Multi-model combined CT image processing method and system

Abstract

The invention provides a multi-model combination CT image processing method and system, and relates to the technical field of medical image processing, and the method comprises the steps: inputting a lesion CT image into a lesion detection model, obtaining a region where a lesion is located in the lesion CT image, and obtaining a lesion image block according to the region where the lesion is located; inputting the lesion image blocks into a coding module of the trained canceration recognition model to obtain a feature map of the lesion image blocks; inputting the feature map into a canceration probability identification module to obtain a probability feature vector; inputting the feature map into an illness state development stage recognition module to obtain an illness state development stage feature vector; inputting the feature map, the probability feature vector and the disease development stage feature vector into a decoding module to obtain a canceration recognition result; and obtaining a lesion analysis report according to the canceration identification result of each lesion image block. According to the invention, the probability of occurrence of cancerization of the lesion and the development stage of the lesion can be determined, and the performance of the cancerization recognition model and the judgment accuracy of the lesion are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a multi-model combined CT image processing method and system. Background Art

[0002] In the related art, CN119313677A discloses a medical injury CT image processing method and system, which relates to the field of medical image processing, and the method includes: performing data preprocessing on the CT image; determining the noise type of the noise in the CT image; determining the noise intensity of the noise in the CT image; determining a training data set according to the noise type and the noise intensity; training a noise processing model according to the training data set to obtain a trained noise processing model; processing the CT image according to the trained noise processing model to obtain a denoised CT image; determining the lesion area at the lesion position in the denoised CT image, and determining the second CT value of the pixel at the lesion position; determining the lesion severity coefficient according to the lesion area and the second CT value; generating an injury report according to the lesion severity coefficient. According to this scheme, the accuracy and effectiveness of the detailed analysis of the lesion position in the CT image can be improved.

[0003] CN119205714A discloses a chest CT image processing method and system, which relates to the technical field of medical image processing. The main scheme is: using dual-energy CT scanning to obtain image data of two groups of X-rays with different energies and attenuation measurement values ​​of each pixel point, respectively calculating the concentration values ​​of soft tissue and blood vessels and generating a concentration map; generating an initial image based on the soft tissue and blood vessel concentration maps and calculating pixel values, obtaining target image pixel values ​​by contrast stretching and segmenting to obtain a binary image, distinguishing blood vessel areas and soft tissue areas in the binary image by identifying and analyzing connected areas in the binary image; extracting the center line of the blood vessel area and calculating the local diameter, respectively comparing it with the blood vessel expansion threshold and the blood vessel contraction threshold to determine abnormal blood vessel expansion or contraction; extracting features from the soft tissue area to form a first feature vector and comparing it with a preset second feature vector to determine abnormalities; automatically marking or highlighting abnormalities to improve diagnostic accuracy and efficiency.

[0004] CN119151967B discloses a medical image analysis method and system based on plain scan CT data, and relates to the technical field of medical image analysis. The medical image analysis method based on plain scan CT data obtains the chest plain scan CT data of the patient, and the chest plain scan CT data includes chest cross-sectional image data of several scanning positions; establishes a three-dimensional model of the patient's chest based on the chest plain scan CT data of the patient; and performs segmentation processing on the three-dimensional model of the patient's chest to obtain a three-dimensional model of the patient's lungs. The present invention automatically reconstructs a three-dimensional model of healthy lungs by using a deep learning model, and compares it with the actual lung model of the patient, thereby providing a more accurate disease diagnosis. This automated comparison reduces human errors and improves the efficiency of analysis, so that doctors can quickly and accurately identify the lesion area and the degree of the disease, thereby formulating a more effective treatment plan.

[0005] Therefore, although the related technology can automatically process CT images through computer image processing methods, it is unable to identify the development stage of lesions in CT images, making it difficult to obtain accurate detection results and subsequent treatment plans.

[0006] The information disclosed in the background technology section of this application is only intended to deepen the understanding of the general background technology of this application, and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art already known to those skilled in the art. Summary of the invention

[0007] The present invention provides a multi-model combined CT image processing method and system, which can solve the technical problem that the related technology cannot identify the development stage of the lesion in the CT image.

[0008] According to a first aspect of the present invention, a multi-model combined CT image processing method is provided, comprising: Inputting the lesion CT image into the lesion detection model, obtaining the area where the lesion is located in the lesion CT image, and obtaining the lesion image block according to the area where the lesion is located; Inputting the lesion image block into the encoding module of the trained cancer recognition model to obtain a feature map of the lesion image block; Inputting the feature map into a canceration probability recognition module to obtain a probability feature vector of canceration of the lesion in the lesion image block; Inputting the characteristic graph into a disease progression stage recognition module to obtain a disease progression stage characteristic vector of the lesion in the lesion image block; Inputting the feature map, the probability feature vector and the disease progression stage feature vector into a decoding module of a canceration recognition module to obtain a canceration recognition result of a lesion in a lesion image block; According to the canceration recognition results of the lesions in each lesion image block, a lesion analysis report of the lesion CT image is obtained.

[0009] According to a second aspect of the present invention, there is provided a multi-model combined CT image processing system, comprising: A detection module is used to input the lesion CT image into the lesion detection model, obtain the area where the lesion is located in the lesion CT image, and obtain the lesion image block according to the area where the lesion is located; An input module, used for inputting the lesion image block into the encoding module of the trained cancer recognition model to obtain a feature map of the lesion image block; A probability module, used for inputting the feature map into a canceration probability recognition module to obtain a probability feature vector of canceration of a lesion in a lesion image block; A stage module, used for inputting the feature map into a disease progression stage recognition module to obtain a disease progression stage feature vector of the lesion in the lesion image block; A result module, used for inputting the feature map, the probability feature vector and the disease progression stage feature vector into a decoding module of a canceration recognition module to obtain a canceration recognition result of a lesion in a lesion image block; The report module is used to obtain a lesion analysis report of the lesion CT image according to the canceration recognition results of the lesions in each lesion image block.

[0010] By adopting the above technical solution, the present invention can achieve the following technical effects: According to the present invention, the probability feature vector of each lesion becoming cancerous can be determined by the cancer probability recognition module of the cancer recognition model to describe the probability of the lesion becoming cancerous, and the disease progression stage feature vector of each lesion can be determined by the disease progression stage recognition module to describe the development stage of the lesion, and then the accuracy of the cancer recognition result can be improved based on the probability of the lesion becoming cancerous and the development stage of the lesion, so that the cancer recognition result can more accurately describe the risk of cancer, improve the performance of the cancer recognition model and the accuracy of the judgment of the lesion. When determining the image authenticity loss function, the image authenticity loss function can be determined by adversarial training, so that the realism of the image generated by the image generation model can be improved while improving the discrimination ability of the discrimination model, so that the performance of the two models can be balanced, the realism of the first generated lesion CT image or the second generated lesion CT image can be improved, and the accuracy of subsequent comparison can be improved. When determining the disease characteristic loss function, the sample lesion CT images of cancer patients and non-cancer patients can be used to train the cancer recognition model separately, and during the training, the image generation model can be used to predict the generated lesion CT images of the next disease development stage based on the sample lesion CT images of any disease development stage. Therefore, the error between the generated lesion CT images and the real lesion CT images can be used to determine the accuracy of the encoding module, the cancer probability recognition module and the disease development stage recognition module. When the number of samples is small, the generated images can be used to increase the number and intensity of training of the encoding module, the cancer probability recognition module and the disease development stage recognition module to improve the training effect. In the training process, higher weights can be given to critical disease development stages to improve the accuracy of the cancer recognition model in a targeted manner. When determining the first cancer risk loss function, the error between the first cancer identification result and the cancer risk labeling information can be solved, and the error factors of the first sample probability feature vector and the first sample disease progression stage feature vector can be removed to determine the actual error between the cancer risk of the lesion in the first sample feature map and the labeling information. The error of the first sample probability feature vector and the first sample disease progression stage feature vector, as well as the error of the first cancer identification result, can be reduced during the training process, thereby improving the overall accuracy of the cancer identification model. When determining the second cancer risk loss function, the error between the second cancer identification result and the cancer risk labeling information can be solved, and the error factors of the second sample probability feature vector and the second sample disease progression stage feature vector can be removed to determine the actual error between the cancer risk of the lesion in the second sample feature map and the labeling information. The error of the second sample probability feature vector and the second sample disease progression stage feature vector, as well as the error of the second cancer identification result, can be reduced during the training process, thereby improving the overall accuracy of the cancer identification model.

[0011] It should be understood that the above general description and the following detailed description are exemplary and explanatory only and do not limit the present invention. Other features and aspects of the present invention will become more apparent from the following detailed description of exemplary embodiments with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other embodiments can be obtained based on these drawings without creative work. Figure 1 A schematic diagram of a process flow of a multi-model combined CT image processing method according to an embodiment of the present invention is exemplarily shown; Figure 2 A block diagram of a multi-model combined CT image processing system according to an embodiment of the present invention is exemplarily shown. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0015] Figure 1 A schematic flow chart of a multi-model combined CT image processing method according to an embodiment of the present invention is exemplarily shown, and the method comprises: Step S101, inputting the lesion CT image into the lesion detection model, obtaining the lesion area in the lesion CT image, and obtaining the lesion image block according to the lesion area; Step S102, inputting the lesion image block into the encoding module of the trained cancer recognition model to obtain a feature map of the lesion image block; Step S103, inputting the feature map into a canceration probability recognition module to obtain a probability feature vector of canceration of the lesion in the lesion image block; Step S104, inputting the feature map into a disease progression stage recognition module to obtain a disease progression stage feature vector of the lesion in the lesion image block; Step S105, inputting the feature map, the probability feature vector and the disease progression stage feature vector into a decoding module of a canceration recognition module to obtain a canceration recognition result of a lesion in a lesion image block; Step S106, obtaining a lesion analysis report of the lesion CT image according to the canceration recognition result of the lesion in each lesion image block.

[0016] According to the multi-model combined CT image processing method of an embodiment of the present invention, the probability feature vector of each lesion becoming cancerous can be determined by the cancer probability recognition module of the cancer recognition model, so as to describe the probability of the lesion becoming cancerous, and the disease progression stage feature vector of each lesion can be determined by the disease progression stage recognition module, so as to describe the development stage of the lesion. Furthermore, based on the probability of the lesion becoming cancerous and the development stage of the lesion, the accuracy of the cancer recognition result can be improved, so that the cancer recognition result can more accurately describe the risk of cancer, thereby improving the performance of the cancer recognition model and the accuracy of the judgment of the lesion.

[0017] According to one embodiment of the present invention, in step S101, the lesion detection model is a deep learning neural network model, for example, a convolutional neural network model, which can be used to detect the area where the lesion is located in the CT image. For example, the lesion can be framed by a rectangular frame, and the area where the rectangular frame is located is the area where the lesion is located. Furthermore, the rectangular frame can be screenshotted to obtain a lesion image block.

[0018] According to one embodiment of the present invention, in step S102, the cancer recognition model is a deep learning neural network model, which may include an encoding module, a cancer probability recognition module, a disease progression stage recognition module and a decoding module, each of which may include multiple levels, such as a convolution layer, an activation layer, a pooling layer, etc., and each module has a specific function. The encoding module can be used to encode the lesion image block to obtain multiple feature maps, each of which may have a lower resolution than the lesion image block, and each feature map is regarded as an image obtained by observing the lesion image block from a specific perspective, so feature maps of multiple perspectives can be obtained, that is, feature maps of multiple feature channels.

[0019] According to one embodiment of the present invention, in step S103, the canceration probability identification module may process multiple feature maps to obtain a probability feature vector of the lesion becoming cancerous. For example, the vector is composed of two components, the first component is the probability of the lesion becoming cancerous, and the second component is the probability of the lesion not becoming cancerous, and the sum of the values ​​of the two components is 1.

[0020] According to one embodiment of the present invention, in step S104, the disease progression stage identification module can be used to identify the progression stage of the lesion, and the progression stage feature vector can be composed of multiple components, and the number of components is consistent with the number of progression stages of the lesion. For example, the canceration stage is divided into the precancerous lesion stage, the atypical growth stage, the canceration stage, etc., and the progression stage feature vector can include three components, each component is the probability that the progression stage of the lesion is in the above three stages, and the sum of the values ​​of the three components is 1. If a lesion is not a cancerous lesion, its recovery stage can also include three stages, for example, the early recovery stage, the mid-term recovery stage, and the full recovery stage, etc., and the progression stage feature vector can also include three components. In order to distinguish it from a cancerous lesion, each component is the opposite number of the probability that the progression stage of the lesion is in the above three stages, and the sum of the values ​​of the three components is -1.

[0021] According to one embodiment of the present invention, in step S105, the feature map, probability feature vector and disease progression stage feature vector can be processed by the decoding module of the canceration identification module to obtain a canceration identification result. The canceration identification result can be information used to describe the risk of canceration, for example, a canceration risk coefficient score value. The higher the value, the higher the risk of canceration. For example, if the tumor is in an advanced stage and has a greater risk of spread, the value is higher, for example, close to 1. If the tumor is in an early stage and has a lower risk of spread, the value is lower, for example, close to 0.5. If the lesion is not a tumor, the value is even lower, for example, close to 0.

[0022] According to one embodiment of the present invention, in step S106, the canceration identification results of the lesions of each lesion image block can be summarized to obtain a lesion analysis report of the lesion CT image. For example, text can be generated to describe the risk of each lesion, and lesions with a higher risk of canceration can be emphasized to prompt patients and doctors to observe and treat, which is conducive to the subsequent formulation of an accurate treatment plan.

[0023] According to one embodiment of the present invention, before using the canceration recognition model for the above processing, the canceration recognition model may be trained. The training steps of the canceration recognition model include: obtaining first sample lesion CT images of cancer patients at multiple disease development stages, and second sample lesion CT images of non-cancer patients at multiple disease development stages; inputting the first sample lesion CT images of multiple disease development stages into the encoding module of the canceration recognition model to obtain a first sample feature map; inputting the first sample feature map into the canceration probability recognition module to obtain a first sample probability feature vector; inputting the first sample feature map into the disease development stage recognition module to obtain a first sample disease development stage feature vector; inputting the second sample lesion CT images of multiple disease development stages into the encoding module of the canceration recognition model to obtain a second sample feature map; inputting the second sample feature map into the canceration probability recognition module to obtain a second sample probability feature vector; inputting the second sample feature map into the disease development stage recognition module to obtain a second sample disease development stage feature vector; according to the first sample probability feature vector, the first sample disease development stage feature vector, and the second sample probability feature vector , the second sample disease progression stage feature vector, the image generation model and the discrimination model to obtain a disease progression similarity loss function; the first sample probability feature vector, the first sample disease progression stage feature vector and the first sample feature map are input into the decoding module of the canceration recognition module to obtain a first canceration recognition result; the second sample probability feature vector, the second sample disease progression stage feature vector and the second sample feature map are input into the decoding module of the canceration recognition module to obtain a second canceration recognition result; according to the first sample probability feature vector, the first sample disease progression stage feature vector, the first canceration recognition result, the second sample probability feature vector, the second sample disease progression stage feature vector and the second canceration recognition result, a canceration recognition loss function is obtained; according to the disease progression similarity loss function and the canceration recognition loss function, a loss function of the canceration recognition model is obtained; according to the loss function of the canceration recognition model, the canceration recognition model is trained to obtain a trained canceration recognition model.

[0024] According to one embodiment of the present invention, during training, the first sample lesion CT images of multiple cancer patients can be combined into one data set, and the second sample lesion CT images of multiple non-cancer patients can be combined into another data set. During each training, the first sample lesion CT images of a cancer patient at multiple disease development stages and the second sample lesion CT images of a non-cancer patient at multiple disease development stages can be selected to train the canceration recognition model.

[0025] According to one embodiment of the present invention, the first sample feature map, the first sample probability feature vector and the first sample disease progression stage feature vector of the first sample lesion CT image can be obtained respectively through the encoding module, the canceration probability recognition module and the disease progression stage recognition module of the canceration recognition model. Similarly, the second sample feature map, the second sample disease progression stage feature vector and the second sample probability feature vector of the second sample lesion CT image can also be obtained.

[0026] According to one embodiment of the present invention, during training, other neural network models can be used to assist in training, thereby increasing the intensity of training and increasing the amount of training when there are fewer samples. According to the first sample probability feature vector, the first sample disease progression stage feature vector, the second sample probability feature vector, the second sample disease progression stage feature vector, the image generation model and the discriminant model, a disease progression similarity loss function is obtained, including: inputting the first sample lesion CT image of the i-th disease progression stage, and its corresponding first sample probability feature vector and first sample disease progression stage feature vector into the image generation model to obtain the first generated lesion CT image of the i+1-th disease progression stage, where i is a positive integer; inputting the second sample lesion CT image of the j-th disease progression stage, and its corresponding second sample probability feature vector and second sample disease progression stage feature vector into the image generation model to obtain the second generated lesion CT image of the j+1-th disease progression stage, where j is a positive integer; according to the first generated lesion CT image of the i+1-th disease progression stage, according to the first sample lesion CT image of the i+1-th disease progression stage, and the j+1-th disease progression stage feature vector, The second generated lesion CT image of the j+1th disease development stage, the second sample lesion CT image of the j+1th disease development stage and the discriminant model are used to obtain an image authenticity loss function; the first generated probability feature vector and the first generated disease development stage feature vector of the first generated lesion CT image of the i+1th disease development stage are obtained, and the second generated probability feature vector and the second generated disease development stage feature vector of the second generated lesion CT image of the j+1th disease development stage are obtained; the disease characteristic loss function is obtained according to the first sample probability feature vector, the first sample disease development stage feature vector, the first generated probability feature vector, the first generated disease development stage feature vector, the second sample probability feature vector, the second sample disease development stage feature vector, the second generated probability feature vector and the second generated disease development stage feature vector; the disease development similarity loss function is obtained according to the image authenticity loss function and the disease characteristic loss function.

[0027] According to one embodiment of the present invention, an image generation model can be used to predict a CT image of the next stage of disease progression based on a first sample lesion CT image of a cancer patient at a certain stage of disease progression, a first sample probability feature vector, and a first sample disease progression stage feature vector, thereby determining the accuracy of the image generation model and the accuracy of the first sample probability feature vector and the first sample disease progression stage feature vector based on the comparison between the predicted CT image of the next stage of disease progression and the first sample lesion CT image actually taken at the next stage of disease progression. Similarly, an image generation model can be used to predict a CT image of the next stage of disease progression based on a second sample lesion CT image of a non-cancer patient at a certain stage of disease progression, a second sample probability feature vector, and a second sample disease progression stage feature vector, thereby determining the accuracy of the image generation model and the accuracy of the second sample probability feature vector and the second sample disease progression stage feature vector based on the comparison between the predicted CT image of the next stage of disease progression and the second sample lesion CT image actually taken at the next stage of disease progression.

[0028] According to one embodiment of the present invention, the authenticity of the image generated by the image generation model can be first trained, that is, the generated image is similar to the actual captured image in terms of image effect, so that the comparison result can be more accurate in the subsequent comparison process. The training can be assisted by a discriminant model. According to the first generated lesion CT image of the i+1th disease development stage, the first sample lesion CT image of the i+1th disease development stage, the second generated lesion CT image of the j+1th disease development stage, the second sample lesion CT image of the j+1th disease development stage and the discriminant model, an image authenticity loss function is obtained, including: inputting the first generated lesion CT image, the first sample lesion CT image, the second generated lesion CT image or the second sample lesion CT image into the discriminant model to obtain the authenticity discrimination probability distribution; according to formula (1), the image authenticity loss function is obtained , (1) in, is the authenticity discrimination probability distribution obtained when the image input to the discrimination model is the first sample lesion CT image or the second sample lesion CT image, is the authenticity discrimination probability distribution obtained when the image input to the discrimination model is the first generated lesion CT image or the second generated lesion CT image, Indicates that The direction of minimization adjusts the parameters of the image generation model, Indicates that The direction of maximization adjusts the parameters of the discriminant model.

[0029] According to one embodiment of the present invention, in theory, the probability that the discriminant model judges the first sample lesion CT image or the second sample lesion CT image as a real image is 100%, and the probability that the discriminant model judges the first generated lesion CT image or the second generated lesion CT image as a real image is 0. However, as the training process proceeds, the realism of the generation model becomes higher and higher, resulting in an increase in the probability that the discriminant model judges the first generated lesion CT image or the second generated lesion CT image as a real image. Moreover, as the discriminant model also improves its discriminant ability during training, the probability that the discriminant model judges the first generated lesion CT image or the second generated lesion CT image as a real image may be reduced again. Ultimately, the performance of the discriminant model and the image generation model can be balanced, so that when the discriminant model has a high discriminant accuracy, it is still difficult to judge whether the first generated lesion CT image or the second generated lesion CT image generated by the image generation model is a real image, that is, the realism of the image generated by the image generation model is made higher.

[0030] According to an embodiment of the present invention, according to the above training method for balancing the performance of the image generation model and the discriminant model, the discriminant model can be adjusted in a direction that maximizes the image authenticity loss function represented by formula (1), and the image generation model can be adjusted in a direction that minimizes the image authenticity loss function. In formula (1), if the input image is the first sample lesion CT image or the second sample lesion CT image, then in the training of the discriminant model, Maximize, that is, as close to 1 as possible, so that Maximize, that is, maximize the image authenticity loss function, so as to improve the accuracy of the discriminant model in judging the real image. If the input image is the first generated lesion CT image or the second generated lesion CT image, then in the training of the discriminant model, Minimize, that is, as close to 0 as possible, so that Maximize, thereby improving the accuracy of the discriminant model in judging the generated images.

[0031] According to one embodiment of the present invention, on the other hand, in formula (1), if the input image is the first generated lesion CT image or the second generated lesion CT image, then in the training of the image generation model, Maximize, that is, as close to 1 as possible, so that Minimization increases the probability that the first generated lesion CT image or the second generated lesion CT image generated by the image generation model is recognized as a real image by the discrimination model, thereby improving the realism of the image generated by the image generation model and improving the performance of the image generation model.

[0032] In this way, the image authenticity loss function can be determined through adversarial training, so that the realism of the images generated by the image generation model can be improved while the discrimination ability of the discrimination model can be improved, thereby balancing the performance of the two models, improving the realism of the first generated lesion CT image or the second generated lesion CT image, and improving the accuracy of subsequent comparisons.

[0033] According to one embodiment of the present invention, the first generated probability feature vector and the first generated disease progression stage feature vector of the first generated lesion CT image of the i+1th disease progression stage can be obtained, and the second generated probability feature vector and the second generated disease progression stage feature vector of the second generated lesion CT image of the j+1th disease progression stage can be obtained. For example, the above-mentioned first generated probability feature vector, the first generated disease progression stage feature vector, the second generated probability feature vector and the second generated disease progression stage feature vector can also be obtained through the encoding module, the canceration probability identification module and the disease progression stage identification module, which will not be repeated here.

[0034] According to one embodiment of the present invention, the accuracy of the encoding module, the canceration probability identification module, and the disease progression stage identification module can be determined based on the first sample probability feature vector, the first sample disease progression stage feature vector, the first generated probability feature vector, the first generated disease progression stage feature vector, the second sample probability feature vector, the second sample disease progression stage feature vector, the second generated probability feature vector, and the second generated disease progression stage feature vector, and then the encoding module, the canceration probability identification module, and the disease progression stage identification module can be trained.

[0035] According to an embodiment of the present invention, according to the first sample probability feature vector, the first sample disease progression stage feature vector, the first generated probability feature vector, the first generated disease progression stage feature vector, the second sample probability feature vector, the second sample disease progression stage feature vector, the second generated probability feature vector and the second generated disease progression stage feature vector, obtaining a disease characteristic loss function includes: obtaining a disease characteristic loss function according to formula (2): , (2) in, is the first sample probability feature vector corresponding to the first sample lesion CT image at the i-th stage of disease development, is the labeled cancer probability feature vector of cancer patients, n is the number of cancer patients’ disease progression stages, is the first sample disease progression stage feature vector corresponding to the first sample lesion CT image at the i-th disease progression stage, is the first disease progression stage labeling feature vector corresponding to the first sample lesion CT image at the i-th disease progression stage, is the first generated probability feature vector of the first generated lesion CT image at the i+1th stage of disease development, is the first generated disease progression stage feature vector of the first generated lesion CT image at the i+1th disease progression stage, The first disease progression stage labeling feature vector corresponding to the first sample lesion CT image of the i+1th disease progression stage, is the second sample probability feature vector corresponding to the second sample lesion CT image at the jth stage of disease development, is the labeled cancer probability feature vector of non-cancer patients, m is the number of disease progression stages of non-cancer patients, is the second sample disease progression stage feature vector corresponding to the second sample lesion CT image at the jth disease progression stage, The second disease progression stage labeling feature vector corresponding to the second sample lesion CT image at the jth disease progression stage, is the second generated probability feature vector of the second generated lesion CT image at the j+1th stage of disease development, is the second sample disease progression stage feature vector of the second generated lesion CT image at the j+1th disease progression stage, The second disease progression stage labeling feature vector corresponding to the second sample lesion CT image at the j+1th disease progression stage, , , , , , , and are preset weights, i, j, n, m are all positive integers, and i≤n, j≤m.

[0036] According to one embodiment of the present invention, in formula (2), is the first sample probability feature vector corresponding to the first sample lesion CT image at the i-th stage of disease development and the labeled cancer probability feature vector (for example, ), during the training process, the error can be reduced, the accuracy of the encoding module and the cancer probability recognition module can be improved, and can be used as the weight of the error. Since the patient is a cancer patient, the larger the value of the disease progression stage, the closer the disease progression is to cancer, that is, the more critical the disease is. Therefore, a higher weight can be assigned to increase the weight of the error of the first sample probability feature vector with a larger value in the disease progression stage during training, and specifically improve the accuracy of the encoding module and the canceration probability recognition module in processing the situation with a larger value in the disease progression stage. The errors of the first sample probability feature vectors in each disease progression stage can be weighted and summed to obtain , as one of the disease characteristic loss functions.

[0037] According to one embodiment of the present invention, in formula (2), is the error between the first sample disease progression stage feature vector corresponding to the first sample lesion CT image at the i-th disease progression stage and the first disease progression stage annotation feature vector corresponding to the first sample lesion CT image at the i-th disease progression stage (i.e., the vector used to describe the disease progression stage obtained based on the annotation information of the first sample lesion CT image at the i-th disease progression stage), The meaning of is as described above and will not be repeated here. The error weighted summation of the first sample disease progression stage feature vector corresponding to the first sample lesion CT image at each disease progression stage is obtained. , as an item in the disease feature loss function, is reduced during the training process to improve the accuracy of the encoding module and the disease progression stage recognition module.

[0038] According to one embodiment of the present invention, in formula (2), is the error between the first generated probability feature vector of the first generated lesion CT image at the i+1th stage of disease development and the annotated cancer probability feature vector, The meaning of is as described above and will not be repeated here. The error weighted summation of the first generated probability feature vectors of the first generated lesion CT images at each stage of disease development is obtained. , as an item of the disease characteristic loss function, during the training process, this item is reduced to improve the accuracy of the encoding module, the cancer probability recognition module and the image generation model. In addition, when the number of samples is small, the images generated by the image generation model can be used to increase the number and intensity of training of the encoding module and the cancer probability recognition module, thereby improving the training effect.

[0039] According to one embodiment of the present invention, in formula (2), is the error between the first generated disease progression stage feature vector of the first generated lesion CT image at the i+1th disease progression stage and the first disease progression stage labeled feature vector corresponding to the first sample lesion CT image at the i+1th disease progression stage, The meaning of is as described above and will not be repeated here. The error weighted summation of the first generated disease progression stage feature vector of the first generated lesion CT image at each disease progression stage is obtained. , as an item of the disease characteristic loss function, during the training process, this item is reduced to improve the accuracy of the encoding module, the disease progression stage recognition module and the image generation model. In addition, when the number of samples is small, the images generated by the image generation model can be used to increase the number and intensity of training for the encoding module and the disease progression stage recognition module, thereby improving the training effect.

[0040] According to one embodiment of the present invention, in formula (2), is the second sample probability feature vector corresponding to the second sample lesion CT image at the jth stage of disease development and the annotated cancer probability feature vector of the non-cancer patient (for example, ), during the training process, the error can be reduced, the accuracy of the encoding module and the cancer probability recognition module can be improved, and can be used as the weight of the error. Since the patient is a non-cancer patient, the larger the value of the disease progression stage, the closer the disease progression is to recovery, that is, the less critical the disease is. Therefore, a lower weight can be assigned. Conversely, the smaller the value of the disease progression stage, the higher the weight can be assigned. This increases the weight of the error of the second sample probability feature vector with a smaller value of the disease progression stage during training, and specifically improves the accuracy of the encoding module and the canceration probability recognition module in processing situations with larger values ​​of the disease progression stage. The errors of the second sample probability feature vectors at each disease progression stage can be weighted and summed to obtain , as one of the disease characteristic loss functions.

[0041] According to one embodiment of the present invention, in formula (2), is the error between the second sample disease progression stage feature vector corresponding to the second sample lesion CT image at the jth disease progression stage and the second disease progression stage annotation feature vector corresponding to the second sample lesion CT image at the jth disease progression stage (i.e., the vector used to describe the disease progression stage obtained based on the annotation information of the second sample lesion CT image at the jth disease progression stage), The meaning of is as described above and will not be repeated here. The weighted sum of the errors of the second sample disease progression stage feature vectors corresponding to the second sample lesion CT images at each disease progression stage is obtained. As an item in the disease characteristic loss function, this item is reduced during the training process to improve the accuracy of the encoding module and the disease progression stage recognition module.

[0042] According to one embodiment of the present invention, in formula (2), is the error between the second generated probability feature vector of the second generated lesion CT image at the j+1th stage of disease progression and the labeled canceration probability feature vector of non-cancer patients, The meaning of is as described above and will not be repeated here. The error weighted summation of the second generated probability feature vectors of the second generated lesion CT images at each stage of disease progression is obtained. , as an item of the disease characteristic loss function, during the training process, this item is reduced to improve the accuracy of the encoding module, the cancer probability recognition module and the image generation model. In addition, when the number of samples is small, the images generated by the image generation model can be used to increase the number and intensity of training of the encoding module and the cancer probability recognition module, thereby improving the training effect.

[0043] According to one embodiment of the present invention, in formula (2), is the error between the second sample disease progression stage feature vector of the second generated lesion CT image at the j+1th disease progression stage and the second disease progression stage labeled feature vector corresponding to the second sample lesion CT image at the j+1th disease progression stage, The meaning of is as described above and will not be repeated here. The error weighted summation of the second sample disease progression stage feature vector of the second generated lesion CT image at each disease progression stage is obtained. , as an item of the disease characteristic loss function, during the training process, this item is reduced to improve the accuracy of the encoding module, the disease progression stage recognition module and the image generation model. In addition, when the number of samples is small, the images generated by the image generation model can be used to increase the number and intensity of training for the encoding module and the disease progression stage recognition module, thereby improving the training effect.

[0044] According to one embodiment of the present invention, the above multiple items are weighted and summed to obtain the disease characteristic loss function. During the training process, the parameters of the cancer recognition model and the image generation model can be adjusted by the gradient descent method, thereby improving the accuracy of the encoding module, the cancer probability recognition module, the disease development stage recognition module and the image generation model.

[0045] In this way, the cancer recognition model can be trained separately using sample lesion CT images of cancer patients and non-cancer patients, and during the training, the generated lesion CT images of the next disease development stage can be predicted based on the sample lesion CT images of any disease development stage through the image generation model, so that the error between the generated lesion CT image and the real lesion CT image can be used to determine the accuracy of the encoding module, the cancer probability recognition module and the disease development stage recognition module. When the number of samples is small, the generated images can be used to increase the number and intensity of training of the encoding module, the cancer probability recognition module and the disease development stage recognition module to improve the training effect. In the training process, higher weights can be given to critical disease development stages to improve the accuracy of the cancer recognition model in a targeted manner.

[0046] According to an embodiment of the present invention, after obtaining the image authenticity loss function and the disease characteristic loss function, the two may be weightedly summed to obtain the disease progression similarity loss function.

[0047] According to an embodiment of the present invention, the first sample probability feature vector, the first sample disease progression stage feature vector and the first sample feature map can be processed by a decoding module to obtain a first canceration recognition result. Similarly, the second sample probability feature vector, the second sample disease progression stage feature vector and the second sample feature map can be processed by a decoding module to obtain a second canceration recognition result. Furthermore, a canceration recognition loss function can be obtained based on the error between the first canceration recognition result and the second canceration recognition result.

[0048] According to one embodiment of the present invention, a canceration identification loss function is obtained based on the first sample probability feature vector, the first sample disease progression stage feature vector, the first canceration identification result, the second sample probability feature vector, the second sample disease progression stage feature vector and the second canceration identification result, including: obtaining a first canceration risk loss function based on the first sample probability feature vector, the first sample disease progression stage feature vector and the first canceration identification result; obtaining a second canceration risk loss function based on the second sample probability feature vector, the second sample disease progression stage feature vector and the second canceration identification result; obtaining a canceration identification loss function based on the first canceration risk loss function and the second canceration risk loss function.

[0049] According to an embodiment of the present invention, a first cancer risk loss function is obtained according to the first sample probability feature vector, the first sample disease progression stage feature vector, and the first cancer identification result, including: obtaining the first cancer risk loss function according to formula (3): , (3) in, is the first canceration recognition result of the first sample lesion CT image at the i-th stage of disease development, is the cancer risk annotation information of the first sample lesion CT image at the i-th stage of disease development, is the first sample probability feature vector corresponding to the first sample lesion CT image at the i-th stage of disease development, for The transposed vector of is the labeled cancer probability feature vector of cancer patients, is the first sample disease progression stage feature vector corresponding to the first sample lesion CT image at the i-th disease progression stage, for The transposed vector of is the first disease progression stage labeling feature vector corresponding to the first sample lesion CT image of the i-th disease progression stage, n is the number of disease progression stages of cancer patients, i and n are both positive integers, and i≤n.

[0050] According to one embodiment of the present invention, in formula (3), is the error between the first cancer recognition result of the first sample lesion CT image at the i-th stage of disease progression and the cancer risk annotation information (for example, the manually annotated cancer risk coefficient score). During the training process, the reduction of this error can reduce the first cancer risk loss function, thereby improving the accuracy of the decoding module, the cancer probability recognition module, the disease progression stage recognition module and the encoding module. Further, in the process of decoding to obtain the first cancer recognition result, the first sample probability feature vector and the first sample disease progression stage feature vector are referenced, and these two vectors also have errors. Therefore, in order to determine the cancer risk in the first sample feature map, the error factors of the above two vectors can be excluded, that is, the cosine similarity of the two and the corresponding annotated feature vector is used as the denominator, thereby removing the error factors of the two vectors and determining the actual error between the cancer risk of the lesion in the first sample feature map and the annotation. In addition, using the two cosine similarities as the denominator can also increase the value of the cosine similarity during the training process, reduce the error of the first sample probability feature vector and the first sample disease progression stage feature vector, and improve the overall accuracy of the cancer recognition model. The meaning of is as described above and will not be repeated here. right By weighted summation, the first cancer risk loss function can be obtained.

[0051] In this way, the error between the first cancer identification result and the cancer risk annotation information can be solved, and the error factors of the first sample probability feature vector and the first sample disease progression stage feature vector can be removed to determine the actual error between the cancer risk of the lesion in the first sample feature map and the annotation information. The error of the first sample probability feature vector and the first sample disease progression stage feature vector, as well as the error of the first cancer identification result, can also be reduced during the training process to improve the overall accuracy of the cancer identification model.

[0052] According to an embodiment of the present invention, obtaining a second cancer risk loss function according to the second sample probability feature vector, the second sample disease progression stage feature vector and the second cancer recognition result includes: obtaining the second cancer risk loss function according to formula (4): , (4) in, is the second canceration recognition result of the second sample lesion CT image at the jth stage of disease development, is the cancer risk annotation information of the second sample lesion CT image at the jth stage of disease development, is the second sample probability feature vector corresponding to the second sample lesion CT image at the jth stage of disease development, is the labeled cancer probability feature vector of non-cancer patients, m is the number of disease progression stages of non-cancer patients, is the second sample disease progression stage feature vector corresponding to the second sample lesion CT image at the jth disease progression stage, The second disease progression stage labeling feature vector corresponding to the second sample lesion CT image at the jth disease progression stage, for The transposed vector of for The transposed vector of , j and m are both positive integers, and j≤m.

[0053] According to one embodiment of the present invention, is the error between the second canceration recognition result of the second sample lesion CT image at the jth stage of disease development and the canceration risk annotation information (for example, the manually annotated canceration risk coefficient score) of the second sample lesion CT image at the jth stage of disease development. During the training process, the reduction of this error can reduce the second canceration risk loss function, thereby improving the accuracy of the decoding module, the canceration probability recognition module, the disease development stage recognition module and the encoding module. Further, in the process of decoding to obtain the second canceration recognition result, the second sample probability feature vector and the second sample disease development stage feature vector are referenced, and these two vectors also have errors. Therefore, in order to determine the canceration risk in the second sample feature map, the error factors of the above two vectors can be excluded, that is, the cosine similarity of the two and the corresponding annotated feature vector is used as the denominator, thereby removing the error factors of the two vectors and determining the actual error between the canceration risk of the lesion in the second sample feature map and the annotation. In addition, by using the two cosine similarities as the denominator, the value of the cosine similarity can also be increased during the training process, reducing the error between the second sample probability feature vector and the second sample disease development stage feature vector, and improving the overall accuracy of the canceration recognition model. The meaning of is as described above and will not be repeated here. right By weighted summation, the second cancer risk loss function can be obtained.

[0054] In this way, the error between the second cancer identification result and the cancer risk annotation information can be solved, and the error factors of the second sample probability feature vector and the second sample disease progression stage feature vector can be removed to determine the actual error between the cancer risk of the lesion in the second sample feature map and the annotation information. The error between the second sample probability feature vector and the second sample disease progression stage feature vector, as well as the error of the second cancer identification result, can also be reduced during the training process to improve the overall accuracy of the cancer identification model.

[0055] According to one embodiment of the present invention, the first cancer risk loss function and the second cancer risk loss function may be weighted and summed to obtain a cancer recognition loss function. Further, the disease progression similarity loss function and the cancer recognition loss function may be weighted and summed to obtain a loss function of a cancer recognition model, and then the parameters of the cancer recognition model may be adjusted by a gradient descent method to reduce the loss function of the cancer recognition model, so as to train the cancer recognition model. After multiple trainings, the trained cancer recognition model may be obtained.

[0056] According to the multi-model combined CT image processing method of the embodiment of the present invention, the probability feature vector of each lesion becoming cancerous can be determined by the cancer probability recognition module of the cancer recognition model to describe the probability of the lesion becoming cancerous, and the disease progression stage recognition module can be used to determine the disease progression stage feature vector of each lesion to describe the development stage of the lesion, and then based on the probability of the lesion becoming cancerous and the development stage of the lesion, the accuracy of the cancer recognition result can be improved, so that the cancer recognition result can more accurately describe the risk of cancer, improve the performance of the cancer recognition model and the accuracy of the judgment of the lesion. When determining the image authenticity loss function, the image authenticity loss function can be determined by adversarial training, so that the realism of the image generated by the image generation model can be improved while improving the discrimination ability of the discrimination model, so that the performance of the two models can be balanced, the realism of the first generated lesion CT image or the second generated lesion CT image can be improved, and the accuracy of subsequent comparison can be improved. When determining the disease characteristic loss function, the sample lesion CT images of cancer patients and non-cancer patients can be used to train the cancer recognition model separately, and during the training, the image generation model can be used to predict the generated lesion CT images of the next disease development stage based on the sample lesion CT images of any disease development stage. Therefore, the error between the generated lesion CT images and the real lesion CT images can be used to determine the accuracy of the encoding module, the cancer probability recognition module and the disease development stage recognition module. When the number of samples is small, the generated images can be used to increase the number and intensity of training of the encoding module, the cancer probability recognition module and the disease development stage recognition module to improve the training effect. In the training process, higher weights can be given to critical disease development stages to improve the accuracy of the cancer recognition model in a targeted manner. When determining the first cancer risk loss function, the error between the first cancer identification result and the cancer risk labeling information can be solved, and the error factors of the first sample probability feature vector and the first sample disease progression stage feature vector can be removed to determine the actual error between the cancer risk of the lesion in the first sample feature map and the labeling information. The error of the first sample probability feature vector and the first sample disease progression stage feature vector, as well as the error of the first cancer identification result, can be reduced during the training process, thereby improving the overall accuracy of the cancer identification model. When determining the second cancer risk loss function, the error between the second cancer identification result and the cancer risk labeling information can be solved, and the error factors of the second sample probability feature vector and the second sample disease progression stage feature vector can be removed to determine the actual error between the cancer risk of the lesion in the second sample feature map and the labeling information. The error of the second sample probability feature vector and the second sample disease progression stage feature vector, as well as the error of the second cancer identification result, can be reduced during the training process, thereby improving the overall accuracy of the cancer identification model.

[0057] Figure 2A block diagram of a multi-model combined CT image processing system according to an embodiment of the present invention is exemplarily shown, and the system includes: A detection module is used to input the lesion CT image into the lesion detection model, obtain the area where the lesion is located in the lesion CT image, and obtain the lesion image block according to the area where the lesion is located; An input module, used for inputting the lesion image block into the encoding module of the trained cancer recognition model to obtain a feature map of the lesion image block; A probability module, used for inputting the feature map into a canceration probability recognition module to obtain a probability feature vector of canceration of a lesion in a lesion image block; A stage module, used for inputting the feature map into a disease progression stage recognition module to obtain a disease progression stage feature vector of the lesion in the lesion image block; A result module, used for inputting the feature map, the probability feature vector and the disease progression stage feature vector into a decoding module of a canceration recognition module to obtain a canceration recognition result of a lesion in a lesion image block; The report module is used to obtain a lesion analysis report of the lesion CT image according to the canceration recognition results of the lesions in each lesion image block.

[0058] It should be understood by those skilled in the art that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functional and structural principles of the present invention have been demonstrated and explained in the embodiments, and the embodiments of the present invention may be deformed or modified in any way without departing from the principles.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-model combined CT image processing method, characterized in that: include: Inputting the lesion CT image into the lesion detection model, obtaining the area where the lesion is located in the lesion CT image, and obtaining the lesion image block according to the area where the lesion is located; Inputting the lesion image block into the encoding module of the trained cancer recognition model to obtain a feature map of the lesion image block; Inputting the feature map into a canceration probability recognition module to obtain a probability feature vector of canceration of the lesion in the lesion image block; Inputting the characteristic graph into a disease progression stage recognition module to obtain a disease progression stage characteristic vector of the lesion in the lesion image block; Inputting the feature map, the probability feature vector and the disease progression stage feature vector into a decoding module of a canceration recognition module to obtain a canceration recognition result of a lesion in a lesion image block; According to the canceration recognition results of the lesions in each lesion image block, a lesion analysis report of the lesion CT image is obtained.

2. The multi-model combined CT image processing method according to claim 1, characterized in that: The training steps of the cancer recognition model include: Acquire first sample lesion CT images of cancer patients at multiple disease progression stages, and second sample lesion CT images of non-cancer patients at multiple disease progression stages; Inputting first sample lesion CT images at multiple disease development stages into the encoding module of the canceration recognition model to obtain a first sample feature map; Inputting the first sample feature map into a canceration probability recognition module to obtain a first sample probability feature vector; Inputting the first sample feature graph into a disease progression stage recognition module to obtain a first sample disease progression stage feature vector; Inputting the second sample lesion CT images at multiple disease development stages into the encoding module of the canceration recognition model to obtain a second sample feature map; Inputting the second sample feature map into a canceration probability recognition module to obtain a second sample probability feature vector; Inputting the second sample characteristic graph into a disease progression stage identification module to obtain a second sample disease progression stage characteristic vector; Obtaining a disease progression similarity loss function according to the first sample probability feature vector, the first sample disease progression stage feature vector, the second sample probability feature vector, the second sample disease progression stage feature vector, an image generation model, and a discriminant model; Inputting the first sample probability feature vector, the first sample disease progression stage feature vector and the first sample feature graph into a decoding module of a canceration recognition module to obtain a first canceration recognition result; Inputting the second sample probability feature vector, the second sample disease progression stage feature vector and the second sample feature graph into a decoding module of a canceration recognition module to obtain a second canceration recognition result; Obtaining a canceration recognition loss function according to the first sample probability feature vector, the first sample disease progression stage feature vector, the first canceration recognition result, the second sample probability feature vector, the second sample disease progression stage feature vector, and the second canceration recognition result; Obtaining a loss function of a canceration recognition model according to the disease progression similarity loss function and the canceration recognition loss function; The canceration recognition model is trained according to the loss function of the canceration recognition model to obtain a trained canceration recognition model.

3. The multi-model combined CT image processing method according to claim 2, characterized in that: According to the first sample probability feature vector, the first sample disease progression stage feature vector, the second sample probability feature vector, the second sample disease progression stage feature vector, the image generation model and the discrimination model, a disease progression similarity loss function is obtained, including: Inputting the first sample lesion CT image of the i-th disease progression stage, and its corresponding first sample probability feature vector and first sample disease progression stage feature vector into the image generation model, to obtain the first generated lesion CT image of the i+1-th disease progression stage, where i is a positive integer; Input the second sample lesion CT image of the j-th disease progression stage, and its corresponding second sample probability feature vector and second sample disease progression stage feature vector into the image generation model to obtain the second generated lesion CT image of the j+1-th disease progression stage, where j is a positive integer; Obtain an image authenticity loss function according to a first generated lesion CT image at the i+1th stage of disease progression, a first sample lesion CT image at the i+1th stage of disease progression, a second generated lesion CT image at the j+1th stage of disease progression, a second sample lesion CT image at the j+1th stage of disease progression, and a discriminant model; Obtaining a first generated probability feature vector and a first generated disease progression stage feature vector of a first generated lesion CT image at the i+1th disease progression stage, and obtaining a second generated probability feature vector and a second generated disease progression stage feature vector of a second generated lesion CT image at the j+1th disease progression stage; Obtaining a disease characteristic loss function according to the first sample probability feature vector, the first sample disease progression stage feature vector, the first generated probability feature vector, the first generated disease progression stage feature vector, the second sample probability feature vector, the second sample disease progression stage feature vector, the second generated probability feature vector, and the second generated disease progression stage feature vector; The disease progression similarity loss function is obtained according to the image authenticity loss function and the disease characteristic loss function.

4. The multi-model combined CT image processing method according to claim 3, characterized in that: According to the first generated lesion CT image at the i+1th stage of disease progression, the first sample lesion CT image at the i+1th stage of disease progression, the second generated lesion CT image at the j+1th stage of disease progression, the second sample lesion CT image at the j+1th stage of disease progression and the discriminant model, an image authenticity loss function is obtained, including: Inputting the first generated lesion CT image, the first sample lesion CT image, the second generated lesion CT image or the second sample lesion CT image into the discrimination model to obtain a authenticity discrimination probability distribution; According to the formula ； Obtaining image authenticity loss function ,in, is the authenticity discrimination probability distribution obtained when the image input to the discrimination model is the first sample lesion CT image or the second sample lesion CT image, is the authenticity discrimination probability distribution obtained when the image input to the discrimination model is the first generated lesion CT image or the second generated lesion CT image, Indicates that The direction of minimization adjusts the parameters of the image generation model, Indicates that The direction of maximization adjusts the parameters of the discriminant model.

5. The multi-model combined CT image processing method according to claim 3, characterized in that: Obtaining a disease characteristic loss function according to the first sample probability feature vector, the first sample disease progression stage feature vector, the first generated probability feature vector, the first generated disease progression stage feature vector, the second sample probability feature vector, the second sample disease progression stage feature vector, the second generated probability feature vector, and the second generated disease progression stage feature vector, includes: According to the formula ； Obtain disease characteristic loss function ,in, is the first sample probability feature vector corresponding to the first sample lesion CT image at the i-th stage of disease development, is the labeled cancer probability feature vector of cancer patients, n is the number of cancer patients’ disease progression stages, is the first sample disease progression stage feature vector corresponding to the first sample lesion CT image at the i-th disease progression stage, is the first disease progression stage labeling feature vector corresponding to the first sample lesion CT image at the i-th disease progression stage, is the first generated probability feature vector of the first generated lesion CT image at the i+1th stage of disease development, is the first generated disease progression stage feature vector of the first generated lesion CT image at the i+1th disease progression stage, The first disease progression stage labeling feature vector corresponding to the first sample lesion CT image of the i+1th disease progression stage, is the second sample probability feature vector corresponding to the second sample lesion CT image at the jth stage of disease development, is the labeled cancer probability feature vector of non-cancer patients, m is the number of disease progression stages of non-cancer patients, is the second sample disease progression stage feature vector corresponding to the second sample lesion CT image at the jth disease progression stage, The second disease progression stage labeling feature vector corresponding to the second sample lesion CT image at the jth disease progression stage, is the second generated probability feature vector of the second generated lesion CT image at the j+1th stage of disease development, is the second sample disease progression stage feature vector of the second generated lesion CT image at the j+1th disease progression stage, The second disease progression stage labeling feature vector corresponding to the second sample lesion CT image at the j+1th disease progression stage, , , , , , , and is the preset weight, Wherein, i, j, n, and m are all positive integers, and i≤n, j≤m.

6. The multi-model combined CT image processing method according to claim 3, characterized in that: Obtaining a canceration recognition loss function according to the first sample probability feature vector, the first sample disease progression stage feature vector, the first canceration recognition result, the second sample probability feature vector, the second sample disease progression stage feature vector, and the second canceration recognition result, including: Obtaining a first canceration risk loss function according to the first sample probability feature vector, the first sample disease progression stage feature vector, and the first canceration identification result; Obtaining a second canceration risk loss function according to the second sample probability feature vector, the second sample disease progression stage feature vector, and the second canceration identification result; A canceration identification loss function is obtained according to the first canceration risk loss function and the second canceration risk loss function.

7. The multi-model combined CT image processing method according to claim 6, characterized in that: Obtaining a first cancer risk loss function according to the first sample probability feature vector, the first sample disease progression stage feature vector, and the first cancer identification result includes: According to the formula ； Obtain the first cancer risk loss function ,in, is the first canceration recognition result of the first sample lesion CT image at the i-th stage of disease development, is the cancer risk annotation information of the first sample lesion CT image at the i-th stage of disease development, is the first sample probability feature vector corresponding to the first sample lesion CT image at the i-th stage of disease development, for The transposed vector of is the labeled cancer probability feature vector of cancer patients, is the first sample disease progression stage feature vector corresponding to the first sample lesion CT image at the i-th disease progression stage, for The transposed vector of is the first disease progression stage labeling feature vector corresponding to the first sample lesion CT image of the i-th disease progression stage, n is the number of disease progression stages of cancer patients, i and n are both positive integers, and i≤n.

8. The multi-model combined CT image processing method according to claim 6, characterized in that: Obtaining a second canceration risk loss function according to the second sample probability feature vector, the second sample disease progression stage feature vector and the second canceration identification result includes: According to the formula ； Obtain the second cancer risk loss function ,in, is the second canceration recognition result of the second sample lesion CT image at the jth stage of disease development, is the cancer risk annotation information of the second sample lesion CT image at the jth stage of disease development, is the second sample probability feature vector corresponding to the second sample lesion CT image at the jth stage of disease development, is the labeled cancer probability feature vector of non-cancer patients, m is the number of disease progression stages of non-cancer patients, is the second sample disease progression stage feature vector corresponding to the second sample lesion CT image at the jth disease progression stage, The second disease progression stage labeling feature vector corresponding to the second sample lesion CT image at the jth disease progression stage, for The transposed vector of for The transposed vector of , j and m are both positive integers, and j≤m.

9. A multi-model combined CT image processing system, used to execute the method according to any one of claims 1 to 8, characterized in that: include: A detection module is used to input the lesion CT image into the lesion detection model, obtain the area where the lesion is located in the lesion CT image, and obtain the lesion image block according to the area where the lesion is located; An input module, used for inputting the lesion image block into the encoding module of the trained cancer recognition model to obtain a feature map of the lesion image block; A probability module, used for inputting the feature map into a canceration probability recognition module to obtain a probability feature vector of canceration of a lesion in a lesion image block; A stage module, used for inputting the feature map into a disease progression stage recognition module to obtain a disease progression stage feature vector of the lesion in the lesion image block; A result module, used for inputting the feature map, the probability feature vector and the disease progression stage feature vector into a decoding module of a canceration recognition module to obtain a canceration recognition result of a lesion in a lesion image block; The report module is used to obtain a lesion analysis report of the lesion CT image according to the canceration recognition results of the lesions in each lesion image block.

Images