Intelligent interaction method for patient in radiotherapy process, computing device and storage medium

Abstract

The invention discloses an intelligent interaction method for a patient in a radiotherapy process, computing equipment and a storage medium, and the method comprises the steps: carrying out the natural language interaction between the patient and an interaction agent in the radiotherapy process; collecting multi-modal and multi-dimensional patient real-time feedback data and carrying out preprocessing; performing real-time fine tuning optimization on a plurality of pre-training basic models in the interactive agent according to the collected real-time feedback data; the finely-adjusted and optimized interaction agent generates targeted education content, health suggestions and emotion support according to the real-time feedback data of the patient, and outputs the education content, the health suggestions and the emotion support in a natural language mode; and repeating the steps until the radiotherapy is finished. In the radiotherapy process, the interaction agent capable of interacting with the patient is introduced, real-time dynamic personalized health education, treatment suggestions and emotion support services are provided for the patient, and emotion problems such as anxiety of the patient are relieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of radiotherapy, and particularly relates to an intelligent interaction method, a computing device, and a storage medium for a patient during radiotherapy. Background Art

[0002] With the rapid development of medical technology, radiotherapy has become an important means for treating major diseases such as cancer. However, the radiotherapy process is complex, and patients often face double challenges, both physically and psychologically, during the treatment. For example, due to the lack of understanding of the treatment method, possible side effects, and treatment course, many patients will feel anxious, fearful, or even resistant. In addition, in the traditional medical model, the communication time between doctors and patients is limited, making it difficult to fully meet the patients' needs in terms of education, emotional support, and personalized health management.

[0003] In the prior art, although there are some rule-based health education systems or simple medical advice tools, these systems usually lack the ability of personalization and cannot be dynamically adjusted according to the actual state, treatment progress, and emotional needs of patients. Some intelligent systems rely on predefined content libraries or fixed feedback mechanisms, making it difficult to adapt to the changing health and psychological states of patients during radiotherapy. This limitation may lead to a decrease in patients' compliance with treatment advice and even affect the treatment effect.

[0004] Therefore, the existing systems applied in the radiotherapy process have the following problems:

[0005] First, the existing intelligent agent systems usually only provide rule-based or predefined content and lack personalized dynamic adjustment.

[0006] Second, many intelligent health systems are based on fixed models or manually set feedback mechanisms, making it difficult to handle the individual differences of patients. Especially during long-term treatment, the adaptability of the system is poor.

[0007] Third, emotion recognition and health management are usually separate modules and lack an effective joint optimization mechanism. Summary of the Invention

[0008] To solve the above technical problems, the present invention proposes an intelligent interaction method, a computing device, and a storage medium for a patient during radiotherapy.

[0009] To achieve the above object, the technical solution of the present invention is as follows:

[0010] In a first aspect, the present invention discloses an intelligent interaction method for a patient during radiotherapy, including:

[0011] Step S1: During radiotherapy, the patient conducts natural language interaction with an interaction intelligent agent;

[0012] Step S2: Collect real-time feedback data of patients in multiple modalities and multiple dimensions and perform preprocessing. The feedback data includes one or more types of multimodal data among voice data, text data, and image data, and the feedback data includes one or more types of multidimensional data among treatment feedback, emotional expression, and health data;

[0013] Step S3: Several pre-trained basic models in the interactive agent are fine-tuned and optimized according to the collected real-time feedback data;

[0014] Step S4: The fine-tuned and optimized interactive agent generates targeted educational content, health advice, and emotional support based on the real-time feedback data of the patient and outputs them in natural language;

[0015] Step S5: Repeat Step S1 - Step S4 until the radiotherapy ends.

[0016] Based on the above technical solutions, the following improvements can be made:

[0017] As a preferred solution, the fine-tuning methods include one or more of supervised learning fine-tuning method, reinforcement learning fine-tuning method, unsupervised learning fine-tuning method, meta-learning fine-tuning method, and transfer learning fine-tuning method.

[0018] As a preferred solution, during the fine-tuning process using the fine-tuning method, the basic model is adjusted in any of the following ways;

[0019] Method 1: Freeze the lower layers of the basic model to prevent its weights from being updated and retain the general feature representation of the basic model;

[0020] Method 2: Only fine-tune the higher layers related to the task in the basic model.

[0021] As a preferred solution, during the fine-tuning process using the supervised learning fine-tuning method, a labeled dataset is used;

[0022] Use the cross-entropy loss function L cross_entropy Minimize the error between the predicted value of the fine-tuned basic model and the actual label. The specific formula is as follows:

[0023]

[0024] Where:

[0025] y i Is the actual label;

[0026] p i Is the predicted value of the model.

[0027] As a preferred solution, during the fine-tuning process using the reinforcement learning fine-tuning method, the optimization objective is as follows:

[0028]

[0029] Wherein:

[0030] γ t is the discount factor;

[0031] R t is the reward obtained at time step t;

[0032] T is the total number of time steps of the task;

[0033] θ is a parameter.

[0034] As a preferred solution, the interactive agent includes the following basic models:

[0035] A text generation model for converting the patient's voice data into text tasks;

[0036] An image analysis model for analyzing the patient's image data;

[0037] A multi-modal model for jointly modeling data of different modalities during radiotherapy.

[0038] As a preferred solution, the interactive agent can track the interaction situation of the patient during each radiotherapy process and perform periodic fine-tuning and optimization on the basic model.

[0039] As a preferred solution, the preprocessing operation in step S2 includes one or more of denoising, standardization, and missing value filling.

[0040] In a second aspect, the present invention discloses a computing device, including:

[0041] One or more processors;

[0042] A memory;

[0043] And one or more programs, wherein one or more programs are stored in the memory and configured to be executed by one or more processors, and one or more programs include instructions for any of the above intelligent interaction methods of the patient during radiotherapy.

[0044] In a third aspect, the present invention discloses a storage medium storing one or more computer-readable programs, and one or more programs include instructions adapted to be loaded and executed by the memory for any of the above intelligent interaction methods of the patient during radiotherapy.

[0045] The present invention discloses an intelligent interaction method, a computing device, and a storage medium for a patient during radiotherapy, which have the following beneficial effects:

[0046] First, during the radiotherapy process, the present invention introduces an interactive agent that can interact with patients, providing real-time dynamic personalized health education, treatment advice, and emotional support services to alleviate emotional problems such as patients' anxiety.

[0047] Second, the present invention introduces an interactive agent based on a foundation model, and the foundation model can be adaptively micro-adjusted according to the real-time feedback data of patients, significantly improving the treatment experience and compliance of patients, and at the same time providing more efficient auxiliary support for medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and thus should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0049] Figure 1 It is a flowchart of the intelligent interaction method provided for the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] The preferred embodiments of the present invention will be described in detail below with reference to the drawings.

[0051] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0052] In addition, the expression of "including" elements is an "open-ended" expression, which only means the existence of corresponding components or steps, and should not be construed as excluding additional components or steps.

[0053] In order to achieve the purpose of the present invention, in some embodiments of the intelligent interaction method for patients during the radiotherapy process, as Figure 1 shown, the intelligent interaction method includes:

[0054] Step S101: During the radiotherapy process, the patient conducts natural language interaction with the interactive agent;

[0055] Step S102: Collect and preprocess multi-modal and multi-dimensional real-time feedback data of the patient. The feedback data includes: multi-modal data such as voice data, text data, and image data, and the feedback data includes: multi-dimensional data such as treatment feedback, emotional expressions (such as anxiety, depression, etc.), and health data;

[0056] Step S103: A number of pre-trained foundation models in the interactive agent are fine-tuned in real time according to the collected real-time feedback data;

[0057] Step S104: The fine-tuned interactive agent generates targeted educational content, health advice, and emotional support based on the patient's real-time feedback data and outputs them in natural language;

[0058] Step S105: Repeat Steps S101 - S104 until the radiotherapy ends.

[0059] The intelligent interaction method of the present invention is implemented based on an interactive agent. The interactive agent is applied to the education and interaction with patients during radiotherapy and can adjust educational content, health advice, and emotional support according to the personalized needs and feedback data of patients.

[0060] The interactive agent supports the integration of various data forms of patients (such as text, voice, images, etc.) and realizes more comprehensive and accurate health management and advice generation through a multi-modal model.

[0061] The interactive agent can real-time capture the patient's emotional state (such as anxiety, depression, etc.) and generate targeted emotional support strategies based on the fine-tuned model, such as soothing language, positive incentives, etc., to enhance the patient's mental health and treatment experience.

[0062] The data collection in Step S102 is the basis for collecting the patient's real-time feedback data (such as health data, treatment records, emotional feedback, etc.) to ensure the accuracy and integrity of the data for subsequent processing.

[0063] The real-time feedback data combines multiple dimensions (such as treatment feedback, emotional expression, and health data) and multiple modalities (such as voice data, text data, imaging data). Compared with the traditional single data source collection method, it can comprehensively understand the patient's state.

[0064] The above-mentioned collection of feedback data can be obtained through sensors, voice devices, and imaging acquisition devices.

[0065] Perform preprocessing operations on the data, including: denoising, standardization, missing value filling, etc. operations to ensure data quality.

[0066] Provide high-quality input data for subsequent fine-tuning to ensure that the model can obtain real and effective information during training.

[0067] To be able to quickly adapt to different tasks, it is necessary to select a pre-trained foundation model as the core of the interactive agent. Specifically, a foundation model pre-trained on a large-scale medical dataset can be selected, such as GPT, BERT, CLIP, etc. based on the Transformer architecture. The model has strong generality and transfer ability. The knowledge obtained by training this model on a large-scale dataset can provide powerful initial capabilities for specific tasks in a short time.

[0068] By leveraging existing pre-trained models, the training time and computational cost of the model can be significantly reduced, while having strong generalization ability in the medical field.

[0069] Compared with traditional training from scratch, fine-tuning with a pre-trained model can significantly improve efficiency and accuracy. In addition, the model itself can learn a large amount of general medical knowledge, laying a foundation for subsequent personalized optimization.

[0070] In some specific embodiments, a language model based on the Transformer architecture (such as BERT, GPT, etc.) or a multimodal model (such as CLIP) can be selected as the foundation model. This model is pre-trained on a large amount of medical data (such as patient historical health data, treatment feedback, medical literature, etc.).

[0071] The interactive agent includes the following foundation models:

[0072] A text generation model for converting the patient's speech data into text tasks, such as GPT or T5 can be used;

[0073] An image analysis model for analyzing the patient's image data (such as skin lesion images, radiotherapy progress images, etc.), such as Vision Transformer (ViT) or CLIP model can be used;

[0074] A multimodal model for jointly modeling different modalities of data during radiotherapy, such as CLIP.

[0075] Of course, in some other embodiments, the foundation model can also be an RNN / LSTM model or a knowledge graph-based model, etc.

[0076] Although the RNN / LSTM model is slightly inferior to the Transformer in processing large-scale data, similar functions can be achieved through targeted optimization on small-scale medical data.

[0077] Knowledge graph-based models (such as UMLS, SNOMED CT) combined with semantic networks directly provide accurate advice and educational content for patients.

[0078] Compared with large-scale general datasets, small-scale but high-quality radiotherapy-specific datasets can be directly utilized, or multi-domain medical data (such as imaging, drug interactions, health management data, etc.) can be integrated to improve the performance of the model in specific scenarios.

[0079] The present invention fine-tunes the base model using real-time feedback data of patients. Through fine-tuning, the model can be adaptively optimized according to the personalized data of patients (such as emotional state, health feedback), so as to provide more targeted education and health advice.

[0080] Fine-tuning is achieved by retraining the pre-trained base model and adjusting the weights of the model using the feedback data of patients (such as treatment side effects, emotional state, health status, etc.).

[0081] Fine-tuning enables the model to be optimized according to individual needs, thereby improving the accuracy and effectiveness of personalized services. Especially in the medical scenario, fine-tuning can help the intelligent agent respond to the emotional changes of patients in a timely manner, provide emotional support, and optimize the treatment plan.

[0082] Compared with traditional systems that rely on fixed rules or predefined models and cannot perform real-time dynamic adjustments, the present invention uses fine-tuning to optimize the prediction ability and interaction strategy of the base model and provide dynamic personalized services.

[0083] The above fine-tuning methods include but are not limited to: supervised learning fine-tuning methods (such as cross-entropy loss), reinforcement learning fine-tuning methods, unsupervised learning fine-tuning methods, meta-learning fine-tuning methods, transfer learning fine-tuning methods, etc. The present invention optimizes the base model through fine-tuning methods, enabling the interactive intelligent agent to adapt to factors such as the treatment process, health status, and emotional needs of different patients.

[0084] The fine-tuning method is to selectively "freeze" or "unfreeze" certain layers of the model. When training the base model, the lower layers of the model (such as convolutional layers or embedding layers) usually have learned some general features, while the higher layers are responsible for specific tasks. The base model is adjusted by any of the following methods;

[0085] Method 1: Freeze the base layers. Freeze the lower layers of the base model to prevent their weights from being updated, retain the general feature representation of the base model, and reduce the computational overhead.

[0086] Method 2: Unfreeze the task-specific layers. Only fine-tune the higher layers related to the task in the base model to ensure that the model can be optimized for the task.

[0087] For example, given a pre-trained model, the goal of fine-tuning is to optimize the model's performance on the target task, that is, to optimize the model parameters θ. The optimization goal of fine-tuning can be described by the following formula:

[0088]

[0089] Where:

[0090] θ is the parameter of the model;

[0091] L is the loss function, and cross-entropy loss can be used;

[0092] M task (X) is the prediction result of the model on the input X;

[0093] Y is the target label or the true label.

[0094] The supervised learning fine-tuning method is introduced in detail below. For the tasks of patient education and health management during radiotherapy, fine-tuning can be performed based on the patient's specific data (such as symptom feedback, emotional state, health records).

[0095] Input data: the patient's historical health records, treatment feedback, emotional state data, medical images, etc.

[0096] Target output: personalized treatment suggestions, educational materials, emotional support content, etc.

[0097] During the fine-tuning process using the supervised learning fine-tuning method, a labeled dataset is used;

[0098] Use the cross-entropy loss function L cross_entropy Minimize the error between the predicted value of the fine-tuned base model and the actual label. The specific formula is as follows:

[0099]

[0100] Where:

[0101] y i is the actual label;

[0102] p i is the predicted value of the model.

[0103] The reinforcement learning fine-tuning method is introduced in detail below. Online interaction is performed according to the patient's feedback, and the strategy is adjusted according to the patient's behavioral or emotional feedback. In this method, the intelligent agent obtains feedback through interaction with the patient and gradually adjusts its strategy to maximize the long-term reward (such as patient satisfaction, treatment compliance, etc.).

[0104] Reinforcement learning process:

[0105] State: The current health status, emotional state, treatment progress, etc. of the patient.

[0106] Action: Actions that the intelligent agent can take, such as providing health advice, emotional support, or treatment adjustment.

[0107] Reward: Rewards or punishments given based on the patient's feedback (such as improved mood, treatment effect).

[0108] During the fine-tuning process using the reinforcement learning fine-tuning method, the optimization objective is as follows:

[0109]

[0110] Where:

[0111] γ t is the discount factor;

[0112] R t is the reward obtained at time step t;

[0113] T is the total number of time steps of the task;

[0114] θ is a parameter.

[0115] Under this framework, the model will learn how to gradually optimize its behavioral strategy according to the patient's emotional response and treatment feedback to achieve long-term optimal results.

[0116] For the unsupervised learning fine-tuning method, self-supervised learning techniques are used to optimize the model by mining the internal structural relationships of the patient input data (such as the patient's positive / negative feedback on treatment suggestions, emotional changes in continuous interactions).

[0117] For the meta-learning fine-tuning method, meta-learning can train a model with fast transfer ability on multiple small sample datasets, replacing traditional supervised learning and reinforcement learning methods. In the case of rapid changes in patient data, meta-learning can quickly adapt to the data distribution of new patients.

[0118] For the transfer learning fine-tuning method, parameter-efficient fine-tuning methods (such as LoRA or Adapter) are used to adjust only part of the weights of the model, reducing the computational overhead while enhancing the adaptability.

[0119] The present invention enables the interactive intelligent agent to continuously self-optimize according to the actual feedback of the patient through real-time fine-tuning.

[0120] The interaction with the patient is a key step in realizing personalized education and treatment suggestions.

[0121] The present invention uses an interactive agent to interact with patients through natural language processing. According to the patient's input (such as treatment feedback, emotional expression, etc.), the intelligent agent generates corresponding feedback based on the fine-tuned model. These feedbacks can be treatment suggestions, emotional support, health reminders, and other contents.

[0122] Through natural language interaction, patients can communicate with the system more naturally, reducing the operational burden on patients. At the same time, it can be adjusted in real time according to the patient's reaction, improving the patient's compliance and treatment experience.

[0123] Different from traditional medical advice systems, the present invention can adjust the advice content according to the patient's immediate emotional changes and feedback during the interaction process. This ability to flexibly respond to patient needs makes this system more advantageous in practical applications.

[0124] Furthermore, on the basis of the above embodiments, the interactive agent can track the interaction situation of the patient during each radiotherapy process and perform periodic fine-tuning and optimization of the basic model.

[0125] The present invention can achieve long-term effect evaluation and optimization of the interactive agent. By long-term tracking of the patient's treatment effect, emotional state, and health feedback, the interactive agent continuously collects data and performs periodic fine-tuning on the basic model to continuously optimize the service content and interaction strategy. This long-term optimization mechanism ensures that the interactive agent can provide personalized support at all stages of the patient's treatment and adapt to changes in the patient's health status.

[0126] By periodically collecting the patient's treatment data and interaction records, the model parameters can be continuously updated to achieve dynamic optimization, ensuring the long-term adaptability and service quality of the system.

[0127] Furthermore, in order to further optimize the implementation effect of the present invention, in some other embodiments, the remaining characteristic technologies are the same, except that when the interactive agent interacts with the patient, virtual reality (VR) can be used to provide the patient with an immersive health education experience, and the radiotherapy process is simulated through a virtual environment to help the patient understand the treatment plan and relieve anxiety.

[0128] Furthermore, in order to further optimize the implementation effect of the present invention, in some other embodiments, the remaining characteristic technologies are the same, except that when the interactive agent interacts with the patient, a social robot with voice interaction and body movements can be used to provide the patient with a more interactive and emotional support service.

[0129] In order to further optimize the implementation effect of the present invention, in some other embodiments, the remaining characteristic technologies are the same, except that an online learning method is adopted to enable the interactive agent to immediately update the model weights during each patient interaction and optimize the service ability in real time.

[0130] In order to further optimize the implementation effect of the present invention, in some other embodiments, the remaining characteristic technologies are the same, and the difference lies in that imitation learning is adopted to train the interaction agent by simulating the real conversation between the doctor and the patient, so as to improve the naturalness and professionalism of the interaction.

[0131] In summary, through the real-time fine-tuning and long-term optimization mechanisms, and by combining reinforcement learning and supervised learning methods, the present invention can dynamically adjust according to the personalized data of the patient, realizing personalized education and health management in the true sense.

[0132] Furthermore, the present invention can adaptively adjust according to the treatment feedback, emotional state and health data of the patient, rather than relying on fixed rules, significantly improving the flexibility and accuracy of the system in a changing medical environment.

[0133] Compared with the prior art, the present invention is innovative in the combination of emotional support and treatment suggestions, improving the treatment compliance and overall experience of the patient through fine-tuning strategies, and ultimately improving the treatment effect.

[0134] In addition, in some other embodiments, the present invention also discloses a computing device, including:

[0135] One or more processors;

[0136] A memory;

[0137] And one or more programs, where one or more programs are stored in the memory and configured to be executed by one or more processors, and one or more programs include instructions for the method for determining the dominating vertex of the alternating group network disclosed in the above embodiments.

[0138] The processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array). The processor may also include a main processor and a coprocessor. The main processor is a processor used to process data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor may also include an AI (Artificial Intelligence) processor, and the AI processor is used to process computational operations related to machine learning.

[0139] The memory may include one or more computer-readable storage media, and the computer-readable storage media may be non-transitory. The memory may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash storage devices. In some embodiments, the non-transitory computer-readable storage media in the memory is used to store at least one instruction, and the at least one instruction is used to be executed by the processor to implement the method for determining the dominating vertices of the alternating group network provided in the method embodiments of the present invention.

[0140] In addition, the computing device may optionally further include: a peripheral device interface and at least one peripheral device. The processor, the memory, and the peripheral device interface may be connected through a bus or signal lines. Each peripheral device may be connected to the peripheral device interface through a bus, signal lines, or a circuit board. Schematically, the peripheral devices include, but are not limited to: a radio frequency circuit, a touch display screen, an audio circuit, and a power supply, etc.

[0141] Of course, the computing device may also include fewer or more components, and this embodiment does not limit this.

[0142] In addition, in some other embodiments, the present invention also discloses a storage medium, and the storage medium stores one or more computer-readable programs. The one or more programs include instructions, and the instructions are adapted to be loaded and executed by the memory to implement the method for determining the dominating vertices of the alternating group network disclosed in the above embodiments.

[0143] The present invention discloses an intelligent interaction method, a computing device and a storage medium for a patient during radiotherapy, which have the following beneficial effects:

[0144] First, during radiotherapy, the present invention introduces an interactive intelligent agent that can interact with the patient, providing the patient with real-time dynamic personalized health education, treatment advice and emotional support services, and alleviating the patient's emotional problems such as anxiety.

[0145] Second, the present invention introduces an interactive intelligent agent based on a basic model, and the basic model can be adaptively micro-adjusted according to the patient's real-time feedback data, significantly improving the patient's treatment experience and compliance, and at the same time providing more efficient auxiliary support for medical services.

[0146] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification only illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. An intelligent interaction method for patients during radiotherapy, characterized in that: include: Step S1: During radiotherapy, the patient interacts with the interactive agent using natural language; Step S2: collecting multimodal and multidimensional real-time feedback data of patients and preprocessing the data, wherein the feedback data includes: one or more multimodal data of voice data, text data, and image data, and the feedback data includes: one or more multidimensional data of treatment feedback, emotional expression, and health data; Step S3: several pre-trained basic models in the interactive agent are fine-tuned and optimized in real time according to the collected real-time feedback data; Step S4: The fine-tuned and optimized interactive agent generates targeted educational content, health advice, and emotional support based on the patient's real-time feedback data, and outputs it in natural language; Step S5: Repeat steps S1 to S4 until the radiotherapy is completed.

2. The intelligent interaction method according to claim 1, characterized in that: The fine-tuning method includes: one or more of a supervised learning fine-tuning method, a reinforcement learning fine-tuning method, an unsupervised learning fine-tuning method, a meta-learning fine-tuning method, and a transfer learning fine-tuning method.

3. The intelligent interaction method according to claim 1, characterized in that: During fine-tuning using the fine-tuning method, the base model is adjusted using any of the following methods; Method 1: Freeze the lower layers of the base model to prevent its weights from being updated and retain the general feature representation of the base model; The second approach is to fine-tune only the high-level parts of the base model that are relevant to the task.

4. The intelligent interaction method according to claim 2, characterized in that: In the fine-tuning process using supervised learning fine-tuning methods, labeled datasets are used; Use the cross entropy loss function L cross_entropy Minimize the error between the predicted value of the fine-tuned base model and the actual label. The specific formula is as follows: in: y i is the actual label; p i is the predicted value of the model.

5. The intelligent interaction method according to claim 2, characterized in that: During the fine-tuning process using the reinforcement learning fine-tuning method, the optimization objectives are as follows: in: γ t is the discount factor; R t is the reward obtained at time step t; T is the total number of time steps for the task; θ is a parameter.

6. The intelligent interaction method according to any one of claims 1 to 5, characterized in that: The interactive agent includes the following basic models: Text generation models, used to convert patient speech data into text tasks; An image analysis model, used to analyze the patient's image data; Multimodal models are used to jointly model data from different modalities during radiotherapy.

7. The intelligent interaction method according to any one of claims 1 to 5, characterized in that: The interactive agent can track the patient's interactions during each radiotherapy process and perform periodic fine-tuning and optimization on the basic model.

8. The intelligent interaction method according to any one of claims 1 to 5, characterized in that: The preprocessing operation in step S2 includes one or more of denoising, standardization, and missing value filling.

9. A computing device, characterized in that include: one or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, and one or more of the programs include instructions for the intelligent interaction method for the patient during radiotherapy as described in any of claims 1-8.

10. A storage medium, characterized in that The storage medium stores one or more computer-readable programs, and the one or more programs include instructions, and the instructions are suitable for being loaded by the memory and executing the intelligent interaction method for patients during radiotherapy as described in any one of claims 1-8.

Images