AI prediction system for relapse after radiotherapy of local advanced rectal cancer

Abstract

The invention relates to the technical field of clinical medicine, and discloses an AI prediction system for relapse after radiotherapy of local advanced rectal cancer, and the system comprises a multi-modal biological image database building module which is used for collecting clinical data, image data, radiotherapy target region and dose distribution data, radiobiological data and genomics information of a patient, integrating the data to establish a database; a feature extraction and screening module; a prediction model construction module; a model verification and optimization module; and a clinical application evaluation module. According to the invention, information of patients in multiple aspects such as clinic, image, radiotherapy target area and dose distribution, radiobiology and genomics is widely collected, and the multi-modal biological image database is integrated and established, so that various factors influencing rectum cancer relapse after radiotherapy can be comprehensively covered; a rich, accurate and comprehensive data basis is provided for subsequent accurate analysis and prediction, and the accuracy and reliability of prediction of relapse conditions after radiotherapy are improved.

Description

Technical Field

[0001] The present invention relates to the field of clinical medical technology, and in particular to an AI prediction system for recurrence of locally advanced rectal cancer after radiotherapy. Background Art

[0002] Radiotherapy is one of the important means of treating malignant tumors. About 70% of cancer patients need to receive radiotherapy at different stages of the disease. Therefore, patients who are sensitive to radiotherapy should be screened out before treatment and given individualized radiotherapy plans to improve the efficacy of patients and reduce side effects. This can achieve the dual benefits of prolonging survival and improving the quality of life and capture the characteristics of the disease. In recent years, A1 has shown unique advantages in automatic delineation and plan design of radiotherapy, and has been put into clinical application.

[0003] However, the A1 model predictions in current technologies are mostly single-center, retrospective data, and are mostly based on single-modality images such as imaging omics, which cannot fully capture the characteristics of the disease. Summary of the invention

[0004] In order to make up for the above shortcomings, the present invention provides an AI prediction system for recurrence of locally advanced rectal cancer after radiotherapy, aiming to improve the problem that the A1 model prediction data is squeezed out from single-center, retrospective data and cannot fully capture the characteristics of the disease.

[0005] In a first aspect, the present invention provides the following technical solution, an AI prediction system for recurrence of locally advanced rectal cancer after radiotherapy, comprising:

[0006] Multimodal biological image database establishment module: used to collect patients' clinical data, imaging data, radiotherapy target area and dose distribution data, radiobiology data and genomics information, and integrate the data to establish a database;

[0007] Feature extraction and screening module: used to extract clinical features, imaging features and radiobiological features from the collected data and screen them;

[0008] Prediction model building module: used to build a prediction model for recurrence of locally advanced rectal cancer after radiotherapy based on the screened features;

[0009] Model validation and optimization module: used to validate and optimize the model using prospective, multi-center clinical research cohorts;

[0010] Clinical application evaluation module: used to apply the optimized model to clinical practice and evaluate its role in radiotherapy plan formulation, efficacy monitoring and prognosis evaluation.

[0011] Preferably, the multimodal biological image database establishment module includes a clinical data collection unit, an image data collection unit, a radiotherapy target and dose distribution data acquisition unit and a radiobiological data detection unit. The clinical data collection unit is used to collect patient information, including but not limited to age, gender, maximum tumor diameter, pathological type, and smoking history. The image data acquisition unit is used to obtain the patient's CT, PET-CT or ultrasound medical imaging data. The radiotherapy target and dose distribution data acquisition unit is used to collect dose distribution parameter data of GTV, CTV, PTV, and organs at risk. The radiobiological data detection unit is used to evaluate the patient's DNA damage repair ability and genomic feature data through HRD score detection, comet experiment, and APsite detection.

[0012] Preferably, the feature extraction and screening module includes a clinical feature extraction unit, an imaging genomics feature extraction unit and a radiobiological feature extraction unit. The clinical feature extraction unit is used to extract gender and age clinical features from clinical data. The imaging genomics feature extraction unit is used to process image data through minimum shrinkage and selection operators, including shape features, first-order features, texture features, and wavelet transform features. The radiobiological feature extraction unit is used to calculate the HRD score, perform APsite detection, and then use the feature selection algorithm to screen out features with high correlation with radiotherapy efficacy.

[0013] Preferably, the prediction model building module includes a model architecture design unit and a feature fusion training unit. The model architecture design unit is used to establish a multimodal prediction model, and the feature fusion unit is used to fuse the screened features into a data set and input it into the model for training and learning.

[0014] Preferably, the model validation and optimization module includes an external data validation unit and a model optimization unit, wherein the external data validation unit is used to validate the model using external independent data from a prospective, multi-center clinical research cohort, and the model optimization unit is used to continuously optimize the model through cross-validation, sensitivity analysis, clinical feedback, and data accumulation.

[0015] Preferably, the clinical application evaluation module includes a physician decision evaluation unit and a treatment effect evaluation unit. The physician decision evaluation unit is used to analyze the degree of help the prediction model provides to the physician in evaluating the decision-making process, and the treatment effect evaluation unit is used to conduct a comprehensive evaluation of the treatment effect of the radiotherapy plan after the decision is applied and implemented, including but not limited to the reduction of the patient's tumor, the degree of symptom relief, the extension of survival time, and the quality of life.

[0016] In a second aspect, the present invention provides the following technical solution, an AI prediction method for recurrence of locally advanced rectal cancer after radiotherapy, the method comprising the following steps:

[0017] S1. Establishment of multimodal biological image database: Collect patient clinical data through the multimodal biological image database establishment module, and integrate these data to establish a database;

[0018] S2. Feature extraction and screening: Use the feature extraction and screening module to extract multi-dimensional comprehensive features and screen out features with high correlation with radiotherapy efficacy;

[0019] S3. Prediction model construction: Establish a multimodal prediction model, fuse the screened features to form a data set, input it into the model for training and learning, and obtain a prediction model for recurrence of locally advanced rectal cancer after radiotherapy;

[0020] S4. Model validation and optimization: The model is validated with external independent data from prospective, multicenter clinical research cohorts, and the model is continuously optimized through cross-validation, sensitivity analysis, clinical feedback and data accumulation;

[0021] S5. Clinical application evaluation: Formulate an individualized radiotherapy plan based on the doctor's feedback on the degree of helpfulness of the prediction model in the decision-making process and the evaluation of the treatment effect after the implementation of the radiotherapy plan.

[0022] In the third aspect, the invention provides the following technical solution: a computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the above-mentioned AI prediction method for recurrence of locally advanced rectal cancer after radiotherapy when executing the computer program.

[0023] In a fourth aspect, the present invention provides the following technical solution: a readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-mentioned AI prediction method for recurrence of locally advanced rectal cancer after radiotherapy.

[0024] The present invention has the following beneficial effects:

[0025] 1. In the present invention, by widely collecting clinical, imaging, radiotherapy target and dose distribution, radiation biology and genomics information of patients, and integrating and establishing a multimodal biological image database, it can comprehensively cover various factors affecting the recurrence of rectal cancer after radiotherapy, and provide a rich, accurate and comprehensive data basis for subsequent precise analysis and prediction, thereby improving the accuracy and reliability of the prediction of recurrence after radiotherapy.

[0026] 2. In the present invention, multiple tests are performed on the patient's tumor tissue and peripheral blood samples to evaluate the tumor cell DNA damage repair ability, obtain radiobiological data, and then construct a prediction model specifically for the recurrence of locally advanced rectal cancer after radiotherapy. The complex relationship between these characteristics can be used to provide clinicians with prediction information on recurrence after radiotherapy, and assist in making more reasonable treatment decisions;

[0027] 3. In the present invention, a prospective, multicenter clinical study is used to externally validate the multimodal prediction model, which can test the accuracy and stability of the model in different scenarios and patient groups, and adjust and improve the model based on the validation results, thereby ensuring that the model has good generalization ability and reliability, so that it can more accurately predict recurrence after radiotherapy in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a framework diagram of the AI ​​prediction system for recurrence of locally advanced rectal cancer after radiotherapy proposed by the present invention;

[0029] Figure 2 The architecture diagram of the multimodal biological image database establishment module of the AI ​​prediction system for recurrence of locally advanced rectal cancer after radiotherapy proposed by the present invention;

[0030] Figure 3 This is a diagram of the architecture of the feature extraction and screening module of the AI ​​prediction system for recurrence of locally advanced rectal cancer after radiotherapy proposed by the present invention;

[0031] Figure 4 This is an architecture diagram of a prediction model building module of an AI prediction system for recurrence of locally advanced rectal cancer after radiotherapy proposed by the present invention;

[0032] Figure 5 This is an architecture diagram of the model verification and optimization module of the AI ​​prediction system for recurrence of locally advanced rectal cancer after radiotherapy proposed by the present invention;

[0033] Figure 6 This is an architectural diagram of the clinical application evaluation module of the AI ​​prediction system for recurrence of locally advanced rectal cancer after radiotherapy proposed by the present invention;

[0034] Figure 7 This is a flow chart of the AI ​​prediction method for recurrence of locally advanced rectal cancer after radiotherapy proposed by the present invention. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] Embodiment 1

[0037] Reference Figure 1-Figure 6 In a first embodiment of the present invention, the present invention provides an AI prediction system for recurrence of locally advanced rectal cancer after radiotherapy, comprising:

[0038] Multimodal biological image database establishment module: used to collect patients' clinical data, imaging data, radiotherapy target area and dose distribution data, radiobiology data and genomics information, and integrate the data to establish a database;

[0039] Feature extraction and screening module: used to extract clinical features, imaging features and radiobiological features from the collected data and screen them;

[0040] Prediction model building module: used to build a prediction model for recurrence of locally advanced rectal cancer after radiotherapy based on the screened features;

[0041] Model validation and optimization module: used to validate and optimize the model using prospective, multi-center clinical research cohorts;

[0042] Clinical application evaluation module: used to apply the optimized model to clinical practice and evaluate its role in radiotherapy plan formulation, efficacy monitoring and prognosis evaluation.

[0043] Specifically, through the establishment of a multimodal biological image database module, a wide range of information on patients in terms of clinical, imaging, radiotherapy target and dose distribution, radiobiology, and genomics is collected, and these diverse data are integrated to build a comprehensive database, which can provide a rich, accurate and comprehensive data foundation for subsequent analysis and research, so that the research can cover various factors that affect the recurrence of rectal cancer after radiotherapy, and lay a data foundation for more accurate predictions;

[0044] The feature extraction and screening module can accurately extract clinical features, imaging features, and radiobiological features from the collected clinical data, and screen these features to remove redundant and irrelevant information, highlight key features closely related to the recurrence of rectal cancer after radiotherapy, make subsequent model construction more targeted and accurate, and improve the performance and reliability of the prediction model;

[0045] Through the prediction model building module, based on the key features after screening, a prediction model specifically for the recurrence of locally advanced rectal cancer after radiotherapy is constructed. The complex relationship and pattern between these features can be used to generate a mathematical model that can predict the recurrence after radiotherapy, providing valuable prediction information for clinicians to assist them in making more reasonable treatment decisions.

[0046] The model validation and optimization module uses a prospective, multi-center clinical research cohort to validate and optimize the constructed model, which can test the accuracy and stability of the model in different scenarios and patient groups, and adjust and improve the model based on the validation results, ensuring that the model has good generalization ability and reliability, so that it can more accurately predict recurrence after radiotherapy in practical applications;

[0047] The clinical application evaluation module can intuitively understand the value of the model in actual clinical work, and determine whether it can effectively help doctors develop more personalized and effective radiotherapy plans, whether it can accurately monitor the efficacy, and whether it can make an accurate assessment of the patient's prognosis, thereby ultimately improving the patient's treatment effect and quality of life.

[0048] The multimodal biological image database establishment module includes a clinical data collection unit, an image data collection unit, a radiotherapy target and dose distribution data acquisition unit, and a radiobiological data detection unit. The clinical data collection unit is used to collect patient information, including but not limited to age, gender, maximum tumor diameter, pathological type, and smoking history. The image data collection unit is used to obtain the patient's CT, PET-CT or ultrasound medical imaging data. The radiotherapy target and dose distribution data acquisition unit is used to collect dose distribution parameter data of GTV, CTV, PTV, and organs at risk. The radiobiological data detection unit is used to evaluate the patient's DNA damage repair ability and genome feature data through HRD score detection, comet experiment, and APsite detection.

[0049] Specifically, the clinical data collection unit can obtain the patient's basic information and key background data related to the disease. By collecting this data, it can provide a basis for the patient's individual characteristics for subsequent analysis, help to fully understand the patient's overall condition, and thus better evaluate the effect of radiotherapy and the risk of recurrence, and provide a basic portrait of the individual patient for subsequent research and model building, making the analysis more targeted and comprehensive.

[0050] The imaging data acquisition unit acquires the patient's CT, PET-CT or ultrasound medical imaging data, which can intuitively present the location, size, morphology and other specific conditions of the tumor in the patient's locally advanced rectal cancer. It can provide a visual basis for doctors and researchers, help accurately locate the tumor, evaluate its development stage, and provide an important reference for formulating radiotherapy plans and predicting recurrence, so that the medical team can understand the tumor more clearly, providing intuitive image support for precise treatment and prediction;

[0051] The radiotherapy target area and dose distribution data acquisition unit can accurately grasp the implementation scope and dose distribution of radiotherapy, which can ensure the pertinence and effectiveness of radiotherapy while reducing the risk of damage to normal tissues. It provides a quantitative basis for optimizing radiotherapy plans and evaluating treatment effects, and helps to improve the quality and safety of radiotherapy.

[0052] The radiobiology data detection unit can provide in-depth understanding of the biological characteristics of the tumor and its sensitivity to radiotherapy at the molecular level, which can provide a deeper biological basis for the subsequent prediction of radiotherapy effects and recurrence risks, help to formulate more individualized treatment strategies, and provide precision medicine with information at the molecular level, making treatment more targeted and forward-looking.

[0053] The feature extraction and screening module includes a clinical feature extraction unit, an imaging feature extraction unit and a radiobiological feature extraction unit. The clinical feature extraction unit is used to extract gender and age clinical features from clinical data. The imaging feature extraction unit is used to process image data through minimum shrinkage and selection operators, including shape features, first-order features, texture features, and wavelet transform features. The radiobiological feature extraction unit is used to calculate the HRD score, perform APsite detection, and then use the feature selection algorithm to screen out features that are highly correlated with radiotherapy efficacy.

[0054] Specifically, the clinical feature extraction unit can screen out key elements related to the basic conditions of individual patients, and provide a preliminary understanding of the potential impact of some inherent attributes of patients on diseases and treatments, providing basic individual feature information of patients for subsequent comprehensive analysis, which is helpful to consider treatment plans and predict results from a macro level;

[0055] The imaging genomics feature extraction unit can extract detailed information that can reflect the morphology, structure and internal heterogeneity of tumors from medical images, providing a more microscopic and quantitative basis for the diagnosis, treatment and prognosis of the disease, and then converting complex imaging data into quantifiable and analyzable features, providing richer imaging information support for precision medicine;

[0056] The radiobiological feature extraction unit reveals the characteristics of tumor cells and their response mechanisms to radiotherapy at the molecular biological level. By extracting these deep-level features, the effect of radiotherapy and the risk of recurrence can be predicted more accurately, providing a molecular biological basis for the formulation of personalized radiotherapy plans, which helps to improve the targetedness and effectiveness of treatment.

[0057] The prediction model building module includes a model architecture design unit and a feature fusion training unit. The model architecture design unit is used to establish a multimodal prediction model, and the feature fusion unit is used to fuse the screened features into a data set and input it into the model for training and learning.

[0058] Specifically, the overall structure and algorithm framework of the model are determined through the model architecture design unit to adapt to the processing and analysis of complex data from different sources and types, and can effectively integrate and process multimodal data such as clinical, imaging and radiobiology, and mine potential patterns and relationships in the data, providing a solid foundation and effective framework for subsequent model training and prediction, so that the model can accurately capture the interaction between various features and improve the accuracy and reliability of prediction;

[0059] Through the feature fusion training unit, features from different dimensions and fields are organically combined, allowing the model to learn more comprehensive and integrated information. Through fusion training, the model can fully utilize the advantages of multiple features and reduce the limitations of a single feature, thereby improving the generalization ability and predictive performance of the model. It can learn more in-depth and accurate patterns from the fused feature data, generate more accurate prediction results, and provide more valuable reference for clinical decision-making.

[0060] The model validation and optimization module includes an external data validation unit and a model optimization unit. The external data validation unit is used to validate the model using external independent data from prospective, multicenter clinical research cohorts. The model optimization unit is used to continuously optimize the model through cross-validation, sensitivity analysis, clinical feedback, and data accumulation.

[0061] Specifically, the external data verification unit can evaluate the universality and accuracy of the model in different medical institutions and different patient groups. By using external independent data for verification, it can avoid overfitting of the model to a specific data set, ensuring that the model has good generalization ability and reliability. The effect achieved is to provide an objective evaluation basis for the actual application of the model, and enhance the credibility and applicability of the model in clinical promotion.

[0062] The model optimization unit continuously optimizes the model through cross-validation, sensitivity analysis, clinical feedback and data accumulation, thereby continuously improving the performance and accuracy of the model to make it more in line with actual clinical needs. Cross-validation helps to select the optimal model parameters, sensitivity analysis can identify the sensitivity of the model to different inputs, and clinical feedback and data accumulation provide a practical direction and basis for optimization, so that the model can continue to evolve and improve with the addition of new data and changes in clinical needs, thereby improving its accuracy and practicality.

[0063] The clinical application evaluation module includes a physician decision-making evaluation unit and a treatment effect evaluation unit. The physician decision-making evaluation unit is used to analyze the degree of help the prediction model provides to the physician in evaluating the decision-making process. The treatment effect evaluation unit is used to conduct a comprehensive evaluation of the treatment effect of the radiotherapy plan after the decision is applied and implemented, including but not limited to the reduction of the patient's tumor, the degree of symptom relief, the extension of survival time, and the quality of life.

[0064] Specifically, the doctor decision-making evaluation unit can measure the support and influence of the information and suggestions provided by the model on doctors' decision-making in formulating treatment plans and selecting treatment strategies, and then understand whether the model is truly effective in assisting doctors to make more intelligent and scientific decisions. It can promote the continuous improvement and perfection of the model to better meet the decision-making needs of doctors, improve the accuracy and efficiency of medical decisions, and ultimately provide patients with more optimized treatment plans;

[0065] The treatment effect evaluation unit objectively and comprehensively evaluates the effectiveness of the treatment plan in actual application. It not only pays attention to the changes in the physiological indicators of the disease, but also pays attention to the improvement of the patient's overall health and quality of life. Through comprehensive evaluation, the advantages and disadvantages of the treatment plan are clarified, providing a basis for subsequent treatment adjustments and improvements. The effect achieved is to provide the medical team with clear feedback on the treatment effect, which helps to further optimize the treatment strategy and improve the patient's treatment satisfaction and quality of life.

[0066] Embodiment 2:

[0067] Reference Figure 7 In a second embodiment of the present invention, the present invention provides an AI prediction method for recurrence of locally advanced rectal cancer after radiotherapy, the method comprising the following steps:

[0068] S1. Establishment of multimodal biological image database: Collect patient clinical data through the multimodal biological image database establishment module, and integrate these data to establish a database;

[0069] S2. Feature extraction and screening: Use the feature extraction and screening module to extract multi-dimensional comprehensive features and screen out features with high correlation with radiotherapy efficacy;

[0070] S3. Prediction model construction: Establish a multimodal prediction model, fuse the screened features to form a data set, input it into the model for training and learning, and obtain a prediction model for recurrence of locally advanced rectal cancer after radiotherapy;

[0071] S4. Model validation and optimization: The model is validated with external independent data from prospective, multicenter clinical research cohorts, and the model is continuously optimized through cross-validation, sensitivity analysis, clinical feedback and data accumulation;

[0072] S5. Clinical application evaluation: Formulate an individualized radiotherapy plan based on the doctor's feedback on the degree of helpfulness of the prediction model in the decision-making process and the evaluation of the treatment effect after the implementation of the radiotherapy plan.

[0073] Specifically, by establishing a multimodal biological image database to collect and integrate comprehensive data, extracting and screening relevant features to build a prediction model, and then using external independent data to verify and optimize the model, and finally evaluating the model's assistance to doctors' decision-making and treatment effects to formulate an individualized radiotherapy plan, it provides a scientific, accurate and comprehensive method for predicting the recurrence of rectal cancer after radiotherapy, which can more accurately predict the risk of recurrence, assist doctors in formulating more optimized treatment plans, improve treatment effects and patients' quality of life, and promote the development of rectal cancer treatment towards individualization and precision.

[0074] Embodiment 3

[0075] The third embodiment of the present invention is based on the same inventive concept. The present invention proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the AI ​​prediction method for recurrence of locally advanced rectal cancer after radiotherapy in the above-mentioned embodiment.

[0076] Embodiment 4

[0077] The fourth embodiment of the present invention is based on the same inventive concept. The present invention proposes a computer, which includes: a processor and a memory; the processor and the memory communicate with each other; the memory is used to store instructions; the processor is used to execute the instructions in the memory, and execute the AI ​​prediction method for recurrence of locally advanced rectal cancer after radiotherapy of the above embodiment.

[0078] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0079] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An AI prediction system for recurrence of locally advanced rectal cancer after radiotherapy, characterized by: include: Multimodal biological image database establishment module: used to collect patients' clinical data, imaging data, radiotherapy target area and dose distribution data, radiobiology data and genomics information, and integrate the data to establish a database; Feature extraction and screening module: used to extract clinical features, imaging features and radiobiological features from the collected data and screen them; Prediction model building module: used to build a prediction model for recurrence of locally advanced rectal cancer after radiotherapy based on the screened features; Model validation and optimization module: used to validate and optimize the model using prospective, multi-center clinical research cohorts; Clinical application evaluation module: used to apply the optimized model to clinical practice and evaluate its role in radiotherapy plan formulation, efficacy monitoring and prognosis evaluation.

2. The AI ​​prediction system for recurrence of locally advanced rectal cancer after radiotherapy according to claim 1, characterized in that: The multimodal biological image database establishment module includes a clinical data collection unit, an image data collection unit, a radiotherapy target and dose distribution data acquisition unit and a radiobiological data detection unit. The clinical data collection unit is used to collect patient information, including but not limited to age, gender, maximum tumor diameter, pathological type, and smoking history. The image data collection unit is used to obtain the patient's CT, PET-CT or ultrasound medical imaging data. The radiotherapy target and dose distribution data acquisition unit is used to collect dose distribution parameter data of GTV, CTV, PTV, and organs at risk. The radiobiological data detection unit is used to evaluate the patient's DNA damage repair ability and genome feature data through HRD score detection, comet experiment, and APsite detection.

3. The AI ​​prediction system for recurrence of locally advanced rectal cancer after radiotherapy according to claim 1, characterized in that: The feature extraction and screening module includes a clinical feature extraction unit, an imaging genomics feature extraction unit and a radiobiological feature extraction unit. The clinical feature extraction unit is used to extract gender and age clinical features from clinical data. The imaging genomics feature extraction unit is used to process image data through minimum shrinkage and selection operators, including shape features, first-order features, texture features, and wavelet transform features. The radiobiological feature extraction unit is used to calculate the HRD score, perform APsite detection, and then use the feature selection algorithm to screen out features with high correlation with radiotherapy efficacy.

4. The AI ​​prediction system for recurrence of locally advanced rectal cancer after radiotherapy according to claim 1, characterized in that: The prediction model construction module includes a model architecture design unit and a feature fusion training unit. The model architecture design unit is used to establish a multimodal prediction model, and the feature fusion unit is used to fuse the screened features into a data set and input it into the model for training and learning.

5. The AI ​​prediction system for recurrence of locally advanced rectal cancer after radiotherapy according to claim 1, characterized in that: The model validation and optimization module includes an external data validation unit and a model optimization unit. The external data validation unit is used to validate the model using external independent data from a prospective, multi-center clinical research cohort, and the model optimization unit is used to continuously optimize the model through cross-validation, sensitivity analysis, clinical feedback, and data accumulation.

6. The AI ​​prediction system for recurrence of locally advanced rectal cancer after radiotherapy according to claim 1, characterized in that: The clinical application evaluation module includes a doctor decision evaluation unit and a treatment effect evaluation unit. The doctor decision evaluation unit is used to analyze the degree of help the prediction model provides to the doctor in evaluating the decision-making process. The treatment effect evaluation unit is used to comprehensively evaluate the treatment effect of the radiotherapy plan after the decision is applied and implemented, including but not limited to the reduction of the patient's tumor, the degree of symptom relief, the extension of survival time, and the quality of life.

7. An AI prediction method for recurrence of locally advanced rectal cancer after radiotherapy, characterized in that: An AI prediction system for recurrence of locally advanced rectal cancer after radiotherapy according to any one of claims 1 to 6, the method comprising the following steps: S1. Establishment of multimodal biological image database: Collect patient clinical data through the multimodal biological image database establishment module, and integrate these data to establish a database; S2. Feature extraction and screening: Use the feature extraction and screening module to extract multi-dimensional comprehensive features and screen out features with high correlation with radiotherapy efficacy; S3. Prediction model construction: Establish a multimodal prediction model, fuse the screened features to form a data set, input it into the model for training and learning, and obtain a prediction model for recurrence of locally advanced rectal cancer after radiotherapy; S4. Model validation and optimization: The model is validated with external independent data from prospective, multicenter clinical research cohorts, and the model is continuously optimized through cross-validation, sensitivity analysis, clinical feedback and data accumulation; S5. Clinical application evaluation: Formulate an individualized radiotherapy plan based on the doctor's feedback on the degree of helpfulness of the prediction model in the decision-making process and the evaluation of the treatment effect after the implementation of the radiotherapy plan.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the AI ​​prediction method for recurrence of locally advanced rectal cancer after radiotherapy as described in claim 7 is implemented.

9. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by the processor, the AI ​​prediction method for recurrence of locally advanced rectal cancer after radiotherapy as claimed in claim 7 is implemented.

Images