Artificial intelligence-based rectal cancer radiotherapy gene screening and targeted therapy optimization system
By building a radiotherapy gene screening and targeted therapy optimization system based on artificial intelligence, the problem of insufficient data fusion in the existing technology has been solved, the individualized accuracy of rectal cancer treatment and the stability of predictive models have been improved, and the scientificity and clinical application of the treatment plan have been optimized.
Patent Information
- Application Number
- CN202510515922.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-23
- Publication Date
- 2025-08-05
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art lacks systematic and multi-dimensional data fusion analysis methods in the treatment of rectal cancer, which makes it difficult to achieve individualized treatment, and the data samples of radiotherapy and targeted treatment effect prediction models are insufficient and the generalization ability is poor, which limits its application in clinical practice.
Using a rectal cancer radiotherapy gene screening and targeted therapy optimization system based on artificial intelligence, a patient's genomic information and medical record information is obtained through the data acquisition module, a prediction model is constructed and treatment effect prediction is combined with multi-source data, and a genomic data completion mechanism and a dynamic model update mechanism are introduced to optimize the treatment plan.
It has achieved the individualized accuracy of rectal cancer treatment, improved the stability and generalization ability of the prediction model, and enhanced the scientificity and clinical practical value of the treatment plan.
Smart Images
Figure CN120432191A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical artificial intelligence technology, and in particular to an artificial intelligence-based rectal cancer radiotherapy gene screening and targeted therapy optimization system. Background Art
[0002] Rectal cancer is one of the most common malignant tumors in the digestive system, with an increasing incidence rate year by year. Its treatment strategy usually includes the combined use of multiple methods such as surgery, radiotherapy, chemotherapy and targeted therapy. Radiotherapy plays an important role in preoperative downstaging of rectal cancer, postoperative adjuvant treatment and in inoperable advanced cases. However, due to significant differences in genetic background, tumor molecular characteristics and treatment sensitivity among individual patients, the same radiotherapy regimen can show large differences in efficacy in different patients.
[0003] Currently, clinical treatment decisions are mainly based on imaging assessment, pathological staging and some genetic markers, but there is a lack of systematic, multi-dimensional data fusion and analysis methods, making it difficult to achieve truly individualized treatment. In addition, existing models still have problems such as insufficient data samples, poor model generalization ability, and low utilization of actual efficacy feedback when predicting radiotherapy response or targeted therapy effects, which limits their widespread application in clinical practice. Therefore, there is an urgent need for an intelligent system that integrates multimodal data and has self-learning and optimization capabilities to assist in achieving more accurate recommendations for rectal cancer treatment strategies. Summary of the Invention
[0004] The purpose of the present invention is to address the current deficiencies and propose an artificial intelligence-based rectal cancer radiotherapy gene screening and targeted therapy optimization system.
[0005] The present invention adopts the following technical solutions:
[0006] A rectal cancer radiotherapy gene screening and targeted treatment optimization system based on artificial intelligence, the system includes a data acquisition module, a radiotherapy effect prediction module and a result output module; the data acquisition module is used to collect the patient's genetic information and medical history information; the radiotherapy effect prediction module is used to predict the treatment effect based on the patient's genetic information, medical history information and candidate treatment plans; the result output module is used to provide the predicted treatment plan with the best treatment effect to the staff.
[0007] The data acquisition module includes a genome information acquisition unit and a medical record acquisition unit; the genome information acquisition unit is used to acquire the patient's genome sequencing data and transcriptome data as genetic information; the medical record acquisition unit is used to acquire the patient's tumor medical imaging data as medical record information.
[0008] The radiotherapy effect prediction module includes a prediction model construction unit, a prediction unit and a model update unit; the prediction model construction unit is used to construct a prediction model for predicting the treatment effects achieved by different treatment plans; the prediction unit is used to input the current patient's genetic information, medical history information and multiple candidate treatment plans into the prediction model, output the treatment effect corresponding to each treatment plan, and output the treatment plan with the best treatment effect to the result output module; the model update unit is used to periodically update and optimize the prediction model during the system operation.
[0009] Furthermore, the prediction model construction unit includes a data expansion enhancer unit and a model construction subunit; the data expansion enhancer unit is used to supplement the genetic information of some sample data that are missing genetic information based on sample data in a public database to generate new sample data, so as to expand the training set size required for establishing the prediction model; the model construction subunit is used to use the expanded training set to construct a prediction model.
[0010] Furthermore, the specific workflow of the data expansion and enhancement subunit is as follows:
[0011] S11: Obtaining genomic data, treatment plan data, and treatment effect data from a public database; the genomic data includes genomic sequencing data and transcriptome data, the treatment plan data includes the type and dosage of targeted therapy drugs, and the treatment effect data can be obtained through medical imaging data before and after the treatment plan is implemented;
[0012] S12: Each piece of sample data obtained in the previous step that contains genomic data, treatment plan data, and treatment effect data is used as basic sample data and put into the training set; each piece of sample data that only contains treatment plan data and treatment effect data is used as expandable sample data;
[0013] S13: Match and compare the expandable sample data with the basic sample data to construct the genome data in the expandable sample data; the specific construction method is as follows:
[0014] S131: Calculate the maximum similarity coefficient between the expandable sample data and the basic sample data:
[0015]
[0016] Among them, max() is the maximum value function; sim(i,j) is the similarity coefficient between a certain expandable sample data i and a certain basic sample data j, S i is the treatment plan and treatment effect vector of the expandable sample data i; S j is the treatment plan and treatment effect vector of basic sample data j;
[0017] S132: Acquire the transcriptome data of the basic sample data j, and perturb the expression values of the non-radiotherapy sensitive core genes therein to generate new transcriptome data as the transcriptome data of the expandable sample data i; the specific method of perturbation satisfies:
[0018] G i =G j +α·δ·[1-max(sim(i,j))];
[0019] Among them, G i is the transcriptome data of the expandable sample data i, G j is the transcriptome data of the basic sample data j, α is the perturbation direction coefficient, which is randomly set to 1 or -1; δ is the perturbation intensity coefficient, which is used to control the overall amplitude of the perturbation and is set by pre-experimental settings;
[0020] S133: combining the transcriptome data of the expandable sample data i calculated in the previous step with the genome sequencing data of the basic sample data j as the genome data of the expandable sample data i;
[0021] S14: putting the expandable sample data i of the expanded genome data into the training set;
[0022] S15: Repeat the above steps S13 to S14 until the number of sample data in the training set meets the training requirements.
[0023] Furthermore, the model construction subunit uses the patient's genetic information in the sample data, the medical imaging data before the implementation of the treatment plan, and the treatment plan as model input; uses the medical imaging data after the implementation of the treatment plan as a basis, and combines the changes in the medical imaging data before and after the implementation of the treatment plan to construct treatment effect data reflecting the actual therapeutic effect as model output.
[0024] The beneficial effects achieved by the present invention are:
[0025] The present invention combines the patient's genomic information, medical history information and multiple candidate treatment options to achieve accurate prediction of treatment effects and individualized recommendations; by introducing a genomic data completion mechanism, it effectively expands the scale of training data and improves the stability and generalization ability of the prediction model; at the same time, it establishes a dynamic model update mechanism based on actual efficacy feedback, enhances the system's adaptive optimization capabilities, and improves the scientific nature and clinical practical value of matching radiotherapy with targeted treatment options. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but rather the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate corresponding parts.
[0027] Figure 1 It is a schematic diagram of the overall module of the present invention.
[0028] Figure 2 Schematic diagram of the specific workflow of the data expansion enhancer unit of the present invention.
[0029] Figure 3 Schematic diagram of the specific workflow of the model updating unit of the present invention.
[0030] Figure 4 For the present invention k With max(sim(k,k j ))Schematic diagram of the function of parameter changes. DETAILED DESCRIPTION
[0031] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below in conjunction with its embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention; for those skilled in the art, after reviewing the following detailed description, other systems, methods and / or features of the present embodiment will become apparent; it is intended that all such additional systems, methods, features and advantages are included in this specification; included within the scope of the present invention and protected by the appended claims; additional features of the disclosed embodiments are described in the following detailed description, and these features will be apparent from the following detailed description.
[0032] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", etc. indicating directions or positional relationships, they are based on the directions or positional relationships shown in the drawings. This is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or component referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting this patent. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.
[0033] Example 1:
[0034] like Figure 1As shown, this embodiment provides an artificial intelligence-based rectal cancer radiotherapy gene screening and targeted therapy optimization system, the system comprising a data acquisition module, a radiotherapy effect prediction module, and a result output module; the data acquisition module is used to collect the patient's genetic information and medical history information; the radiotherapy effect prediction module is used to predict the treatment effect based on the patient's genetic information, medical history information, and candidate treatment plans; the result output module is used to provide the predicted treatment plan with the best treatment effect to the staff;
[0035] The data acquisition module includes a genome information acquisition unit and a medical record acquisition unit; the genome information acquisition unit is used to acquire the patient's genome sequencing data and transcriptome data as genetic information; the medical record acquisition unit is used to acquire the patient's tumor medical imaging data as medical record information;
[0036] The radiotherapy effect prediction module includes a prediction model construction unit, a prediction unit, and a model update unit; the prediction model construction unit is used to construct a prediction model for predicting the treatment effects achieved by different treatment plans; the prediction unit is used to input the current patient's genetic information, medical history information, and multiple candidate treatment plans into the prediction model, output the treatment effect corresponding to each treatment plan, and output the treatment plan with the best treatment effect to the result output module; the model update unit is used to periodically update and optimize the prediction model during system operation;
[0037] Furthermore, the prediction model construction unit includes a data expansion and enhancement subunit and a model construction subunit; the data expansion and enhancement subunit is used to supplement the gene information of some sample data that lacks gene information based on sample data in a public database to generate new sample data, so as to expand the training set size required for establishing the prediction model; the model construction subunit is used to use the expanded training set to construct the prediction model;
[0038] Further, such as Figure 2 As shown, the specific workflow of the data expansion and enhancement subunit is as follows:
[0039] S11: Obtaining genomic data, treatment plan data, and treatment effect data from a public database; the genomic data includes genomic sequencing data and transcriptome data, the treatment plan data includes the type and dosage of targeted therapy drugs, and the treatment effect data can be obtained through medical imaging data before and after the treatment plan is implemented;
[0040] S12: Each piece of sample data obtained in the previous step that contains genomic data, treatment plan data, and treatment effect data is used as basic sample data and put into the training set; each piece of sample data that only contains treatment plan data and treatment effect data is used as expandable sample data;
[0041] S13: Match and compare the expandable sample data with the basic sample data to construct the genome data in the expandable sample data; the specific construction method is as follows:
[0042] S131: Calculate the maximum similarity coefficient between the expandable sample data and the basic sample data:
[0043]
[0044] Among them, max() is the maximum value function; sim(i,j) is the similarity coefficient between a certain expandable sample data i and a certain basic sample data j, S i is the treatment plan and treatment effect vector of the expandable sample data i; S j is the treatment plan and treatment effect vector of the basic sample data j; the above formula can be used to measure the similarity between the treatment plan and efficacy characteristics of the expandable sample and the basic sample, so as to select the most representative matching sample when completing the gene information, thereby improving the accuracy and biological plausibility of the completion result;
[0045] S132: Acquire the transcriptome data of the basic sample data j, and perturb the expression values of the non-radiotherapy sensitive core genes therein to generate new transcriptome data as the transcriptome data of the expandable sample data i; the specific method of perturbation satisfies:
[0046] G i =G j +α·δ·[1-max(sim(i,j))];
[0047] Among them, G i is the transcriptome data of the expandable sample data i, G j is the transcriptome data of the basic sample data j, α is the perturbation direction coefficient, which randomly takes a value of 1 or -1; δ is the perturbation intensity coefficient, which is used to control the overall amplitude of the perturbation and is set by pre-experimentation. The above formula can introduce moderate variability while maintaining the biological similarity between samples, generating new transcriptome data that conforms to the real distribution, thereby enhancing the diversity of training data and the generalization ability of the model;
[0048] S133: combining the transcriptome data of the expandable sample data i calculated in the previous step with the genome sequencing data of the basic sample data j as the genome data of the expandable sample data i;
[0049] S14: putting the expandable sample data i of the expanded genome data into the training set;
[0050] S15: Repeat the above steps S13 to S14 until the number of sample data in the training set meets the training requirements;
[0051] Furthermore, the model building subunit uses the patient's genetic information in the sample data, the medical imaging data before the treatment plan is implemented, and the treatment plan as the model input; uses the medical imaging data after the treatment plan is implemented as the basis, and combines the changes in the medical imaging data before and after the treatment plan to construct the treatment effect data reflecting the actual treatment effect as the model output;
[0052] Further, such as Figure 4 As shown, the loss function of the prediction model during training satisfies:
[0053]
[0054] Among them, L is the total loss function of the prediction model, n is the total number of sample data in the training set, ω k is the weighting factor of the kth sample data in the training set, x k is the input feature data of the kth sample data, f(x k ) is the prediction model for x k The calculated predicted output, that is, the treatment effect data predicted by the model; y k is the true treatment effect label of the kth sample data, that is, the treatment effect data in the sample data; for ω k satisfy;
[0055]
[0056] Among them, max(sim(k,k j )) represents the kth sample data and a certain basic sample data k j The maximum similarity coefficient between
[0057] Through the above training method, the influence of sample data on model training can be dynamically adjusted according to its source and credibility, thereby improving the robustness of the prediction model and its adaptability to individual differences;
[0058] Furthermore, the prediction unit pre-screens multiple candidate treatment plans by combining an expert rule base, high-frequency effective plans in a database, clinical guideline recommendations, and historical cases; then, each candidate treatment plan is combined with the current patient's genetic information and medical history information to construct multiple treatment combination input data, which are input into the prediction model in sequence, thereby obtaining treatment effect data corresponding to multiple candidate treatment plans; and selecting the candidate treatment plan corresponding to the optimal treatment effect data as the treatment plan with the optimal treatment effect; the treatment effect data is specifically the tumor shrinkage rate extracted by comparing the medical imaging data before and after the implementation of the treatment plan.
[0059] Example 2:
[0060] This embodiment should be understood to include at least all the features of any of the aforementioned embodiments and be further improved thereon;
[0061] This embodiment provides an artificial intelligence-based rectal cancer radiotherapy gene screening and targeted therapy optimization system, comprising a data acquisition module, a radiotherapy effect prediction module, and a result output module; the data acquisition module is configured to collect a patient's genetic information and medical history information; the radiotherapy effect prediction module is configured to predict the treatment effect based on the patient's genetic information, medical history information, and candidate treatment plans; and the result output module is configured to provide the predicted treatment plan with the optimal treatment effect to staff;
[0062] The data acquisition module includes a genome information acquisition unit and a medical record acquisition unit; the genome information acquisition unit is used to acquire the patient's genome sequencing data and transcriptome data as genetic information; the medical record acquisition unit is used to acquire the patient's tumor medical imaging data as medical record information;
[0063] The radiotherapy effect prediction module includes a prediction model construction unit, a prediction unit, and a model update unit; the prediction model construction unit is used to construct a prediction model for predicting the treatment effects achieved by different treatment plans; the prediction unit is used to input the current patient's genetic information, medical history information, and multiple candidate treatment plans into the prediction model, output the treatment effect corresponding to each treatment plan, and output the treatment plan with the best treatment effect to the result output module; the model update unit is used to periodically update and optimize the prediction model during system operation;
[0064] Further, such as Figure 3 As shown, the model updating unit completes the update of the prediction model in the following manner:
[0065] S21: Acquire patient sample data that is continuously added during the operation of the system, wherein the added sample data includes patient genome data, treatment plan data, actual treatment effect data, and model-predicted treatment effect data; the actual treatment effect data is treatment effect data obtained by comparing clinical medical images before and after the treatment plan is implemented, and the model-predicted treatment effect data is treatment effect data predicted by the prediction model for the selected treatment plan before the patient receives treatment;
[0066] S22: When the newly acquired sample data reaches a preset input threshold, a model update is triggered;
[0067] S23: Use the newly added sample data as a new round of training samples and re-train the prediction model. During the training process, the loss function satisfies:
[0068]
[0069] Among them, L′ is the loss function of the model update stage, m is the total number of new sample data, is the weighting factor of the lth sample data in the newly added sample data, x l is the input feature data of the lth sample data, f(x l ) is the prediction model for x l The calculated predicted output, that is, the treatment effect data predicted by the model; y l is the true treatment effect label of the lth sample data, that is, the actual treatment effect data in the sample data; satisfy;
[0070]
[0071] Among them, f′(x l ) is the treatment effect data predicted by the model in the newly added sample data; γ is the weight control coefficient, which is used to adjust the influence of the prediction error on the sample weighting factor and is set through pre-experimental settings;
[0072] Through the above settings, sample data with larger prediction errors can be given higher training weights during the model update process, thereby guiding the model to focus on learning the error area, so that the prediction model can achieve targeted correction and adaptive optimization.
[0073] The contents disclosed above are only preferred feasible embodiments of the present invention and do not limit the scope of protection of the present invention. Therefore, all equivalent technical changes made using the contents of the present invention description and drawings are included in the scope of protection of the present invention. In addition, the elements therein can be updated as technology develops.
Claims
1. An artificial intelligence-based rectal cancer radiotherapy gene screening and targeted therapy optimization system, characterized by: The system includes a data acquisition module, a radiotherapy effect prediction module, and a result output module; the data acquisition module is used to collect the patient's genetic information and medical history information; the radiotherapy effect prediction module is used to predict the treatment effect based on the patient's genetic information, medical history information, and candidate treatment plans; The result output module is used to provide the predicted treatment plan with the best treatment effect to the staff; The data acquisition module includes a genome information acquisition unit and a medical record acquisition unit; the genome information acquisition unit is used to acquire the patient's genome sequencing data and transcriptome data as genetic information; the medical record acquisition unit is used to acquire the patient's tumor medical imaging data as medical record information; The radiotherapy effect prediction module includes a prediction model construction unit, a prediction unit and a model update unit; the prediction model construction unit is used to construct a prediction model for predicting the treatment effects achieved by different treatment plans; the prediction unit is used to input the current patient's genetic information, medical history information and multiple candidate treatment plans into the prediction model, output the treatment effect corresponding to each treatment plan, and output the treatment plan with the best treatment effect to the result output module; the model update unit is used to periodically update and optimize the prediction model during the system operation.
2. The artificial intelligence-based rectal cancer radiotherapy gene screening and targeted therapy optimization system according to claim 1, characterized in that: The prediction model construction unit includes a data expansion and enhancement subunit and a model construction subunit; the data expansion and enhancement subunit is used to supplement the genetic information of some sample data that lack genetic information based on sample data in a public database to generate new sample data, so as to expand the training set size required for establishing the prediction model; the model construction subunit is used to use the expanded training set to construct a prediction model.
3. The artificial intelligence-based rectal cancer radiotherapy gene screening and targeted therapy optimization system according to claim 1, characterized in that: The specific workflow of the data expansion and enhancement subunit is as follows: S11: Obtaining genomic data, treatment plan data, and treatment effect data from a public database; the genomic data includes genomic sequencing data and transcriptome data, the treatment plan data includes the type and dosage of targeted therapy drugs, and the treatment effect data can be obtained through medical imaging data before and after the treatment plan is implemented; S12: Each piece of sample data obtained in the previous step that contains genomic data, treatment plan data, and treatment effect data is used as basic sample data and put into the training set; each piece of sample data that only contains treatment plan data and treatment effect data is used as expandable sample data; S13: Match and compare the expandable sample data with the basic sample data to construct the genome data in the expandable sample data; the specific construction method is as follows: S131: Calculate the maximum similarity coefficient between the expandable sample data and the basic sample data: Among them, max() is the maximum value function; sim(i,j) is the similarity coefficient between a certain expandable sample data i and a certain basic sample data j, S i is the treatment plan and treatment effect vector of the expandable sample data i; S j is the treatment plan and treatment effect vector of basic sample data j; S132: Acquire the transcriptome data of the basic sample data j, and perturb the expression values of the non-radiotherapy sensitive core genes therein to generate new transcriptome data as the transcriptome data of the expandable sample data i; the specific method of perturbation satisfies: G i =G j +α·δ·[1-max(sim(i,j))]; Among them, G i is the transcriptome data of the expandable sample data i, G j is the transcriptome data of the basic sample data j, α is the perturbation direction coefficient, which is randomly set to 1 or -1; δ is the perturbation intensity coefficient, which is used to control the overall amplitude of the perturbation and is set by pre-experimental settings; S133: combining the transcriptome data of the expandable sample data i calculated in the previous step with the genome sequencing data of the basic sample data j as the genome data of the expandable sample data i; S14: putting the expandable sample data i of the expanded genome data into the training set; S15: Repeat the above steps S13 to S14 until the number of sample data in the training set meets the training requirements.
4. The artificial intelligence-based rectal cancer radiotherapy gene screening and targeted therapy optimization system according to claim 1, characterized in that: The model construction subunit uses the patient's genetic information in the sample data, the medical imaging data before the implementation of the treatment plan, and the treatment plan as model input; uses the medical imaging data after the implementation of the treatment plan as the basis, and combines the changes in the medical imaging data before and after the implementation of the treatment plan to construct treatment effect data reflecting the actual therapeutic effect as model output.