Pelvis and sacrum early warning method and device based on non-enhanced MRI end-to-end deep learning model

Abstract

The invention provides a pelvis and sacrum early warning method and device based on a non-enhanced MRI end-to-end deep learning model, and is applied to the technical field of data processing. The method comprises the following steps: preprocessing preoperative magnetic resonance imaging sequence image information of a target user to generate target MRI sequence image information; processing the tumor state influence factors of the pelvis and the sacrum and the tumor risk adjustment factors of the pelvis and the sacrum to generate dynamic risk parameter information of the target user; processing a preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging based on the training sample set and the verification sample set to generate a target end-to-end deep learning model; and processing the target MRI sequence image information, the physiological parameter information of the target user and the dynamic risk parameter information of the target user based on a target end-to-end deep learning model to generate user early warning information.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a pelvic and sacrum early warning method and device based on a non-enhanced MRI end-to-end deep learning model. Background Art

[0002] Pelvic and sacral tumors (PSTs) are relatively rare, among which metastatic tumors are the most common type due to the hematopoietic function of this area. Due to the rarity of PSTs and the similarity of their clinical manifestations and imaging features, radiologists often lack sufficient clinical experience to make judgments. In the early stages, PSTs are usually small and asymptomatic, but once discovered, they have usually grown larger and compressed surrounding organs, often requiring surgical intervention. For all primary malignant sacral tumors and benign lesions involving the lower segment, wide resection should be selected when the S3 nerve root can be preserved. However, due to the complex anatomical structure, the influence of multiple surrounding organs and the difficulty of surgery in this area.

[0003] In recent years, deep learning models have shown great potential in exploring the properties of tumors. However, existing technologies have not yet found early warning of benign and malignant bone tumors based on deep learning models. In this context, the development of sensitive and effective computer-assisted early warning tools has become critical.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore includes information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0005] The purpose of this application is to provide a pelvic and sacral early warning method and device based on a non-enhanced MRI end-to-end deep learning model, which at least overcomes the problems existing in the prior art to a certain extent, by collecting multi-dimensional data of the target user, covering multiple sequences of preoperative magnetic resonance imaging, physiological parameters, serological markers, microbial communities, and lifestyle information, and preparing deep learning models and sample sets based on non-enhanced MRI. Then, the information is processed finely, and the imaging sequence is normalized and filled into a three-channel image to obtain the target MRI sequence image information; the serological and microbial community information is extracted and fused into tumor status influencing factors, and the lifestyle information is refined into risk adjustment factors, and the two are combined to obtain dynamic risk parameter information. Finally, the preset model is trained using the preprocessed sample set, the segmentation model is trained with the three-channel image, and the diagnosis model is trained in combination with other information, and the early warning information is generated according to the threshold.

[0006] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present invention.

[0007] According to one aspect of the present application, a pelvic and sacrum early warning method based on a non-enhanced MRI end-to-end deep learning model is provided, comprising: obtaining preoperative magnetic resonance imaging sequence image information of a target user, physiological parameter information of the target user, serological marker parameter information of the target user, microbial community parameter information, life habit information of the target user, a preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging, a training sample set, and a verification sample set, wherein the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging includes a preset segmentation model and a preset diagnosis model, and the preoperative magnetic resonance imaging sequence image information of the target user includes T1-weighted imaging information of the target user, T2-weighted imaging information of the target user, and diffusion-weighted imaging information of the target user; preprocessing the preoperative magnetic resonance imaging sequence image information of the target user to generate target MRI sequence image information; The user's serological marker parameter information and the microbial community parameter information are processed to generate factors affecting the tumor status of the pelvis and sacrum; the living habit information of the target user is processed to generate risk adjustment factors for the tumors of the pelvis and sacrum; the factors affecting the tumor status of the pelvis and sacrum and the risk adjustment factors for the pelvis and sacrum are processed to generate dynamic risk parameter information for the target user; the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging is processed based on the training sample set and the verification sample set to generate a target end-to-end deep learning model; the target MRI sequence image information, the physiological parameter information of the target user and the dynamic risk parameter information of the target user are processed based on the target end-to-end deep learning model to generate user warning information, wherein the user warning information is used to characterize the attribute information of the target user's preoperative pelvic and sacrum tumors.

[0008] Another aspect of the present application is a pelvic and sacrum early warning device based on a non-enhanced MRI end-to-end deep learning model, characterized in that it includes: an acquisition module for acquiring preoperative magnetic resonance imaging sequence image information of a target user, physiological parameter information of a target user, serological marker parameter information of a target user, microbial community parameter information, living habit information of a target user, a preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging, a training sample set, and a verification sample set, wherein the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging includes a preset segmentation model and a preset diagnosis model, and the preoperative magnetic resonance imaging sequence image information of the target user includes T1-weighted imaging information of the target user, T2-weighted imaging information of the target user, and diffusion-weighted imaging information of the target user; a processing module for preprocessing the preoperative magnetic resonance imaging sequence image information of the target user to generate a target MRI sequence image information; processing the serological marker parameter information and the microbial community parameter information of the target user to generate the tumor status influencing factors of the pelvis and sacrum; processing the living habit information of the target user to generate the tumor risk adjustment factors of the pelvis and sacrum; processing the tumor status influencing factors of the pelvis and sacrum and the tumor risk adjustment factors of the pelvis and sacrum to generate the dynamic risk parameter information of the target user; processing the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging based on the training sample set and the verification sample set to generate a target end-to-end deep learning model; processing the target MRI sequence image information, the physiological parameter information of the target user and the dynamic risk parameter information of the target user based on the target end-to-end deep learning model to generate user warning information, wherein the user warning information is used to characterize the attribute information of the target user's preoperative pelvic and sacrum tumors.

[0009] According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a second processor, the pelvis and sacrum early warning method based on the non-enhanced MRI end-to-end deep learning model is implemented.

[0010] The present application provides a pelvic and sacrum early warning method and device based on a non-enhanced MRI end-to-end deep learning model. The server collects multi-dimensional data of the target user, covering multiple sequences of preoperative magnetic resonance imaging, physiological parameters, serological markers, microbial communities, and lifestyle information, and prepares deep learning models and sample sets based on non-enhanced MRI. Then the information is processed finely, the imaging sequence is normalized and filled into a three-channel image, and the target MRI sequence image information is obtained; the serological and microbial community information is extracted and fused into tumor status influencing factors, and the lifestyle information is refined into risk adjustment factors, and the two are combined to obtain dynamic risk parameter information. Finally, the preset model is trained using the preprocessed sample set, the segmentation model is trained with the three-channel image, and the diagnosis model is trained in combination with other information, and the early warning information is generated according to the threshold.

[0011] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A flowchart of a pelvic and sacrum early warning method based on a non-enhanced MRI end-to-end deep learning model provided by an embodiment of the present application is shown; Figure 2 A structural schematic diagram of a pelvic and sacrum warning device based on a non-enhanced MRI end-to-end deep learning model provided in one embodiment of the present application is shown. DETAILED DESCRIPTION

[0013] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0014] Combine the following Figure 1 To describe the pelvic and sacral early warning method based on the non-enhanced MRI end-to-end deep learning model according to an exemplary embodiment of the present application. In one embodiment, the present application also proposes a pelvic and sacral early warning method and device based on the non-enhanced MRI end-to-end deep learning model. Figure 1 The flowchart of a pelvic and sacrum early warning method based on a non-enhanced MRI end-to-end deep learning model according to an embodiment of the present application is schematically shown. Figure 1 As shown, the method is applied to a server, comprising: S101, obtaining the target user's preoperative magnetic resonance imaging sequence image information, the target user's physiological parameter information, the target user's serological marker parameter information, the target user's microbial community parameter information, the target user's living habit information, a preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging, a training sample set, and a verification sample set.

[0015] In one embodiment, the preoperative magnetic resonance imaging sequence image information of the target user includes T1-weighted imaging information, T2-weighted imaging information, and diffusion-weighted imaging information (DWI). T1-weighted imaging has a good display of anatomical structures, and can clearly present basic structural information such as the bone morphology and soft tissue contours of the pelvis and sacrum, which helps to determine the location of the tumor and its anatomical relationship with the surrounding normal tissues; T2-weighted imaging is more sensitive to edema, inflammation, and other conditions in the diseased tissue, and can highlight the edema area in the tumor tissue and its surroundings, providing a basis for judging the scope and invasion of the tumor; diffusion-weighted imaging (DWI) can reflect the diffusion movement of water molecules in the tissue, which is of great significance for detecting the diffusion-restricted area of ​​water molecules in the tumor tissue, and helps to identify the active part of the tumor, because the dense area of ​​tumor cells usually has restricted diffusion of water molecules, which appears as a high signal on DWI. The information provided by these different imaging methods complements each other, providing rich data for a comprehensive understanding of the morphology, location, size, and internal characteristics of the tumor.

[0016] The target user's physiological parameter information includes gender, age, tumor location and maximum tumor size. Gender information is of certain significance in tumor research. The incidence, type of disease or prognosis of certain tumors vary between different genders. For example, men are more susceptible to certain malignant pelvic and sacral tumors. Age is an important risk factor. Generally speaking, the risk of tumors increases with age, and different types of tumors are more common in different age groups. For example, primary bone tumors are more likely to occur in children and adolescents, while the proportion of metastatic tumors in adults is relatively high. Tumor location is crucial for diagnosis and treatment. Tumors in different locations have different relationships with surrounding organs and tissues, which affects the difficulty of surgical resection and prognosis. For example, tumors located in the sacrum are close to neural structures, and more caution is required during surgery to avoid nerve damage. The maximum tumor size directly reflects the extent of tumor growth. Larger tumors are more likely to cause symptoms, such as compression of surrounding organs leading to dysfunction. It also indicates that the tumor is more invasive or has a longer course of disease. These physiological parameter information are usually obtained during the user's medical treatment through clinical examination, imaging measurement, and medical records. The date of birth or exact age should be recorded for age, the location of the tumor should be described using a unified anatomical naming and positioning method, and the maximum tumor size needs to have clear measurement standards and methods (such as diameter, volume, etc.). During the data collation process, the data is normalized and the gender is encoded into a specific value (such as 0 for female and 1 for male) to facilitate computer processing and model analysis.

[0017] Serological marker parameter information mainly involves specific markers related to pelvic and sacral tumors, such as bone alkaline phosphatase (BALP) and lactate dehydrogenase (LDH). BALP is of great significance in bone diseases, especially bone tumors. Its level changes are closely related to osteoblast activity. In many bone tumors (such as osteosarcoma, osteoblastoma, etc.), BALP levels usually increase due to tumor cells stimulating increased osteoblast activity. Therefore, it can be used as an important indicator for evaluating tumor bone metabolism status and tumor activity; LDH is an enzyme widely present in human tissues. In tumor cells, due to abnormal metabolism, LDH production and release increase, resulting in increased serum LDH levels. In pelvic and sacral tumors (especially malignant tumors such as Ewing sarcoma, etc.), LDH elevation is more common. Its level is related to tumor load, proliferation activity and prognosis, and can be used to monitor tumor progression and treatment response.

[0018] Microbial community parameter information reflects the characteristics of microbial communities related to pelvic and sacral tumors in the target user. In the pelvic and sacral regions, microbial communities affect the tumor microenvironment in a variety of ways, such as regulating local immune responses, affecting cell signal transduction, and participating in metabolic processes. Some microorganisms produce toxins or metabolites that directly damage cell DNA and promote the proliferation and transformation of tumor cells; or by regulating the host immune system and inhibiting immune surveillance functions, creating favorable conditions for the growth and escape of tumor cells. Therefore, analyzing microbial community parameter information helps to gain a deeper understanding of the pathogenesis of tumors. High-throughput sequencing technologies, such as 16SrRNA sequencing and metagenomic sequencing, are usually used to obtain microbial community parameter information. These technologies can comprehensively and deeply analyze the characteristics of the composition structure of the microbial community (including the relative abundance of different microbial species, diversity index, etc.), gene expression activity (reflecting the functional state of microorganisms), and metabolic pathway activity (revealing the role of microorganisms in metabolic processes).

[0019] The target user's lifestyle information covers multiple aspects such as dietary intake, exercise type, exercise intensity and sleep quality, which are closely related to the risk of pelvic and sacral tumors. In terms of dietary intake, a high-sugar, high-fat and high-salt diet increases the risk of chronic diseases such as obesity and diabetes, which in turn indirectly affects the occurrence and development of tumors. For example, long-term excessive intake of red meat and processed meat is related to the occurrence of certain tumors (such as colorectal cancer), and colorectal cancer metastasizes to the pelvis and sacrum; exercise type and intensity have an important impact on the body's functional state, hormone regulation, and skeletal muscle health. Long-term lack of exercise leads to problems such as slowed body metabolism, obesity, and decreased immune function, increasing the risk of tumors, while moderate exercise (such as weight-bearing exercise helps increase bone density, and aerobic exercise can improve cardiopulmonary function and immunity) has a positive effect on preventing tumors; poor sleep quality (such as long-term insomnia, sleep apnea and other problems) affects the body's endocrine system, immune system and metabolic function, leading to hormone imbalance and increased inflammatory response, thus creating favorable conditions for the occurrence of tumors.

[0020] The preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging is the core tool of the entire study, including a preset segmentation model and a preset diagnosis model. The preset segmentation model is mainly used to automatically segment the tumor area in the preoperative magnetic resonance imaging sequence images. Its structure is usually based on advanced deep learning algorithms, such as the U-Net architecture. U-Net has a unique encoder-decoder structure. The encoder part gradually extracts the feature information of the image through convolutional layers and pooling layers, gradually reduces the resolution of the image, and increases the number of feature channels to obtain the high-level semantic features of the image; the decoder part gradually restores the feature map to the original image size through upsampling layers and convolutional layers, and fuses it with the feature map of the corresponding level in the encoder to obtain more accurate segmentation results. This structure enables the model to effectively process tumor features of different scales and accurately identify tumor boundaries. The preset diagnosis model judges the benign and malignant nature of the tumor based on the segmentation results and other relevant information (such as physiological parameter information, dynamic risk parameter information, etc.). Its structure is based on advanced architectures such as Transformer networks, and combined with multi-instance learning strategies, it can comprehensively process multimodal data and learn the complex relationship between data, thereby improving the accuracy of diagnosis.

[0021] The training sample set and validation sample set are important data resources for building and evaluating deep learning models. These sample sets usually come from multiple medical institutions and cover a large number of user data diagnosed with benign or malignant pelvic and sacral tumors by pathology, including preoperative magnetic resonance imaging sequence image information, physiological parameter information, serological marker parameter information, microbial community parameter information, and lifestyle information. In the process of constructing the sample set, it is necessary to strictly screen according to the predetermined inclusion and exclusion criteria to ensure the representativeness of the sample and data quality. The inclusion criteria include MRI examination to find a single pelvic and sacral lesion, complete preoperative MRI images (including T1-weighted, T2-weighted, DWI and enhanced T1-weighted images, excluding incomplete enhanced MRI sequences), pathological diagnosis of benign or malignant tumors (classified according to the WHO classification criteria), etc.; the exclusion criteria involve multiple lesions, poor image quality (such as the presence of severe artifacts), postoperative recurrence users, etc. By carefully constructing the sample set, rich, diverse and high-quality data is provided for model training, so that the model can learn the characteristic patterns of different types of tumors and the relationship between related factors.

[0022] S102, pre-processing the preoperative magnetic resonance imaging sequence image information of the target user to generate target MRI sequence image information.

[0023] In one embodiment, the preoperative magnetic resonance imaging sequence image information of the target user is normalized and filled to generate image information of the target scale specification, wherein the image information of the target scale specification is used to characterize the size of the target user's T1-weighted imaging information, the target user's T2-weighted imaging information, and the target user's diffusion-weighted imaging information. The main purpose of the normalization process is to map the pixel values ​​of the target user's T1-weighted imaging, T2-weighted imaging, and diffusion-weighted imaging (DWI) information to a unified numerical range, so that the image data between different scanning sequences and different users are comparable. In MRI images, the range of pixel values ​​varies depending on device settings, scanning sequence parameters, and individual differences between users. For example, different MRI scanners produce different signal intensity values ​​for the same tissue, which interferes with subsequent image feature-based analysis and model training. Through normalization processing, these intensity changes caused by device and individual differences are eliminated, the pathological features in the image are highlighted, and the accuracy and consistency of image analysis are improved. For example, the histogram equalization method adjusts the histogram distribution of the image to make the contrast of the image more uniform and enhance the contrast between different tissues in the image, thereby more clearly showing the difference between the tumor area and the surrounding normal tissues.

[0024] Padding is mainly performed on images with a width or height less than 224. In deep learning models, input images are usually required to have uniform size specifications to ensure that the model can correctly process image data. For smaller images, if they are directly input into the model without padding, information loss or model errors will occur during processing. For example, in image analysis models based on convolutional neural networks (CNNs), the operations of convolutional layers and pooling layers rely on fixed-size receptive fields. If the image sizes are inconsistent, the consistency of the receptive fields will be destroyed, affecting the model's extraction and analysis of image features. Through padding, the image is expanded to a size that meets the model input requirements while maintaining the integrity of the image content and avoiding errors introduced by size mismatch.

[0025] There are many ways to fill in, such as boundary filling (copying the pixel values ​​of the image boundary), symmetric filling (symmetrically copying the pixel values ​​with the image boundary as the symmetry axis) or mean filling (filling with the average pixel value of the image). When choosing a filling method, it is necessary to consider the impact on image features. Boundary filling will introduce some unreal information at the edge of the image. If the edge area of ​​the image itself has little effect on the analysis, boundary filling is a simple and effective method. Symmetric filling maintains the symmetry of the image to a certain extent, and is more suitable for some pelvic and sacral regions with symmetrical structural characteristics. Mean filling is relatively smooth and will not introduce obvious boundary effects, but it will slightly blur the edge of the image. For example, if the tumor is mainly located in the central area of ​​the pelvis and sacrum, boundary filling has little effect on the analysis results; if the tumor is close to the edge or the edge features need to be accurately analyzed, symmetric filling or more complex filling strategies are more appropriate. The image size after filling reaches the target scale specification (such as 224×224 or other sizes suitable for model input). At this time, the image information of the target scale specification generated can accurately represent the size of T1-weighted imaging, T2-weighted imaging and DWI imaging information, laying the foundation for subsequent image analysis and model processing.

[0026] The image information of the target scale is processed to generate a three-channel image, wherein the three-channel image is composed of the target user's T1-weighted imaging information, the target user's T2-weighted imaging information, and the target user's diffusion-weighted imaging information. The generation of the three-channel image is to integrate the target scale T1-weighted imaging, T2-weighted imaging, and DWI imaging information after normalization and padding. Each channel corresponds to the information of one imaging method. The purpose of this is to simultaneously include the features of multiple MRI imaging modalities in one image, so that the model can obtain information in the image from multiple angles, thereby more comprehensively and accurately identifying and analyzing the characteristics of pelvic and sacral tumors. For example, the T1-weighted imaging channel provides anatomical information of bones and soft tissues, the T2-weighted imaging channel is more sensitive to edema and inflammation of lesion tissues, and the DWI imaging channel can reflect the diffusion movement of water molecules in the tissue, which helps to judge the activity of tumor tissue. By integrating these three types of imaging information into a three-channel image, the model comprehensively utilizes these complementary information to improve the detection and diagnosis capabilities of tumors.

[0027] In the analysis of pelvic and sacral tumors, three-channel images have obvious advantages. Compared with single-channel images, they can provide richer information. For example, it is difficult for a single T1-weighted image to accurately distinguish the subtle differences between tumor tissue and surrounding tissue, especially when the tumor boundary is unclear or there are multiple tissue types mixed. After combining T2-weighted and DWI imaging information, the model more accurately outlines the tumor boundary by comparing the signal intensity, texture characteristics, and water molecule diffusion of the tumor area in different channels, determines the nature of the tumor (such as benign or malignant), evaluates the degree of tumor invasion (such as whether it invades peripheral nerves, blood vessels and other structures), and monitors the treatment effect of the tumor (such as the changes in tumor tissue after treatment in different imaging channels). In addition, three-channel images also enhance the model's ability to learn image features. When processing multi-channel images, the deep learning model automatically learns the correlation and complementarity between different channels and extracts more representative feature vectors, thereby improving the diagnostic accuracy and generalization ability of the model. For example, the model learns that areas that appear as low signals in T1-weighted imaging, high signals in T2-weighted imaging, and areas with restricted diffusion of water molecules in DWI imaging are more likely to be tumor tissues. This feature learning based on multi-channel information can more accurately identify tumor characteristics and reduce misjudgment.

[0028] The target MRI sequence image information generated based on three-channel images is the final image data form obtained after a series of preprocessing steps, which will serve as the input data of subsequent deep learning models (such as preset segmentation models and preset diagnostic models). These preprocessing steps include normalization, padding, and three-channel image generation, which aim to convert the original preoperative MRI sequence image information into a format more suitable for model processing while retaining the key information in the image. The target MRI sequence image information contains rich tumor-related features, which can provide comprehensive image data support for the model, helping the model to accurately identify tumor areas, analyze tumor features, and perform tasks such as benign and malignant tumor judgment. The generated target MRI sequence image information is highly compatible with the input requirements of the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging. The model is usually designed to receive image inputs of specific sizes, formats, and data types, and the preprocessed target MRI sequence image information just meets these requirements. For example, the model requires the input image size to be 224×224×3 (width×height×number of channels), and the data type to be a normalized value (such as float32). Through the previous normalization, padding, and three-channel image generation operations, the image data input to the model is ensured to be consistent with the model's expectations in terms of size, format, and numerical range, so that the model can effectively process image data, extract valuable feature information, and then make accurate predictions and diagnoses. This adaptability is a key link to ensure the normal operation and accurate output of the entire deep learning model system, which can improve the model's training efficiency and diagnostic accuracy, and provide a reliable basis for the clinical diagnosis and treatment of pelvic and sacral tumors.

[0029] In the diagnosis and analysis of pelvic and sacral tumors, it is crucial to accurately and effectively process the preoperative magnetic resonance imaging (MRI) sequence image information. MRI images contain rich anatomical and pathological information, but due to factors such as imaging equipment, scanning parameters, and individual differences among users, the original image data has differences in size, intensity distribution, etc., which will bring difficulties to subsequent image analysis and model processing. Through normalization and filling processing and three-channel image generation and other operations, the image data is converted into a form that is more suitable for computer processing and model input, improving the consistency and recognizability of image features, thereby providing strong support for accurate diagnosis of tumors.

[0030] S103, processing the serological marker parameter information and the microbial community parameter information of the target user to generate factors affecting the tumor status of the pelvis and sacrum.

[0031] In one embodiment, the target user's serological marker parameter information is subjected to feature extraction processing to generate target serological marker feature information, wherein the target serological marker feature information includes bone alkaline phosphatase (BALP) and lactate dehydrogenase (LDH). BALP plays a key role in bone diseases, especially bone tumors. Its level change is closely related to osteoblast activity. In many bone tumors (such as osteosarcoma, osteoblastoma, etc.), the proliferation and activity of tumor cells stimulate the increase of osteoblast activity, thereby leading to an increase in BALP level. BALP can be used as an important indicator for evaluating the state of tumor bone metabolism and tumor activity. LDH is an enzyme widely present in human tissues. In tumor cells, due to abnormally active metabolism, the production and release of LDH are significantly increased, resulting in an increase in serum LDH levels. In pelvic and sacral tumors (especially malignant tumors such as Ewing sarcoma, etc.), LDH elevation is more common, and its level is closely related to the tumor load, proliferation activity and prognosis, and can be used to monitor tumor progression and treatment response.

[0032] For the BALP and LDH data obtained by the test, feature extraction is required for subsequent analysis. By calculating their absolute values, change rates, and the degree of deviation from the normal reference value range, etc. For example, the concentration change rate of BALP and LDH at different time points is calculated to reflect the dynamic change process of the tumor. If the concentration of BALP or LDH gradually increases over time, it indicates that the tumor is in the progressive stage; conversely, if the concentration decreases, it means that the treatment is effective or the tumor is in remission. The correlation between BALP and LDH and other clinical indicators (such as certain indicators in blood routine, liver and kidney function indicators, etc.) is also analyzed, and the correlation coefficient is extracted as a feature to more comprehensively reflect the relationship between the user's physiological state and the tumor. The extracted features are quantified, and the levels of BALP and LDH are divided into different levels, such as low, medium, and high risk levels, so as to intuitively reflect the degree of tumor risk. At the same time, the characteristics of serological markers are comprehensively quantitatively evaluated in combination with other clinical information of the user (such as age, gender, tumor type, etc.) to improve the accuracy and clinical practicality of the feature information. For example, a higher risk weight is assigned to users who are older and have higher BALP levels, because age itself is also a risk factor for tumor development, and together with increased BALP levels, it indicates a higher tumor risk.

[0033] The microbial community parameter information is processed by feature extraction to generate the gene expression activity feature information of the microbial community and the composition structure feature information of the microbial community. In the pelvic and sacral regions, the microbial community can affect the tumor microenvironment in many ways. For example, some microorganisms can produce toxins or metabolites, which directly damage cell DNA and promote the proliferation and transformation of tumor cells; the microbial community also regulates the host immune system and inhibits the immune surveillance function, thereby creating favorable conditions for the growth and escape of tumor cells. In addition, changes in the composition and function of the microbial community affect the metabolic environment of the tumor and promote the survival and proliferation of tumor cells. High-throughput sequencing technology, such as 16SrRNA sequencing and metagenomic sequencing, is usually used to obtain microbial community parameter information. 16SrRNA sequencing is mainly used to analyze the composition structure of the microbial community. By determining the 16SrRNA gene sequence, different microbial species are identified and their relative abundance is determined. Metagenomic sequencing can more comprehensively reveal the genetic composition and functional potential of the microbial community, not only identifying known microorganisms, but also discovering new microbial species and gene functions. These technologies can provide rich data and conduct in-depth analysis of the composition structure of microbial communities (including the relative abundance of different microbial species, diversity index, etc.), gene expression activity (reflecting the functional status of microorganisms), and metabolic pathway activity (revealing the role of microorganisms in metabolic processes), providing strong support for studying the relationship between microbial communities and tumors.

[0034] The characteristic information of target serological markers, the characteristic information of gene expression activity of microbial communities, and the characteristic information of composition and structure of microbial communities were fused to generate a comprehensive feature vector. The characteristic information of serological markers (such as BALP and LDH related features), the characteristic information of gene expression activity of microbial communities, and the characteristic information of composition and structure each reflect the factors related to pelvic and sacral tumors from different perspectives. However, these information cannot fully and accurately evaluate the tumor status when used alone. Through feature fusion processing, these different sources of information are integrated together, so that the model can learn the synergy and comprehensive influence between them, thereby more comprehensively characterizing the complex biological characteristics of tumors. For example, increased BALP levels indicate enhanced osteogenic activity of bone tumors, but combined with certain gene expression activity features related to immune regulation in the microbial community, the association between changes in immune status in the tumor microenvironment and abnormal bone metabolism is revealed. This comprehensive information is more critical for accurately judging the tumor status. PCA is a dimensionality reduction technology that projects high-dimensional features into low-dimensional space by linearly transforming the original features, while retaining the variance information of the data as much as possible. In this process, feature fusion and dimensionality reduction are achieved, data redundancy is reduced, and model processing efficiency is improved.

[0035] Assume that the serological marker feature vector is (in Indicates BALP-related features, represents LDH-related features), the gene expression activity feature vector of the microbial community is ( represents the expression activity characteristics of the i-th gene), the composition structure characteristic vector of the microbial community is ( represents the relative abundance and other structural characteristics of the jth microorganism). First, determine the weight vector , which correspond to the weights of serological marker characteristics, gene expression activity characteristics and composition structure characteristics respectively. The comprehensive feature vector F is calculated by the following formula: .

[0036] The comprehensive feature vector is processed to generate feature impact correlation information, and the tumor status influencing factors of the pelvis and sacrum are generated based on the feature impact correlation information. To process the fused comprehensive feature vector to generate feature impact correlation information, data analysis and modeling methods are required, for example, prediction models are constructed based on machine learning algorithms, such as logistic regression, decision trees, neural networks, etc. The model learns the complex nonlinear relationship between the comprehensive feature vector and the tumor status (such as benign and malignant, tumor stage, etc.), thereby revealing the correlation between features and their influence on the tumor status. For example, the logistic regression model establishes a mathematical relationship between the feature and the probability of tumor occurrence, analyzes the contribution of each feature in predicting the tumor status (expressed as a regression coefficient), and then determines the feature impact correlation. The decision tree model intuitively shows how the feature affects the classification of the tumor status through different branch decisions. Each node represents a feature, and the branch represents the different value ranges of the feature. By analyzing the structure of the tree and the node splitting rules, the interaction between the features and the order of influence on the tumor status can be understood.

[0037] In the logistic regression model, a positive regression coefficient indicates that the feature is positively correlated with tumor occurrence or progression, and the larger the absolute value of the coefficient, the stronger its influence; a negative coefficient indicates a negative correlation. In the decision tree model, the hierarchical position and branching order of the feature in the tree reflect its importance and order of tumor status classification, and the features close to the root node usually have a greater impact on the classification results. This correlation information helps researchers to deeply understand how serological markers and microbial community characteristics work together on tumor status. For example, it was found that there is a synergistic effect between BALP and specific microbial community composition characteristics. When the BALP level increases and the relative abundance of a certain beneficial microorganism decreases, the risk of tumor occurrence increases significantly. The tumor status influencing factor of the pelvis and sacrum is a comprehensive indicator used to quantify the overall impact of serological markers and microbial community-related characteristics on tumor status. It integrates the feature impact correlation information obtained from the previous analysis and converts multiple complex feature relationships into a single value or vector to more intuitively evaluate the risk of tumor status. This factor serves as an important reference for judging tumor occurrence, development, treatment response and prognosis.

[0038] S104, processing the target user's living habit information to generate pelvic and sacral tumor risk adjustment factors.

[0039] In one embodiment, feature extraction is performed on the target user's living habit information to generate dietary intake features, exercise type features, exercise intensity features, and sleep quality features. The impact of the balance of nutrient intake in the diet on the risk of pelvic and sacral tumors is analyzed. For example, insufficient intake of calcium and vitamin D affects bone health and is related to the occurrence and development of certain bone tumors; while the lack of antioxidant nutrients such as vitamin C, vitamin E, and selenium leads to increased oxidative stress, damages cell DNA, and increases the risk of tumors. Weight-bearing exercise helps increase bone density, enhances bone strength and stability, and has a positive effect on preventing osteoporosis and related tumors (such as giant cell tumors of bone) in the pelvis and sacrum. Aerobic exercise improves cardiopulmonary function, promotes blood circulation, enhances immune system function, and helps reduce the risk of tumors. Flexibility exercises help maintain joint mobility, reduce muscle tension, improve the overall flexibility of the body, and help prevent pelvic and sacral problems caused by poor posture or musculoskeletal imbalance. When extracting features, record the target user's participation in different types of exercise, including information such as sports events, exercise frequency, and duration. Long-term poor sleep quality leads to hormone imbalance, affecting the normal secretion of growth hormone. Growth hormone plays an important role in bone growth and repair, and its abnormality is related to the occurrence of pelvic and sacral tumors. At the same time, insufficient sleep or poor sleep quality will weaken the immune system, reduce the body's immune surveillance and clearance of tumor cells, and increase the risk of tumors.

[0040] The dietary intake characteristics, exercise type characteristics, exercise intensity characteristics and sleep quality characteristics are processed to generate dietary risk values, exercise risk values, sleep quality values ​​and dynamic feature weight values, where the dynamic feature weight values ​​are generated based on the living habits information of different target users in a preset time period. The calculated dietary risk values ​​are graded, such as low risk, medium risk and high risk levels, so as to intuitively evaluate the degree of influence of diet on the risk of pelvic and sacral tumors. In the grading process, individual differences such as age, gender, and basic health conditions are considered to affect dietary risk. For example, due to the decline in digestive function, the elderly have different tolerance and nutrient absorption capacity for certain foods, and the risk under the same dietary pattern is higher than that of young people; women's needs for calcium and vitamin D change during specific physiological periods (such as menopause), which affects dietary risk assessment. According to these individual differences, the dietary risk value is appropriately adjusted so that the assessment results more accurately reflect the actual risk status of different individuals.

[0041] Consider the impact of individual factors on the exercise risk value, such as age, gender, physical condition (whether there are bone and joint diseases, cardiovascular diseases, etc.) and exercise habit history. The elderly are at relatively high risk when doing high-intensity exercise, and the risk score related to exercise intensity should be adjusted according to their age; women's exercise risk changes during special periods such as menstruation or pregnancy, and the assessment model needs to be adjusted accordingly; for users who already have bone and joint diseases (such as osteoporosis, arthritis, etc.) or cardiovascular diseases, certain types of exercise or intensity are not suitable. These limiting factors need to be considered when calculating the exercise risk value to avoid aggravation of the disease or increased tumor risk due to exercise. By comprehensively considering individual factors, the exercise risk value can more accurately reflect the actual situation of each target user and provide a basis for personalized exercise recommendations. Consider the impact of sleep environment and habit factors on sleep quality value and tumor risk. Noise, light, temperature and other bad sleep environment factors are given a certain risk score according to their degree of interference. Users who are in a bad sleep environment for a long time will have a correspondingly reduced sleep quality value and an increased tumor risk. Bad sleeping habits, such as using electronic devices before bedtime and irregular work and rest schedules, can also affect sleep quality. By investigating and understanding the sleep habits of target users, we make appropriate adjustments in the sleep quality value calculation for users with bad habits, in order to more comprehensively assess the relationship between sleep quality and the risk of pelvic and sacral tumors.

[0042] The calculated dynamic feature weight values ​​are applied to the generation of pelvic and sacral tumor risk adjustment factors. The dietary risk value, exercise risk value, and sleep quality value are multiplied by the corresponding dynamic feature weight values, and then weighted summed to obtain the final tumor risk adjustment factor. In this way, the risk adjustment factor can more accurately reflect the comprehensive impact of different lifestyle characteristics on tumor risk at different time points. For example, if a user's recent exercise risk value increases due to changes in exercise habits, and the dynamic weight value of the exercise feature is high, then when calculating the risk adjustment factor, the exercise risk value will contribute more to the overall risk, thereby highlighting the impact of recent changes in exercise habits on the risk of pelvic and sacral tumors, and providing a more targeted basis for personalized tumor risk assessment and intervention.

[0043] The dietary risk value, exercise risk value, sleep quality value and dynamic feature weight value are processed to generate the tumor risk adjustment factor of the pelvis and sacrum. The tumor risk adjustment factor of the pelvis and sacrum is calculated by integrating the dietary risk value, exercise risk value, sleep quality value and their corresponding dynamic feature weight values. Assuming that the dietary risk value is The sports risk value is , the sleep quality value is , the dynamic characteristic weight value of dietary risk is , the dynamic characteristic weight value of motion risk is , the dynamic characteristic weight value of sleep quality risk is , then the calculation formula of the tumor risk adjustment factor is expressed as: This formula integrates the risk information of different lifestyle factors, where the weight value reflects the relative importance of each factor to cancer risk in a specific time period. For example, if the impact of changes in exercise habits on cancer risk is considered more critical in a certain period of time (i.e. larger), then the exercise risk value It will carry a greater weight when calculating the risk adjustment factor.

[0044] S105, processing the tumor status influencing factors of the pelvis and sacrum and the tumor risk adjustment factors of the pelvis and sacrum to generate dynamic risk parameter information of the target user.

[0045] In one embodiment, the tumor status influencing factors of the pelvis and sacrum and the tumor risk adjustment factors of the pelvis and sacrum are processed to generate a tumor morphology change value within a preset time interval information. The present application includes a calculation formula for obtaining the tumor morphology change value, which is: .in, Represents the value of tumor morphology change, represents the volume of the tumor at time t, represents the initial tumor volume, represents the surface area of ​​the tumor at time t, , represents the morphological irregularity of the tumor at time t, Represents the irregularity of the initial tumor morphology, are the weight coefficients of volume change, surface area change and morphological irregularity change, respectively.

[0046] Assume that a user's initial tumor volume , tumor volume at the end of the first month , tumor volume at the end of the second month , tumor volume at the end of the third month The rate of change in tumor size during the first visit period (0-1 month) was The rate of change in tumor size during the second visit period (1-2 months) was The rate of change in tumor size at the third visit cycle (2-3 months) was By calculating the rate of change of tumor size in different access cycles, it was found that the growth rate of the user's tumor accelerated in the second access cycle, and then slowed down slightly in the third access cycle, but it was still growing overall.

[0047] In the first visit cycle, medical imaging analysis (such as MRI images) revealed that the tumor shape was approximately spherical and the surface was relatively smooth. By the second visit cycle, the tumor began to have a slight bulge on one side, the overall shape changed to an ellipse, and the surface was no longer so smooth, with some small depressions. In the third visit cycle, the tumor further deformed, the bulge became more obvious, forming an irregular shape similar to a dumbbell, and the depressions on the surface increased and deepened. In order to quantify these changes, shape descriptors such as circularity are used (assuming that the initial circularity is 0.9, it becomes 0.75 in the second cycle, and drops to 0.6 in the third cycle). The continuous decrease in circularity indicates that the tumor shape gradually deviates from a circle and becomes more complex and irregular, which means that the proliferation rate of tumor cells in local areas is inconsistent, the invasiveness is enhanced, and it is easier to infiltrate surrounding tissues.

[0048] In the first visit cycle, the tumor boundary was clear, the edges were neat, and the boundary with the surrounding tissue was obvious. In the second visit cycle, the tumor boundary began to become blurred in some areas, and some small burr-like protrusions appeared, indicating that tumor cells began to infiltrate the surrounding tissue. In the third visit cycle, the blurred boundary area expanded, the burrs increased and became longer, and the boundary between some areas and the surrounding tissue was difficult to clearly define. The fuzziness index of the boundary was calculated (assuming that the initial boundary fuzziness was 0.15, it became 0.3 in the second cycle, and rose to 0.45 in the third cycle). The increase in boundary fuzziness indicated that the stability of tumor boundary changes decreased, the infiltration of tumor cells into surrounding tissues increased, the difficulty of complete tumor removal during surgery increased, and the risk of tumor recurrence and metastasis increased.

[0049] In another embodiment, the initial tumor surface area , initial tumor morphology irregularity After 3 months of observation, at time t, the tumor volume was measured. , tumor surface area , tumor morphology irregularity At the same time, the weight coefficient is determined based on a large amount of clinical data statistical analysis. . First calculate the volume change rate: Surface area change rate: ; Change rate of morphological irregularity: The tumor morphology change value This value reflects the comprehensive change in the tumor morphology of the user within these 3 months. A higher value This suggests that the tumor has obvious changes in volume, surface area, and morphological irregularity, and requires closer monitoring.

[0050] Changes in tumor volume are directly related to the growth or shrinkage of the tumor. In many pelvic and sacral tumors, rapid volume growth means that tumor cells proliferate actively and are more invasive. For example, malignant tumors (such as osteosarcoma, Ewing sarcoma, etc.) usually have a faster growth rate, while benign tumors (such as osteoid osteoma, osteochondroma, etc.) grow relatively slowly. By monitoring the volume change rate, the growth trend of the tumor can be discovered early and the treatment strategy can be adjusted in time. For example, more active treatment interventions (such as surgery, chemotherapy, radiotherapy, etc.) are required for rapidly growing tumors. Changes in tumor surface area reflect changes in tumor morphology. When the tumor surface becomes more irregular or diffuse, it indicates that the infiltration of tumor cells into surrounding tissues increases, which is of great significance for determining the stage and prognosis of the tumor. For example, in sacral tumors, if the tumor surface area increases rapidly and the morphology becomes complex, it will increase the difficulty of surgical resection and also indicate a poor prognosis for the user. Changes in morphological irregularity are closely related to the malignancy of the tumor. The more irregular the morphology, the higher the heterogeneity of the tumor cells and the more aggressive their biological behavior. For example, during the development of some highly malignant pelvic tumors, the morphological irregularity will gradually increase, showing features such as blurred boundaries, lobed or burr-like shapes, which makes the tumor more likely to invade surrounding important structures such as nerves and blood vessels.

[0051] For the tumor morphology change values ​​within the preset time interval information, characteristic parameter information of different access cycles and dynamic pelvic and sacral structural information are generated, among which the characteristic parameter information of different access cycles includes the rate of change of tumor size, the characteristics of tumor shape change, and the stability characteristics of tumor boundary change. At the initial examination, the bone structure of the pelvis and sacrum was intact, and no obvious bone destruction was found. The surrounding soft tissue morphology was normal, and the nerves and blood vessels were clear, without compression or displacement. The MRI images of the second access cycle showed that there was slight bone absorption in the contact area between the tumor and the sacral bone, which was manifested as a slight decrease in bone density, indicating that the tumor began to affect the bone. In terms of soft tissue, the muscle tissue around the tumor showed mild edema, and the nerves and blood vessels were not obviously compressed, but the vascular blood flow signal was slightly changed, which was a hemodynamic change caused by local metabolic changes of the tumor. In the third access cycle, the scope of sacral bone destruction expanded, the bone cortex became thinner, and small bone defects appeared locally. The nerves were compressed and deformed, the space around the nerves narrowed, and the signal intensity of some nerve fibers changed, indicating that the nerve function was affected. The blood vessels are compressed and displaced significantly, the blood vessel diameter becomes thinner, and the local blood flow rate slows down, which affects the blood supply to the tumor and the nutrient supply to the surrounding tissues, further affecting the growth of the tumor and the function of the surrounding tissues.

[0052] These dynamically changing pelvic and sacral structural information are very important for evaluating the progression of tumors and predicting the prognosis of users. For example, the aggravation of bone destruction affects the stability of the pelvis and increases the risk of fractures; nerve compression and functional changes cause users to experience symptoms such as lower limb pain, numbness, and weakness, which seriously affect the quality of life; blood vessel compression affects the tumor's uptake and treatment effect of chemotherapy drugs, and also affects the metabolism and repair of surrounding normal tissues.

[0053] The characteristic parameter information of different access cycles and the dynamic pelvic and sacral structural information are processed to generate the dynamic risk parameter information of the target user. The dynamic risk parameter information is calculated by weighted summation, assuming that the weight of the tumor size change rate is 0.4, the weight of the tumor shape change feature is 0.3, the weight of the tumor boundary change stability feature is 0.2, and the weight of the pelvic and sacral structure change is 0.1. Taking the third access cycle as an example, the tumor size change rate in this cycle is , shape change characteristics (roundness is 0.6), boundary change stability characteristics (fuzziness is 0.45), pelvic and sacral structure changes (the combined score of bone destruction and soft tissue invasion is assumed to be 0.35). Then the dynamic risk parameter information . The larger the value of the dynamic risk parameter information, the higher the dynamic tumor risk of the user during the access cycle. By regularly calculating (such as calculating once a month or every two months) and monitoring the dynamic risk parameter information, the changing trend of tumor risk can be grasped in real time. If the dynamic risk parameter information continues to rise, you should be highly vigilant, such as increasing the dose of chemotherapy drugs, replacing more effective targeted drugs, or performing surgical intervention in a timely manner; if the dynamic risk parameter information remains relatively stable or decreases, you can continue to maintain treatment and observe closely. At the same time, users are managed in layers according to the dynamic risk parameter information, and the follow-up frequency can be increased for high-risk users (such as changing from once every three months to once a month).

[0054] S106: Process the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging based on the training sample set and the verification sample set to generate a target end-to-end deep learning model.

[0055] In one embodiment, the training sample set is preprocessed to generate a training sample set in a target format, wherein the training sample set in the target format includes three-channel image information of other users, physiological parameter information of other users, and dynamic risk parameter information of other users; the collected training sample set is cleaned to remove data records with errors or missing values. For example, if it is found that the age data in the physiological parameter information of a user is missing, or the serological marker test results are obviously abnormal and the accuracy cannot be verified, the sample is removed. For image data, check whether there are problems such as image damage and low resolution to ensure that the image quality meets the requirements. Then, all types of data are converted into a format suitable for model processing. For the user's physiological parameter information (such as gender, age, tumor location, maximum tumor size, etc.), it is numerically encoded. For example, 0 represents female and 1 represents male; age is represented by actual age value; tumor location is encoded according to a predefined anatomical coding system; the maximum tumor size is uniformly converted into a value in cubic centimeters. For serological marker parameter information (such as bone alkaline phosphatase and lactate dehydrogenase, etc.), the concentration values ​​obtained by detection are standardized so that their numerical range is within a reasonable range, which is convenient for subsequent calculation and comparison. Microbial community parameter information (such as microbial community composition structure and gene expression activity characteristics obtained by high-throughput sequencing, etc.) is converted into vector form after data analysis and feature extraction. Lifestyle information (such as dietary intake, exercise type, exercise intensity, sleep quality, etc.) is also quantified.

[0056] For example, dietary intake is converted into nutrient intake values ​​according to food types and intake amounts; exercise types are assigned corresponding values ​​according to the intensity of different exercises and the degree of impact on the body; exercise intensity is calculated by comprehensive calculation of exercise duration, exercise frequency and exercise intensity indicators (such as heart rate changes, energy consumption, etc.) to obtain a quantitative value; sleep quality is comprehensively evaluated based on factors such as sleep duration, sleep efficiency, sleep stage distribution and sleep disorders to obtain a quantitative score. Finally, the processed user data is combined into a training sample set in the target format according to the preset rules, ensuring that each sample contains the three-channel image information of other users (given by Weighted imaging, weighted imaging and diffusion-weighted imaging information), physiological parameter information and dynamic risk parameter information.

[0057] In order to avoid the deviation caused by sample imbalance during model training, the training sample set is processed by sample balancing. The number of samples of different categories (such as benign tumors and malignant tumors) is counted. If the number of samples of a certain category is too small, the oversampling technology (such as SMOTE algorithm) is used to increase the number of samples of this category, so that the number of samples of each category is relatively balanced. At the same time, in order to improve the generalization ability of the model, the training samples are enhanced. For image data, random rotation, flipping, scaling and other operations are used to increase the diversity of images. For example, the pelvic and sacral MRI images of a user are randomly rotated 10-20 degrees, or flipped horizontally or vertically, while keeping the tumor feature information in the image unchanged, so as to generate new training samples. For physiological parameter information and dynamic risk parameter information, random noise within a certain range can be added to simulate the change of data, so that the model can learn more robust feature patterns. The training sample set processed in this way can better represent the actual data distribution and help train a deep learning model with better performance.

[0058] The preset segmentation model is processed based on the three-channel image information of other users to generate the segmentation result of the preset tumor lesion area. The preset segmentation model adopts the U-Net structure, and its encoder part extracts features from the input three-channel image information through a series of convolutional layers and pooling layers. The convolutional layer is used to extract local features in the image. For example, the edge details of the tumor area can be detected through a 3×3 convolution kernel, and the texture features of a wider area can be captured by a 5×5 convolution kernel. The pooling layer gradually reduces the image resolution in a 2×2 maximum pooling manner, while expanding the receptive field, so that the model can obtain more global image feature information. In each layer of the encoder, the number of feature channels will gradually increase, so that more advanced and abstract image semantic features can be learned. For example, after the three-channel image input by the first layer encoder is convolved, 16 feature channels are generated, and the number of feature channels in subsequent layers becomes 32, 64, 128, etc. The decoder part gradually restores the low-resolution feature map to the original image size through upsampling layers and convolution layers. The upsampling layer uses a bilinear interpolation method to increase the image resolution, and at the same time, the feature map of the corresponding level in the encoder is fused with the feature map in the decoder through a jump connection, so that the model can better utilize feature information of different scales and accurately identify the boundaries of tumor lesions. For example, in a certain layer of the decoder, the upsampled feature map is first spliced ​​with the feature map of the corresponding resolution in the encoder, and then the features are further refined through a 3×3 convolution layer, gradually restoring the accurate segmentation results of the tumor area.

[0059] Assume that a user's pelvic and sacral MRI images are processed, and the segmentation results of the tumor lesion area are obtained after the segmentation model. In the form of image visualization, the tumor area is marked on the original MRI image, for example, the general shape and range of the tumor are clearly outlined with a red outline. At the same time, the accuracy of the segmentation results is quantified by calculating evaluation indicators such as the Dice coefficient and the intersection over union (IoU). If the actual tumor area is 100 square pixels, the intersection of the tumor area segmented by the model and the actual area is 80 square pixels, and the union is 120 square pixels, then the Dice coefficient is 2×80 / (100+120)≈0.727, and the IoU is 80 / 120≈0.667. Generally speaking, the higher the Dice coefficient and IoU value, the more accurate the segmentation result and the stronger the model's ability to recognize the tumor area.

[0060] The preset diagnostic model is processed based on the segmentation results of the preset tumor lesion area, the physiological parameter information of other users, and the dynamic risk parameter information of other users to generate a training result. The preset segmentation model and the preset diagnostic model are processed based on the verification sample set to generate a verification result. If the identification information contained in the verification result and the identification information contained in the training result are risk factors that characterize sacral lesions, the trained preset segmentation model and the preset diagnostic model are used as the target end-to-end deep learning model. The verification sample set also undergoes similar preprocessing steps as the training sample set to ensure that the data format and quality meet the requirements. Then, the three-channel image information of the verification sample set is input into the preset segmentation model to obtain the segmentation result of the tumor lesion area; the segmentation result is input into the preset diagnostic model together with the physiological parameter information and dynamic risk parameter information in the verification sample set to obtain the prediction result. For example, for a user in the verification set, the segmentation model accurately outlines the boundary of the tumor, and the diagnostic model predicts that the probability of the tumor being benign is 0.7 and the probability of being malignant is 0.3. According to the threshold, it is judged as a user with a benign tumor.

[0061] Calculate the performance indicators such as accuracy, AUC (area under the receiver operating characteristic curve), sensitivity, and specificity in the validation results. Assuming that there are 80 users in the validation set, there are actually 30 users with benign tumors and 50 users with malignant tumors, and the model correctly predicts 25 users with benign tumors and 40 users with malignant tumors, the accuracy is (25+40) / 80=0.8125. Calculate the AUC value by drawing the ROC curve. If the AUC is 0.85, it means that the model has good performance in distinguishing benign and malignant tumors. The sensitivity is 40 / 50=0.8, and the specificity is 25 / 30≈0.83. If these indicators in the validation results meet the expected requirements, and the identification information contained in the validation results (such as the benign and malignant judgment results of the tumor, the risk level classification, etc.) is consistent with the identification information contained in the training results in characterizing the risk factors affecting sacral lesions, that is, the performance of the model on the validation set and the training set is stable and meets expectations, then the trained preset segmentation model and preset diagnosis model are used as the target end-to-end deep learning model. If the verification result is not ideal, for example, the accuracy is lower than the expected threshold, or there is overfitting (such as high accuracy on the training set but low accuracy on the verification set) or underfitting (such as low accuracy on both the training set and the verification set) on the verification set, the model needs to be adjusted, such as adjusting the model's hyperparameters (learning rate, number of iterations, network structure, etc.), optimizing data augmentation methods, increasing the amount of training data, etc., and then retraining and revalidating until a satisfactory target end-to-end deep learning model is obtained.

[0062] S107, processing the target MRI sequence image information, the physiological parameter information of the target user, and the dynamic risk parameter information of the target user based on the target end-to-end deep learning model to generate user warning information.

[0063] In one embodiment, the physiological parameter information of the target user is subjected to feature extraction processing to generate the physiological characteristic information of the target user, wherein the physiological characteristic information of the target user includes the target user's gender characteristic information, age characteristic information, tumor location characteristic information and maximum tumor size characteristic information. For the target user's physiological parameter information, feature extraction processing is first performed. Assume that the target user is a 55-year-old male, whose tumor is located on the right side of the sacrum, and the maximum tumor diameter is 5 cm measured by imaging (the volume is calculated to be approximately 65.45 cubic centimeters). The gender characteristic information is encoded as 1 (indicating male), and the age characteristic information directly uses the actual age of 55 years old as the value. The tumor location characteristic information divides the sacrum into several regions according to a pre-set anatomical coding system, such as the right sacrum is encoded as 3. The maximum tumor size characteristic information is in cubic centimeters, that is, 65.45. In this way, the target user's physiological characteristic information vector [1,55,3,65.45] is generated, and these features will serve as important inputs for subsequent model processing.

[0064] The target MRI sequence image information is processed based on the target segmentation model to generate the target user's tumor lesion area segmentation result. The target user's three-channel MRI sequence image information is processed based on the target segmentation model (such as a model using the U-Net architecture). The encoder part of the model gradually extracts image features through convolutional layers and pooling layers. For example, in the convolutional layer, convolution kernels of different sizes (such as 3×3, 5×5, etc.) are used to detect the edges, textures and other detailed features of the tumor area. The pooling layer downsamples in a 2×2 manner to reduce the image resolution while expanding the receptive field to obtain more global image feature information. The decoder part restores the low-resolution feature map to the original image size through the upsampling layer and the convolution layer, and fuses the feature maps of different levels through jump connections to accurately identify the boundaries of the tumor lesion area.

[0065] After being processed by the segmentation model, the segmentation result of the tumor lesion area of ​​the target user is obtained. In terms of image visualization, the tumor area is clearly marked on the original MRI image, for example, the shape and range of the tumor are outlined with a green contour line. Assuming that the original MRI image size is 512×512 pixels, the position, shape, size and other information of the segmented tumor area in the image can be intuitively presented through the segmentation results. At the same time, evaluation indicators are calculated to quantify the accuracy of the segmentation results, such as the Dice coefficient and intersection over union (IoU). If the actual tumor area is 80 square pixels, the intersection of the tumor area segmented by the model and the actual area is 65 square pixels, and the union is 90 square pixels, then the Dice coefficient is , IoU is The higher Dice coefficient and IoU value indicate that the segmentation model is more accurate in identifying the tumor region of the target user.

[0066] Based on the target diagnosis model, the target user's tumor lesion area segmentation results, the target user's physiological characteristics information and the target user's dynamic risk parameter information are processed to generate a target prediction value. The target prediction value is processed based on the preset threshold to generate user warning information, which is used to characterize the attribute information of the target user's preoperative pelvic and sacral tumors. The target diagnosis model receives the tumor lesion area segmentation results, physiological characteristics information and dynamic risk parameter information as input. For the segmentation results, the model will further extract geometric features such as the area, perimeter, shape complexity of the tumor area, and the position characteristics of the tumor area in the image, and convert these features into vector form. Physiological characteristic information [1,55,3,65.45] and dynamic risk parameter information (assuming 0.45, which is calculated based on the previous calculation and represents the dynamic risk level of the current tumor) are also input into the model. The model automatically learns the association and importance weights between these different sources of information through the self-attention mechanism of the Transformer network, so as to comprehensively judge the benign and malignant nature of the tumor.

[0067] Assume that the weights of the tumor lesion area segmentation result features, physiological feature information, dynamic risk parameter information and their interaction features on the prediction results are (These weights are obtained through model training and learning). After model calculation, the target prediction value P is obtained. According to the preset calculation formula (assuming Respectively represent the specific function transformation of segmentation result features, physiological feature information, dynamic risk parameter information and interactive features): Assume that the values ​​after function transformation are is the sigmoid function, then .

[0068] The preset threshold is set to 0.5, because According to the model prediction, the target user's tumor is more likely to be malignant. Therefore, the generated user warning information is "Preoperative pelvic and sacral tumors may be malignant, please further check and evaluate". This warning information can help doctors and users understand the potential risks of tumors in advance so that they can take appropriate treatment measures or adjust treatment plans in time. If the predicted value is less than the preset threshold, the warning information may be "Preoperative pelvic and sacral tumors tend to be benign, but regular review is still required". In this way, based on the processing results of the target end-to-end deep learning model, targeted preoperative tumor attribute warning information is provided to the target user.

[0069] The server collects multi-dimensional information of the target user, including preoperative magnetic resonance imaging sequence images (T1, T2, diffusion-weighted imaging), physiological parameters, serological markers, microbial communities and lifestyle habits, and prepares a preset end-to-end deep learning model and sample set based on non-enhanced MRI. Subsequently, each piece of information is processed in a targeted manner. The imaging sequence images are normalized, padded and converted into three-channel images to generate the target MRI sequence image information; the serological and microbial community information is extracted and fused to obtain the tumor status influencing factors; the features are extracted from the lifestyle information to calculate the relevant risk value and weight value, and the tumor risk adjustment factor is generated. The two are integrated to obtain dynamic risk parameter information.

[0070] Next, the preset model is trained with the preprocessed training sample set, and the segmentation model is trained with three-channel images to obtain the segmentation results of the tumor lesion area. The diagnosis model is trained in combination with other information. After verification by the validation sample set, the model that meets the risk factor identification of sacral lesions becomes the target model. The target model is used to combine multiple types of information to calculate the prediction value, and the user warning information is generated according to the preset threshold.

[0071] In one embodiment, if Figure 2 As shown, the present application also provides a pelvic and sacral early warning device based on a non-enhanced MRI end-to-end deep learning model, comprising: An acquisition module 201 is used to acquire preoperative magnetic resonance imaging sequence image information of a target user, physiological parameter information of the target user, serological marker parameter information of the target user, microbial community parameter information, life habit information of the target user, a preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging, a training sample set, and a verification sample set, wherein the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging includes a preset segmentation model and a preset diagnosis model, and the preoperative magnetic resonance imaging sequence image information of the target user includes T1-weighted imaging information of the target user, T2-weighted imaging information of the target user, and diffusion-weighted imaging information of the target user; The processing module 202 is used to pre-process the preoperative magnetic resonance imaging sequence image information of the target user to generate target MRI sequence image information; process the serological marker parameter information and the microbial community parameter information of the target user to generate the tumor status influencing factors of the pelvis and sacrum; process the living habit information of the target user to generate the tumor risk adjustment factors of the pelvis and sacrum; process the tumor status influencing factors of the pelvis and sacrum and the tumor risk adjustment factors of the pelvis and sacrum to generate the dynamic risk parameter information of the target user; process the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging based on the training sample set and the verification sample set to generate a target end-to-end deep learning model; process the target MRI sequence image information, the physiological parameter information of the target user and the dynamic risk parameter information of the target user based on the target end-to-end deep learning model to generate user warning information, wherein the user warning information is used to characterize the attribute information of the target user's preoperative pelvic and sacrum tumors.

[0072] Each embodiment in the present application is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for evaluating the pelvic and sacral early warning method, electronic device, electronic device, and readable storage medium embodiment based on the non-enhanced MRI end-to-end deep learning model, since it is basically similar to the above-mentioned pelvic and sacral early warning method embodiment based on the non-enhanced MRI end-to-end deep learning model, the description is relatively simple, and the relevant parts can be referred to the partial description of the above-mentioned pelvic and sacral early warning method embodiment based on the non-enhanced MRI end-to-end deep learning model.

Claims

1. A pelvic and sacral early warning method based on a non-enhanced MRI end-to-end deep learning model, characterized in that: include: Obtaining preoperative magnetic resonance imaging sequence image information of the target user, physiological parameter information of the target user, serological marker parameter information of the target user, microbial community parameter information, life habit information of the target user, a preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging, a training sample set, and a verification sample set, wherein the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging includes a preset segmentation model and a preset diagnosis model, and the preoperative magnetic resonance imaging sequence image information of the target user includes T1-weighted imaging information of the target user, T2-weighted imaging information of the target user, and diffusion-weighted imaging information of the target user; Preprocessing the preoperative magnetic resonance imaging sequence image information of the target user to generate target MRI sequence image information; Processing the serological marker parameter information and the microbial community parameter information of the target user to generate factors affecting the tumor status of the pelvis and sacrum; Processing the target user's lifestyle information to generate pelvic and sacral tumor risk adjustment factors; Processing the tumor status influencing factors of the pelvis and sacrum and the tumor risk adjustment factors of the pelvis and sacrum to generate dynamic risk parameter information of the target user; Processing the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging based on the training sample set and the verification sample set to generate a target end-to-end deep learning model; The target MRI sequence image information, the target user's physiological parameter information and the target user's dynamic risk parameter information are processed based on the target end-to-end deep learning model to generate user warning information, wherein the user warning information is used to characterize the attribute information of the target user's preoperative pelvic and sacral tumors.

2. The method according to claim 1, characterized in that Preprocessing the preoperative magnetic resonance imaging sequence image information of the target user to generate target MRI sequence image information includes: Normalizing and filling the preoperative magnetic resonance imaging sequence image information of the target user to generate image information of a target scale, wherein the image information of the target scale is used to characterize the size of the T1-weighted imaging information of the target user, the T2-weighted imaging information of the target user, and the diffusion-weighted imaging information of the target user; Processing the image information of the target scale specification to generate a three-channel image, wherein the three-channel image is composed of T1-weighted imaging information of the target user, T2-weighted imaging information of the target user, and diffusion-weighted imaging information of the target user; Generate target MRI sequence image information based on the three-channel images.

3. The method according to claim 1, characterized in that The serological marker parameter information and the microbial community parameter information of the target user are processed to generate factors affecting the tumor status of the pelvis and sacrum, including: Performing feature extraction processing on the serological marker parameter information of the target user to generate target serological marker feature information, wherein the target serological marker feature information includes bone-derived alkaline phosphatase and lactate dehydrogenase; Performing feature extraction processing on the microbial community parameter information to generate gene expression activity feature information of the microbial community and composition structure feature information of the microbial community; Performing feature fusion processing on the characteristic information of the target serological marker, the characteristic information of the gene expression activity of the microbial community, and the characteristic information of the composition structure of the microbial community to generate a comprehensive feature vector; Processing the comprehensive feature vector to generate feature impact correlation information; The tumor status influencing factors of the pelvis and sacrum are generated based on the feature impact correlation information.

4. The method according to claim 3, characterized in that The target user's lifestyle information is processed to generate pelvic and sacral tumor risk adjustment factors, including: Performing feature extraction processing on the target user's life habit information to generate dietary intake features, exercise type features, exercise intensity features, and sleep quality features; The dietary intake feature, the exercise type feature, the exercise intensity feature and the sleep quality feature are processed to generate a dietary risk value, an exercise risk value, a sleep quality value and a dynamic feature weight value, wherein the dynamic feature weight value is generated based on the life habit information of different target users within a preset time period; The dietary risk value, the exercise risk value, the sleep quality value and the dynamic feature weight value are processed to generate tumor risk adjustment factors for the pelvis and sacrum.

5. The method according to claim 4, characterized in that The tumor status influencing factors of the pelvis and sacrum and the tumor risk adjustment factors of the pelvis and sacrum are processed to generate dynamic risk parameter information of the target user, including: Processing the tumor status influencing factors of the pelvis and sacrum and the tumor risk adjustment factors of the pelvis and sacrum to generate tumor morphology change values ​​within a preset time interval information; Processing the tumor morphology change values ​​within the preset time interval information to generate characteristic parameter information of different access cycles and structural information of the dynamic pelvis and sacrum, wherein the characteristic parameter information of different access cycles includes the change rate of tumor size, the change characteristics of tumor shape, and the stability characteristics of tumor boundary changes; Processing characteristic parameter information of different access cycles and dynamic pelvic and sacral structural information to generate dynamic risk parameter information of target users; The method also includes a calculation formula for obtaining a tumor morphology change value, and the calculation formula is: ； in, Represents the value of tumor morphology change, represents the volume of the tumor at time t, represents the initial tumor volume, represents the surface area of ​​the tumor at time t, , represents the morphological irregularity of the tumor at time t, Represents the irregularity of the initial tumor morphology, are the weight coefficients of volume change, surface area change and morphological irregularity change, respectively.

6. The method according to claim 1, characterized in that The preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging is processed based on the training sample set and the verification sample set to generate a target end-to-end deep learning model, including: Preprocessing the training sample set to generate a training sample set in a target format, wherein the training sample set in the target format includes three-channel image information of other users, physiological parameter information of other users, and dynamic risk parameter information of other users; Processing the preset segmentation model based on the three-channel image information of other users to generate a segmentation result of a preset tumor lesion area; The preset diagnosis model is processed based on the segmentation result of the preset tumor lesion area, the physiological parameter information of other users, and the dynamic risk parameter information of other users to generate a training result; Based on the verification sample set, the preset segmentation model and the preset diagnosis model are processed respectively to generate a verification result; If the identification information contained in the verification result and the identification information contained in the training result characterize the risk factors affecting sacral lesions, the trained preset segmentation model and preset diagnosis model are used as the target end-to-end deep learning model.

7. The method according to claim 5, characterized in that The target MRI sequence image information, the physiological parameter information of the target user, and the dynamic risk parameter information of the target user are processed based on the target end-to-end deep learning model to generate user warning information, including: Performing feature extraction processing on the physiological parameter information of the target user to generate physiological feature information of the target user, wherein the physiological feature information of the target user includes gender feature information, age feature information, tumor location feature information, and maximum tumor size feature information of the target user; Processing the target MRI sequence image information based on the target segmentation model to generate a tumor lesion area segmentation result for the target user; Based on the target diagnosis model, the target user's tumor lesion area segmentation result, the target user's physiological characteristic information and the target user's dynamic risk parameter information are processed to generate a target prediction value; Processing the target prediction value based on a preset threshold value to generate user warning information; The method includes a calculation formula for obtaining a target prediction value, and the calculation formula is: ； in, is the sigmoid function, are weight coefficients, representing the characteristics of the tumor lesion area segmentation results. , physiological characteristics information , Dynamic risk parameter information And the interaction characteristics between them The weight of the impact on the prediction results.

8. A pelvic and sacral early warning device based on a non-enhanced MRI end-to-end deep learning model, characterized in that: The device comprises: An acquisition module, used to acquire preoperative magnetic resonance imaging sequence image information of a target user, physiological parameter information of the target user, serological marker parameter information of the target user, microbial community parameter information, life habit information of the target user, a preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging, a training sample set, and a verification sample set, wherein the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging includes a preset segmentation model and a preset diagnosis model, and the preoperative magnetic resonance imaging sequence image information of the target user includes T1-weighted imaging information of the target user, T2-weighted imaging information of the target user, and diffusion-weighted imaging information of the target user; A processing module is used to pre-process the preoperative magnetic resonance imaging sequence image information of the target user to generate target MRI sequence image information; process the serological marker parameter information and the microbial community parameter information of the target user to generate the tumor status influencing factors of the pelvis and sacrum; process the living habit information of the target user to generate the tumor risk adjustment factors of the pelvis and sacrum; process the tumor status influencing factors of the pelvis and sacrum and the tumor risk adjustment factors of the pelvis and sacrum to generate the dynamic risk parameter information of the target user; process the preset end-to-end deep learning model based on non-enhanced magnetic resonance imaging based on the training sample set and the verification sample set to generate a target end-to-end deep learning model; process the target MRI sequence image information, the physiological parameter information of the target user and the dynamic risk parameter information of the target user based on the target end-to-end deep learning model to generate user warning information, wherein the user warning information is used to characterize the attribute information of the preoperative pelvic and sacrum tumors of the target user.

9. An electronic device, characterized in that: include: a first processor; and a memory for storing executable instructions of the first processor; Wherein, the first processor is configured to execute the pelvis and sacrum early warning method based on the non-enhanced MRI end-to-end deep learning model as described in any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the second processor, the pelvis and sacrum early warning method based on the non-enhanced MRI end-to-end deep learning model described in any one of claims 1 to 7 is implemented.

Images