment.

[0039] Therefore, please refer to Figure 1 This invention provides a method for fusing continuous and discontinuous dynamic imaging information from liver CT images, comprising:

[0040] S1: Acquire liver CT images within a preset period;

[0041] For further details, please refer to Figure 2 The acquisition of liver CT images within a preset period specifically includes:

[0042] S1.1: CT imaging, acquiring current liver CT images;

[0043] S1.2: Analyze the current liver CT images to determine the number of image acquisition periods;

[0044] Please see Figure 3 Specifically, it includes the following steps:

[0045] (1) Determine the corresponding nature of the lesion, including the location, size, and severity of the lesion;

[0046] (2) Segment the current stage liver CT images according to the corresponding nature of the lesions;

[0047] CT image segmentation technology is an existing technology and will not be elaborated upon here.

[0048] (3) Collect CT images of normal liver, and based on these CT images, obtain the differences between the current liver CT images and normal liver CT images in terms of the corresponding nature of the lesions. When the difference reaches the threshold of 25%, the number of collection periods is determined to be 5; when the difference reaches the threshold of 35%, the number of collection periods is determined to be 7; and when the difference is greater than the threshold of 35%, the number of collection periods is determined to be 10.

[0049] It should be noted that determining the appropriate number of collection periods based on pathological differences can reflect the characteristics of liver changes with the minimum and most suitable number of collection periods. Too few collection periods will result in unclear characteristics, while too many collection periods will lead to a waste of resources.

[0050] S1.3: Acquire all liver CT images within the preset period and perform inter-slice alignment processing;

[0051] S1.4: Introduce liver image recognition features, construct a recognition objective function, and extract the recognition feature regions of all liver CT images.

[0052] Furthermore, after introducing liver image recognition features, the liver image recognition features are amplified to minimize the loss value output by the recognition objective function.

[0053] The target function is constructed through pixel grayscale difference analysis, and the constructed function is as follows:

[0054] Loss value = [{"pixel grayscale difference" - feature} / {pixel grayscale of liver tumor - pixel grayscale of normal liver}] ∈ (0, 1).

[0055] like Figure 7 and Figure 8 As shown, the magnified image features are more obvious, and the pixel grayscale differences are smaller.

[0056] S2: Extract the identification feature points from all liver CT images and perform feature segmentation on the liver CT images based on the extracted identification feature points;

[0057] For further details, please refer to Figure 4 Extracting the identification feature points from all liver CT images specifically includes:

[0058] (1) Construct a recognition model using the XGBoost model algorithm;

[0059] (2) Input all the recognition feature regions of each liver CT image respectively, and use the liver image recognition features as the recognition benchmark to initially identify the recognition feature points of all liver CT images;

[0060] (3) Input the respective recognition feature points into the recognition model, perform self-learning and self-optimization, and obtain the optimal parameters of the recognition model. Each optimal parameter is the recognition feature point of each recognition feature region.

[0061] The feature point extraction and recognition process utilizes the XGBoost model algorithm from machine learning. This algorithm establishes a tree-like model and, relying on input sample data, performs self-learning and self-optimization to obtain the optimal model parameters. (Please refer to [link / reference]). Figure 9 The training of the model includes the following process:

[0062] (1) Input the recognition feature regions of each liver CT image into the XGboost model;

[0063] (2) The XGboost model performs the first round of learning: finding the first split node. Find one feature among all features as the split node to split and calculate the loss value. Find another feature as the split node to split and calculate the loss value at the same time, until a minimum loss value is obtained. Split according to the feature corresponding to the minimum loss value to obtain the first tree model;

[0064] (3) The XGboost model performs a second round of learning: finding the second split node. From the remaining features, the feature with the smallest loss value is selected as the next-level split node, and a new tree-like model is obtained by splitting.

[0065] (4) Following the above method, after several rounds of model learning, the tree model gradually grows stronger and new tree models are continuously formed on the basis of further splitting in the optimization, while obtaining the loss value and node score.

[0066] (5) The model stops splitting when the error in the objective function of the obtained XGboost model is minimized and the node score reaches the optimal value, and the optimal parameters of the model are obtained.

[0067] For further details, please refer to Figure 5 The feature segmentation of liver CT images based on extracted recognition feature points specifically includes:

[0068] (1) Construct a planar model that integrates dimensional information;

[0069] (2) Based on the above steps, identify the identification feature areas and identification feature points of normal liver CT images;

[0070] (3) Using the identification feature point as the feature segmentation center point, the corresponding identification feature area is the first range area, and the identification feature area of ​​the normal liver CT image is the second range area. The identification feature point and the identification feature point of the normal liver CT image are fused and connected to perform feature segmentation on the first range area and the second range area.

[0071] (4) Obtain the feature recognition region after feature segmentation, and obtain the feature recognition points of the feature recognition region according to the above steps.

[0072] S3: Compare the segmented liver CT images, extract the liver CT images with dynamic changes, and arrange the extracted liver CT images in time alignment.

[0073] For further details, please refer to Figure 6 By comparing the segmented liver CT images, the liver CT images with dynamic changes are extracted, specifically including:

[0074] (1) Construct a plane model with fused coordinate information;

[0075] (2) Input the feature recognition points after feature segmentation of each CT image into the coordinate information fusion plane model;

[0076] (3) Connect the input feature recognition points one by one according to the time sequence of collection;

[0077] (4) Obtain the slope values ​​of each segment, and define that when the slope value does not belong to (-1,1), the subsequent liver CT image has dynamic changes compared to the previous liver CT image, and extract it.

[0078] S4: Obtain the dynamic change curve of the arranged liver CT images;

[0079] It should be noted that in the coordinate information fusion planar model, the extracted feature recognition points are connected by a smooth curve to obtain the dynamic change curve of the arranged liver CT images.

[0080] S5: Obtain the standard floating line of the dynamic change curve;

[0081] It should be noted that: in the coordinate information fusion planar model, the input feature recognition points are averaged to obtain the standard floating line of the dynamic change curve.

[0082] like Figure 10 As shown ( Figure 10 In the figure, the horizontal axis represents the change in time / day, and the vertical axis represents the degree of liver deterioration compared to a healthy liver (range / degree - this range can be customized, preferably 0-500). In the figure, 1 is the dynamic change curve of liver deterioration before feature segmentation, 2 is the dynamic change curve, and (A) is the standard floating line. It can be clearly seen from the figure that the curve features after feature segmentation are more obvious and the fluctuation is greater than that of the standard floating line, which is more beneficial for subsequent analysis of the development of liver deterioration.

[0083] S6: Using the standard floating line as a benchmark, construct the maximum and minimum loss difference of the dynamic change curve, and establish the allowable range of loss difference;

[0084] Furthermore, the maximum and minimum loss differences of the dynamic change curves are obtained using the following formulas:

[0085] Y max ′=Y max(动态变化曲线图) -Y (标准浮动线) ;

[0086] Y min ′=Y min(动态变化曲线图) -Y (标准浮动线) ;

[0087] H = (Y min ′, Y max ′);

[0088] Among them, Y max ' represents the maximum loss difference; Y min ' represents the minimum loss difference; H represents the allowable range of the loss difference; Y represents the longitudinal position coordinate value of the coordinate information fusion plane model.

[0089] S7: Obtain the minimum and maximum values ​​during the dynamic change period, and define the liver transformation warning line when the maximum value reaches the threshold, thereby realizing the probability analysis of fused images of liver CT images.

[0090] Furthermore, the minimum and maximum values ​​during the dynamic change period can be obtained using the following formula:

[0091]

[0092]

[0093] Among them, y min The minimum value during the dynamic change period; y max Y represents the maximum value during the dynamic change period. max ' represents the maximum loss difference; Y min ' represents the minimum loss difference; Y represents the longitudinal positional coordinate value of the coordinate information fusion planar model;

[0094] The system trains on the calculated minimum and maximum values ​​within the dynamic change period. The minimum and maximum values ​​within the dynamic change period are defined as being within the allowable range of loss difference to be considered as qualified for training. Otherwise, the selection and calculation of the minimum and maximum values ​​within the dynamic change period are repeated.

[0095] This invention provides a method for fusing continuous and discontinuous dynamic imaging information from liver CT images. Combining machine learning algorithms, it extracts and identifies feature points to filter and segment images, obtaining dynamic change curves and standard floating lines. It constructs the maximum and minimum loss differences of the dynamic change curves, establishes an allowable range for the loss difference, and obtains the minimum and maximum values ​​within the dynamic change period. Based on the maximum value, it judges threshold differences, achieving probabilistic analysis of the fused liver CT images. This addresses the problem that in existing liver disease diagnosis and treatment, doctors can only judge treatment effectiveness based on the intuitive appearance of tissues in dynamically enhanced images. On the one hand, this evaluation process is relatively crude, resulting in a heavy workload and low efficiency for doctors. On the other hand, the information-assisted diagnosis and treatment process does not perform accurate three-dimensional image fusion, resulting in insufficiently comprehensive lesion information and easy neglect of key feature points, thus increasing the difficulty of diagnosis and treatment.

[0096] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for fusing continuous and discontinuous dynamic imaging information from liver CT images, characterized in that, include: Liver CT images were acquired during a preset period. This includes CT imaging, acquiring current CT images of the liver; Analyze the current liver CT images to determine the number of image acquisition periods; acquire all liver CT images within the preset period and perform inter-slice alignment processing; By introducing liver image recognition features, constructing a recognition objective function, and extracting the recognition feature regions of all the liver CT images; Extracting identification feature points from all the liver CT images and performing feature segmentation on the liver CT images based on the extracted identification feature points includes: constructing a dimensional information fusion planar model; constructing an identification model using the XGBoost model algorithm to identify the identification feature regions and identification feature points of normal liver CT images; using the identification feature points as feature segmentation center points, the corresponding identification feature regions as a first range region, and the identification feature regions of normal liver CT images as a second range region, and fusing and connecting the identification feature points of the identification feature points with those of normal liver CT images to perform feature segmentation on the first range region and the second range region; Obtain the feature recognition region after feature segmentation, and obtain the feature recognition points of the feature recognition region based on the recognition model constructed using the XGBoost model algorithm; The process of extracting dynamically changing liver CT images by comparing the segmented liver CT images includes: constructing a coordinate information fusion plane model; inputting the feature recognition points after feature segmentation of each CT image into the coordinate information fusion plane model, where the extracted feature recognition points are connected by a smooth curve to obtain a dynamic change curve of the arranged liver CT images; averaging the input feature recognition points in the coordinate information fusion plane model to obtain a standard floating line of the dynamic change curve; connecting the input feature recognition points one by one according to the acquisition time sequence; obtaining the slope value of each segment, and defining that the current liver CT image has dynamic changes compared to the previous liver CT image when the slope value is not in (-1,1), and extracting the current liver CT image; and arranging the extracted liver CT images in a time-aligned manner. Obtain dynamic change curves of the arranged liver CT images; Obtain the standard floating line of the dynamic change curve; Using the standard floating line as a reference, the maximum and minimum loss differences of the dynamic change curve are constructed, and the allowable range of the loss difference is established. The maximum and minimum loss differences of the dynamic change curve are obtained using the following formula. in, This represents the maximum loss difference. H represents the minimum loss difference; H is the allowable range of the loss difference; Y is the longitudinal positional coordinate value of the coordinate information fusion planar model. The minimum and maximum values ​​during the dynamic change period are obtained, and a liver degeneration warning line is defined when the maximum value reaches a threshold. This enables probabilistic analysis of fused liver CT images. The minimum and maximum values ​​during the dynamic change period are obtained using the following formula. Where n is the nth liver CT image; y min The minimum value during the dynamic change period; y max This represents the maximum value during the dynamic change period. This represents the maximum loss difference. Y represents the minimum loss difference; Y is the longitudinal positional coordinate value of the coordinate information fusion planar model. The calculated minimum and maximum values ​​within the dynamic change period are then used for training. The minimum and maximum values ​​within the dynamic change period are defined as meeting the allowable range of the loss difference to be considered as qualified for training. Otherwise, the selection and calculation of the minimum and maximum values ​​within the dynamic change period are repeated.

2. The method for fusing continuous and discontinuous dynamic imaging information of liver CT images according to claim 1, characterized in that: Analyzing the liver CT images from the current period to determine the specific period of image acquisition includes... Determine the corresponding nature of the lesion, including the location, size, and severity of the lesion. The liver CT image at the current stage is segmented according to the corresponding nature of the lesion; A normal liver CT image is acquired, and based on this CT image, the difference between the current liver CT image and the normal liver CT image in the corresponding nature of the lesion is obtained. When the difference reaches a threshold of 25%, the number of acquisition periods is determined to be 5; when the difference reaches a threshold of 35%, the number of acquisition periods is determined to be 7; and when the difference is greater than the threshold of 35%, the number of acquisition periods is determined to be 10.

3. The method for fusing continuous and discontinuous dynamic imaging information of liver CT images according to claim 2, characterized in that: After introducing the liver image recognition features, the liver image recognition features are amplified to minimize the loss value output by the recognition objective function.

4. The method for fusing continuous and discontinuous dynamic imaging information of liver CT images according to claim 3, characterized in that: Extracting the identification feature points from all the liver CT images specifically includes, A recognition model is constructed using the XGBoost model algorithm; Input all the recognition feature regions of each of the liver CT images respectively, and use the recognition features of the liver images as the recognition benchmark to initially identify the recognition feature points of all the liver CT images; Each identification feature point is input into the identification model for self-learning and self-optimization to obtain the optimal parameters of the identification model. Each optimal parameter is the identification feature point of each identification feature region.

Images