A server, a medical imaging scan system and a scan control method thereof

Abstract

The application discloses a kind of server, medical imaging scanning system and its scanning control method, server includes: interactive unit, for receiving the natural language input of user, natural language input characterizes the adjustment demand of user to current scanning protocol;Agent unit is used to determine target task according to the natural language input of user, and target task is planned as multiple target subtasks, and the order combination of multiple target subtasks forms decision tree;According to decision tree, target subtask is executed in sequence, and in response to at least one intermediate result corresponding to target subtask does not satisfy set threshold value, adjust decision tree or re-plan target task, to generate the recommended scanning parameter of current scanning protocol.In addition, the application also provides medical imaging scanning system and the scanning control method of medical imaging scanning system.The embodiments of the application can realize that user uses text description to retrieve scanning protocol, improve examination efficiency.

Description

Technical Field

[0001] This application relates to the field of medical imaging, specifically to a server, a medical imaging scanning system, and a scanning control method for the medical imaging scanning system. Background Technology

[0002] Magnetic Resonance Imaging (MRI), as a non-invasive medical imaging technique, plays a vital role in clinical diagnosis. Adjusting the MRI protocol parameters is a crucial step in the MRI process, directly impacting image quality and diagnostic outcomes. With the continuous development of MRI technology, the scanning sequences and protocol parameters used for imaging are becoming increasingly complex, placing higher demands on the professional knowledge and technical skills of operators.

[0003] In particular, parameter tuning for magnetic resonance imaging (MRI) sequences is a complex task, posing a significant challenge for general users. On one hand, tuning MRI protocol parameters relies heavily on the operator's experience and expertise. To master MRI sequence techniques, radiologists need to invest considerable time in theoretical learning and training. In daily scans, even experienced technicians typically adjust only a few simple parameters, such as field of view, number of slices, slice thickness, slice spacing, saturation band, and flip angle. For complex parameters, they heavily depend on training and technical support from the equipment manufacturer. On the other hand, MRI involves a wide variety of sequences that are constantly being updated, each with its unique technical characteristics, image representation, and clinical applications. This presents a high learning curve for radiologists, requiring significant time and effort to familiarize themselves with and understand. Furthermore, the sheer number of protocol parameters, with their strong and complex interactions and dependencies, makes the optimization and tuning process intricate. An ideal parameter combination must consider not only tissue characteristics and application scenarios but also the performance characteristics of the equipment itself, resulting in situations where even seemingly reasonable parameter settings may lead to unsatisfactory scan results. Therefore, the experience and skill level of the radiology technician are among the key factors determining the scan's effectiveness. Stable MRI scans require operators to possess profound theoretical knowledge and rich practical experience, enabling them to accurately identify and resolve problems. Inexperienced radiology technicians may misunderstand or incorrectly set parameters, making it difficult to determine the optimal protocol for different scenarios, resulting in a significant reduction in scan quality.

[0004] In recent years, some automatic lookup functions have emerged for adjusting MRI protocol parameters, such as matching the patient's height, weight, and other information to a database for the corresponding scanning protocol. While this method considers user-related parameter information, it remains limited to preset parameter selections, lacking flexibility and personalization, and failing to enable natural language communication and feedback between the user and the system. Furthermore, the matching strategy remains unchanged, unable to continuously improve the parameter adjustment strategy based on historical experience and user feedback. Radiology technicians often need to consider numerous factors when adjusting parameters; relying solely on matching patient information still cannot achieve satisfactory scanning results.

[0005] Therefore, there is an urgent need to improve the existing magnetic resonance protocol parameter adjustment methods so that they can understand user needs, provide natural language interaction, and have continuous learning and optimization capabilities, thereby improving the efficiency and quality of magnetic resonance imaging. Summary of the Invention

[0006] The purpose of this application is to provide a server, a medical imaging scanning system, and a scanning control method for the medical imaging scanning system to solve the problems of complex sequence parameter adjustment and strong dependence on operator experience in magnetic resonance imaging.

[0007] To achieve the above objectives, a first aspect of this application provides a server suitable for a medical imaging scanning system. The server includes: an interaction unit for receiving natural language input from a user, the natural language input representing the user's adjustment needs for the current scanning protocol; and an intelligent agent unit for determining a target task based on the user's natural language input, and planning the target task into multiple target sub-tasks, the sequential combination of the multiple target sub-tasks forming a decision tree; executing the target sub-tasks sequentially according to the decision tree, and adjusting the decision tree or re-planning the target task in response to at least one intermediate result corresponding to a target sub-task failing to meet a set threshold, to generate recommended scanning parameters for the current scanning protocol.

[0008] In some embodiments of the first aspect, the intelligent agent includes a long-term memory knowledge base and a short-term memory knowledge base. The long-term memory knowledge base is used to store at least one of the following: prior scanning protocol, parameter tuning template, parameter tuning rules, and parameter tuning priority. The short-term memory knowledge base is used to store at least one of the following: real-time feedback from the user, real-time status of the object being detected, and real-time status information of the medical imaging scanning system.

[0009] In some embodiments of the first aspect, the server further includes: an update unit for updating the long-term memory knowledge base based on at least one of recommended scanning parameters of the current scanning protocol, the decision tree, and the user's real-time feedback.

[0010] In some embodiments of the first aspect, the server further includes a tool unit capable of performing at least one of artifact detection, motion monitoring, protocol preparation, SAR estimation, PNS estimation, patient cooperation estimation, and physiological signal capture; the agent unit is capable of calling the tool unit to complete the target sub-task.

[0011] In some embodiments of the first aspect, the recommended scanning parameters of the current scanning protocol include a first set of recommended parameters and a second set of recommended parameters. The intelligent agent unit is also used to generate a first image quality estimate corresponding to the first set of recommended parameters and a second image quality estimate corresponding to the second set of recommended parameters. The image quality indicators corresponding to the first image quality estimate and the second image quality estimate are different.

[0012] In some embodiments of the first aspect, the interactive unit simultaneously displays the initial scanning parameters and recommended scanning parameters of the current scanning protocol.

[0013] A second aspect of this application provides a medical imaging scanning system, comprising: a scanner on which a user performs a scan of a target object; a user interface configured to receive natural language input from a user, the natural language input representing the user's adjustment requirements for the current scanning protocol; a processor; and a non-transient computer-readable storage medium operatively communicating with the processor, wherein the non-transient computer-readable storage medium contains one or more programming instructions that, when executed, cause the processor to:

[0014] The target task is determined based on the user's natural language input, and the target task is planned into multiple target sub-tasks. The sequential combination of multiple target sub-tasks forms a decision tree. According to the decision tree, the target sub-tasks are executed sequentially. If the intermediate result corresponding to at least one target sub-task does not meet the set threshold, the decision tree is adjusted or the target task is re-planned to generate recommended scanning parameters for the current scanning protocol. The recommended scanning parameters for the current scanning protocol are displayed through the user interaction interface. In response to the user's confirmation operation on the user interaction interface, the scanner is controlled to execute the current scanning protocol according to the recommended scanning parameters.

[0015] In some embodiments of the second aspect, the processor includes an intelligent agent, which includes a long-term memory knowledge base and a short-term memory knowledge base. The long-term memory knowledge base is used to store at least one of a priori scanning protocol, parameter tuning template, parameter tuning rules, and parameter tuning priority. The short-term memory knowledge base is used to store at least one of real-time user feedback, real-time status of the object being detected, and real-time status information of the medical imaging scanning system.

[0016] A third aspect of this application provides a scanning control method for a medical imaging scanning system, comprising: receiving natural language input from a user, the natural language input representing the user's adjustment requirements for the current scanning protocol; determining a target task based on the user's natural language input; planning the target task into multiple target sub-tasks, and sequentially combining the multiple target sub-tasks to form a decision tree; executing the target sub-tasks sequentially according to the decision tree, and adjusting the decision tree or re-planning the target task in response to at least one intermediate result corresponding to a target sub-task failing to meet a set threshold, to generate recommended scanning parameters for the current scanning protocol; and controlling the medical imaging scanning system to execute the current scanning protocol according to the recommended scanning parameters to complete the imaging scan of the object to be detected.

[0017] In some implementations of the third aspect, the adjustment requirements for the current scanning protocol include at least one of: improving the quality of the scan images corresponding to the current scanning protocol, increasing the scanning speed, meeting clinical requirements, and correcting user operation errors.

[0018] The beneficial effects of this application are as follows: The target task is determined based on the user's natural language input. By introducing a scanning protocol agent based on a large language model, the user can directly interact with the scanning protocol agent through natural language to communicate their needs for the scanning protocol, significantly reducing the learning threshold for MRI parameter adjustment. The target task is planned into multiple sub-tasks, and the sequential combination of these sub-tasks forms a decision tree. The sub-tasks are executed sequentially according to the decision tree. Based on the user's needs, the strategy is intelligently and dynamically adjusted using relevant knowledge, standards, and tools, effectively solving the problem of complex protocol parameter optimization. In response to at least one sub-task's intermediate result failing to meet a set threshold, the decision tree is adjusted or the target task is replanned. The scanning protocol agent can learn and optimize based on user feedback, continuously improving the accuracy and efficiency of parameter adjustment, achieving continuous system improvement. The scanning protocol agent can be automatically executed after user confirmation, supporting one-click scanning, simplifying the operation process, improving work efficiency, and reducing the possibility of human error. This application, through the assistance of the agent, reduces the reliance on the experience and technical level of radiology technicians, enabling even inexperienced operators to obtain better scanning results, improving the overall quality and consistency of MRI examinations. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a server module diagram for a medical imaging scanning system provided in Embodiment 1 of this application;

[0021] Figure 2 This is a schematic diagram of the medical imaging scanning system provided in Embodiment 2 of this application;

[0022] Figure 3 This is a schematic diagram of the construction method of the intelligent agent of the medical imaging scanning system provided in Embodiment 3 of this application;

[0023] Figure 4 This is a schematic flowchart of the scanning control method for the medical imaging scanning system provided in Embodiment 4 of this application;

[0024] Figure 5 This is a schematic flowchart of the scanning control method for the medical imaging scanning system provided in Embodiment 5 of this application;

[0025] Figure 6 This is a block diagram of the intelligent agent included in the medical imaging scanning system provided in Embodiment 5 of this application;

[0026] Figure 7 This is a schematic diagram of the user interaction interface provided in Embodiment Six of this application. Detailed Implementation

[0027] Magnetic resonance imaging (MRI) involves a wide variety of sequences, including fast imaging sequences, 3D imaging sequences, motion artifact correction sequences, fat suppression sequences, steady-state free precession sequences, and dual-excitation planar echo imaging sequences. Each sequence has its unique technical characteristics, image performance, and clinical applications. With improvements in hardware, new types of sequences are constantly being introduced. This complexity presents a high learning curve for radiology technicians, requiring them to invest significant time and effort to become familiar with and understand them. Furthermore, the protocol parameters are numerous and have complex interactions and dependencies. An ideal parameter combination must consider not only tissue characteristics and application scenarios but also the performance characteristics of the equipment itself. This makes the optimization and adjustment of protocol parameters intricate, and even with seemingly reasonable parameter settings, scan results may still be unsatisfactory. Magnetic resonance sequence parameter adjustment is a complex task and presents considerable challenges for general users. Inexperienced operators may misunderstand or incorrectly set parameters, leading to failure to achieve image optimization goals and significantly reduced scan results. To address these issues, this application proposes the following improvements to existing magnetic resonance protocol parameter adjustment methods:

[0028] First, construct an intelligent agent. This agent is based on a large language model as the brain for understanding and planning. It has the ability to understand user needs, communicate with users, and has the logical reasoning ability required for protocol tuning. It optimizes protocol parameters based on relevant knowledge and tools. It can intelligently learn and optimize parameter tuning strategies based on user feedback.

[0029] Secondly, the intelligent agent has short-term and long-term memory (knowledge base) required for the tuning protocol, including general descriptions of all basic magnetic resonance protocols, parameter lists and descriptions, tuning templates and examples, tuning rules and priorities, basic magnetic resonance physics knowledge (physical constraints and related formulas), equipment parameters and hardware and software constraints, user input and feedback, patient status, system status and system information, etc., one or more of the following combinations:

[0030] Furthermore, the agent is equipped with the tools required for the protocol tuning. These tools may include one or more of the following combinations: artifact detection, motion monitoring, sequence preparation, specific absorption rate (SAR) calculation, peripheral nervous system (PNS) stimulation calculation, patient compliance estimation, and physiological signal capture.

[0031] Finally, the protocol agent is equipped with the system permissions required to perform parameter tuning. After user confirmation, the protocol agent can automatically apply the parameter tuning results, enabling users to start scanning the adjusted protocol with one click.

[0032] When users such as clinicians and radiology technicians need to adjust protocols, they can directly interact with the intelligent agent in this application through natural language to communicate their needs for the protocols. The intelligent agent can intelligently plan parameter adjustment strategies based on the user's needs, using relevant knowledge, standards and tools for protocol adjustment, and automatically apply the new protocol after user confirmation to meet the user's needs for optimizing scan parameters.

[0033] The following describes specific embodiments of this application. It should be noted that, in order to maintain brevity, this specification cannot provide a detailed description of all features of the actual embodiments. It should be understood that, in the actual implementation of any embodiment, just as in any engineering or design project, various specific decisions are often made to achieve the developer's specific goals and to meet system-related or business-related constraints, and this can change from one embodiment to another. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, some design, manufacturing, or production modifications based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient content of this application.

[0034] References to "an embodiment," "an embodiment," "an exemplary embodiment," etc., in the specification indicate that the described embodiment may include a specific feature, structure, or characteristic; however, not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, such phrases do not necessarily refer to the same embodiment. Additionally, when a specific feature, structure, or characteristic is described in connection with an embodiment, it is believed that the influence of such feature, structure, or characteristic on such feature, structure, or characteristic in conjunction with other embodiments, whether explicitly described or not, is within the knowledge of those skilled in the art.

[0035] Example 1

[0036] This application provides a server suitable for medical imaging scanning systems. For example... Figure 1As shown, the server includes an interaction unit 102 and an agent unit 104. The medical imaging scanning system can be an X-ray imaging device, a computed tomography (CT) device, a magnetic resonance imaging (MRI) device, a positron emission tomography (PET) imaging device, a single-photon emission computed tomography (SPECT) imaging device, and a multimodal imaging system, such as a PET / CT, PET / MR, or SPECT / CT imaging system. The interaction unit 102 and the agent unit 104 can be coupled via one or more configurable wired and / or wireless networks (such as the Internet and / or a virtual private network, wireless telephone network, wireless local area network, wired local area network, wireless wide area network, wired wide area network, etc.). The server can be a Picture Archiving and Communication System (PACS) system. This server is used to store, manage, and transmit medical image data. It receives, stores, and retrieves digital images generated by the aforementioned medical imaging scanning system, and transmits these images to the workstations of doctors or other medical professionals via the network for viewing and analysis. When needed, users can quickly retrieve and use the images with certain authorization. The PACS system also includes auxiliary diagnostic management functions.

[0037] The interaction unit 102 is used to receive natural language input from users, which represents the user's adjustment needs for the current scanning protocol. The interaction unit 102 may include input devices such as a keyboard, mouse, pen, voice input device, and touchscreen. It may also include output devices such as display devices, speakers, and printers. Users may include personnel operating the medical imaging scanning system, such as radiology technicians, clinicians, and operators. Optionally, the interaction unit 102 can be connected to a Radiology Information System (RIS), which has multiple modules such as a registration and appointment module, a queuing and calling module, a technician module, and a reporting module.

[0038] The registration and appointment module allows patients to directly enter their information or retrieve it from the RIS database by means of their name, etc.

[0039] The queuing and calling module can pre-assign examination tasks in the task list, mark them as appointment tasks, and set the examination tasks according to items such as imaging equipment, examination items, examining physicians, patient sources, and appointment time slots.

[0040] The technician module transmits the examination tasks requested by the equipment to the equipment, intuitively displays the waiting status, and tracks the examination progress; or, according to the examination status, it changes the corresponding attributes of the patient.

[0041] The reporting module allows users to write reports and import patient text information or enter patient image information.

[0042] The intelligent agent unit 104, specifically a computing system or similar electronic computing device, manipulates and transforms data representing physical (electronic) quantities within the computing system's memory, registers, or other such information storage, transmission, or display devices. The intelligent agent unit 104 determines a target task based on the user's natural language input and plans the target task into multiple target sub-tasks. The sequential combination of these sub-tasks forms a decision tree. Based on the decision tree, the target sub-tasks are executed sequentially. If an intermediate result corresponding to at least one target sub-task fails to meet a set threshold, the decision tree is adjusted or the target task is replanned to generate recommended scanning parameters for the current scanning protocol. The scanning protocol can be gradient field parameters, radio frequency field parameters, and acquisition timing related to MR scanning. In CT scanning, the scanning protocol can be tube voltage values, tube current values, scanning modes such as helical scanning, plain scanning, or contrast-enhanced scanning, image cycle time, acquisition time, contrast agent regimen, etc.

[0043] The agent unit 104 may include multiple cascaded agents. For example, the agent unit 104 may include agent 1, agent 2, and agent 3. The following example illustrates this by having a user propose improvements to the current scan results after completing a scan.

[0044] Agent 1 is able to rewrite the question corresponding to the user's natural language input:

[0045] 1) Once the problem is identified, check if the user has corrected it. If so, rewrite the problem as the corrected one. Check if the user has added any information to the problem. If so, rewrite the problem with the added information.

[0046] 2) Remove redundant information and check the accuracy of technical terms and professional vocabulary to ensure that these terms are not miswritten or omitted during the rewriting process. Removing redundant information may include deleting repetitive and verbose content or deleting modifiers without semantic meaning.

[0047] 3) Combine the original scanning protocol with the results of 2) to identify the protocol name and scanning area that need to be optimized.

[0048] 4) While ensuring the original intent of the user's input, output the rewritten question.

[0049] Agent 2 connects with Agent 1 and can retrieve relevant information from the database based on the rewritten question. Furthermore, based on this relevant information, it extracts target subtasks that can solve the user's problem.

[0050] 1) Understand the rewritten question, search the database based on the rewritten question, and find content related to the user's question to obtain relevant search information;

[0051] 2) Extract target sub-tasks that can solve user problems from the retrieved relevant information, list the extracted target sub-tasks by item, and form a decision tree.

[0052] Agent 3 connects with Agent 2 to adjust parameter values ​​in the current scanning protocol based on the decision tree.

[0053] 1) Identify the parameters that need to be adjusted, and confirm the position and current value of these parameters in the current scanning protocol. In this embodiment, only parameters that are explicitly present in the current scanning protocol can be adjusted.

[0054] 2) Based on the decision tree, determine the new value for each parameter, ensuring that the new value meets the requirements of the parameter adjustment strategy and is within a reasonable range;

[0055] 3) Output the adjusted parameter values. In this embodiment, the output result includes the adjusted parameters and their new values. In one embodiment, the scanning protocol takes an MR scanning protocol as an example. The scanning parameters of the scanning protocol may include one or more parameters such as the type of magnetic resonance pulse sequence, TR, number of repetitions, TI, etc. The scanning parameters may include one or more parameters such as the radio frequency center frequency, flip angle, etc. In some embodiments, the scanning parameters may include one or more other imaging parameters, such as the radio frequency channel count (or number), image contrast and / or ratio, slice thickness, imaging type (e.g., T1-weighted imaging, T2-weighted imaging, proton density-weighted imaging, etc.), field of view (FOV), off-center frequency offset or similar parameters, or combinations thereof.

[0056] The intelligent agent comprises a long-term memory knowledge base and a short-term memory knowledge base. The long-term memory knowledge base stores information retained for extended periods, which can be stored and retrieved using a vector database. In this embodiment, the long-term memory knowledge base stores one or more combinations of prior scanning protocols, parameter tuning templates, parameter tuning rules, and parameter tuning priorities. The short-term memory knowledge base includes context information generated and temporarily stored during task execution, which is cleared upon task completion. In this embodiment, the short-term memory knowledge base stores one or more combinations of user real-time feedback, the real-time status of the detected object, and the real-time status information of the medical imaging scanning system. User real-time feedback can include the user's clinical needs, user suggestions regarding the current results, and user modification requests.

[0057] The server also includes a tool unit 106, which can perform one or more combinations of artifact detection, motion monitoring, protocol preparation, SAR estimation, PNS estimation, patient cooperation estimation, and physiological signal capture. The agent unit 104 can call the tool unit to complete the target sub-task, that is, the execution of the target sub-task is affected by artifact detection, whether there is motion, whether the protocol is prepared, SAR value, PNS value, patient cooperation, and physiological signals.

[0058] The current scanning protocol's recommended scanning parameters include a first set of recommended parameters and a second set of recommended parameters. The intelligent agent unit 104 is also used to generate a first image quality estimate corresponding to the first set of recommended parameters and a second image quality estimate corresponding to the second set of recommended parameters, wherein the image quality indices corresponding to the first and second image quality estimates are different. Medical image quality indices can be signal-to-noise ratio, contrast, resolution, artifacts, and the corresponding scan time. For example, the first image quality estimate corresponding to the first set of recommended parameters reflects optimal signal-to-noise ratio, and the second image quality estimate corresponding to the second set of recommended parameters reflects optimal resolution. Of course, the recommended scanning parameters of the scanning protocol may also include a third set of recommended parameters and a fourth set of recommended parameters, wherein: the third image quality estimate corresponding to the third set of recommended parameters reflects optimal contrast; and the fourth image quality estimate corresponding to the fourth set of recommended parameters reflects the shortest scan time. Users can choose according to the image quality indices they are interested in.

[0059] The interactive unit 102 simultaneously displays the initial scanning parameters and recommended scanning parameters of the current scanning protocol. Users can observe which parameters of the scanning protocol have changed.

[0060] The server also includes an update unit 108, which is used to update the long-term memory knowledge base based on recommended scanning parameters of the current scanning protocol, decision trees, real-time user feedback, or user needs. The update unit 108 can update the long-term memory knowledge base based on the recommended scanning parameters or decision trees of the current scanning protocol to verify the validity of scanning parameters, perform iterative optimization of the model, or extend protocol compatibility.

[0061] Example 2

[0062] This application provides a medical imaging scanning system, such as... Figure 2 The system includes a scanner 202, a user interface 204, a processor 206, and a non-transient computer-readable storage medium 208. The medical imaging scanning system can perform X-ray imaging, CT imaging, MR imaging, PET imaging, SPECT imaging, etc. The scanner 202 and user interface 204 can be connected to the processor 206 via a network. Alternatively, the scanner 202 and user interface 204 can be directly connected to the processor 206.

[0063] Scanner 202 is used to perform scanning on the object to be detected. In this embodiment, scanner 202 is described as an MR scanner. The MR scanner includes a magnet assembly, a gradient coil assembly, and a radio frequency coil assembly. The magnet assembly can generate a main magnetic field. The magnet assembly is a superconducting magnet, which, after excitation, excites nuclear spins in the patient's body in the detection area.

[0064] The gradient coil assembly, located inside the magnet assembly, generates gradient field components Gx, Gy, and Gz in different directions. These gradient fields can be superimposed on the main magnetic field generated by the magnet assembly, allowing the magnetic orientation of the object's protons to vary according to their position within the gradient field, thereby encoding spatial information into the MR signal generated by the imaged object region. The gradient coil assembly may include an X-coil for generating the gradient field component Gx in the x-direction; a Y-coil for generating the gradient field component Gy in the y-direction; and a Z-coil for generating the gradient field component Gz in the z-direction. In some embodiments, the Z-coil may be designed based on a Maxwell coil, while the X and Y coils may be designed based on a Golay coil configuration. The three sets of coils execute parameters related to the gradient pulses in the scanning protocol to generate three distinct magnetic field components for position encoding. The gradient coil assembly allows for spatial encoding of the MR signal used for image reconstruction.

[0065] An RF coil assembly may include multiple RF coils. Each RF coil may include one or more RF transmitting coils and / or one or more RF receiving coils, each RF coil including one or more RF receiving coil units. The RF transmitting coil may transmit RF pulses toward an object. Under the combined action of a main magnetic field, a gradient magnetic field, and the RF pulses, an object-related MR signal may be generated in response to the pulse sequence. The RF receiving coil may acquire the MR signal from the object. The pulse sequence may be defined by imaging parameters and settings associated with the imaging parameters in the time series. In some embodiments, the imaging parameters may include parameters related to the RF pulses emitted by the RF coils, such as the number of excitations (NEX), bandwidth, etc.; parameters related to the gradient field generated by the gradient coils (e.g., gradient direction, duration of gradient application, etc.); and parameters related to the MR signal (e.g., echo time (TE), echo sequence length (ETL), etc.). In some embodiments, the pulse sequence may be defined by one or more time-related parameters, such as repetition time (TR), acquisition time (TA), etc.

[0066] One or more RF coils can both transmit RF pulses and receive MR signals at different times. In some embodiments, the function, size, type, geometry, location, number, and / or amplitude of the RF coils can be determined or changed according to one or more specific conditions. For example, RF coils can be classified as volumetric coils and local coils depending on their function and size. In some embodiments, the RF receiving coil unit can correspond to a channel used for acquiring MR signals.

[0067] User interaction interface 204 is configured to receive natural language input from a user, wherein the natural language input represents the user's adjustment requirements for the current scanning protocol. User interaction interface 204 may include a mobile device, tablet, laptop, or similar device, or a combination thereof. Optionally, user interaction interface 204 may include input devices, output devices, etc. Optionally, user interaction interface 204 may be part of scanner 202.

[0068] A non-transient computer-readable storage medium 208 is operatively communicateable with a processor 206, wherein the non-transient computer-readable storage medium contains one or more programming instructions. The non-transient computer-readable storage medium 208 may contain a propagated data signal containing computer program code, for example, on baseband or as part of a carrier wave. Such a propagated signal can take many forms, including electromagnetic, optical, etc., or any suitable combination. The non-transient computer-readable storage medium 208 can be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, apparatus, or device to enable communication, propagation, or transmission of a program for use. Program code located on a computer-readable signal medium can be propagated through any suitable medium, including radio, cable, fiber optic cable, RF, etc., or any combination of the above media.

[0069] The processor 206 is capable of executing one or more programming instructions to: determine a target task based on the user's natural language input, and plan the target task into multiple target subtasks, the sequential combination of multiple target subtasks forming a decision tree; execute the target subtasks sequentially according to the decision tree, and in response to at least one target subtask's intermediate result failing to meet a set threshold, adjust the decision tree or re-plan the target task to generate recommended scanning parameters for the current scanning protocol; control the user interface to display the recommended scanning parameters for the current scanning protocol, and in response to the user's confirmation operation on the user interface, control the scanner to execute the current scanning protocol according to the recommended scanning parameters.

[0070] User input in natural language could be something like, "A brain scan is needed for an elderly patient who cannot remain still for extended periods, with a focus on the hippocampus." The large language model analyzes the user input, extracting key information such as the scan site, specific requirements, and areas of interest, clarifying details of the needs through questioning when necessary. This natural language can be speech, text, or a scan order from the RIS system. User input could also include phrases like, "I want to increase the scanning speed for stroke patients," "I want to improve gray-white matter contrast," "I want to further improve image quality," "patient movement," or "abnormal operation," etc.

[0071] The current scanning protocol uses magnetic resonance pulse sequences (also known as magnetic resonance imaging protocols) as an example. Types of magnetic resonance pulse sequences can include spin echo (SE) pulse sequences, gradient focused echo (GRE) pulse sequences, inversion recovery (IR) pulse sequences, multi-echo magnetic resonance pulse sequences, T1ρ preprocessing pulse sequences, T2 preprocessing pulse sequences, diffusion-weighted imaging (DWI) pulse sequences, etc. These magnetic resonance pulse sequences can be defined by one or more parameters, such as the type of magnetic resonance pulse sequence, the time at which the magnetic resonance pulse sequence is applied, the duration of the magnetic resonance pulse sequence, parameters related to radio frequency pulses in the magnetic resonance pulse sequence, the count of radio frequency pulses in the magnetic resonance pulse sequence, repetition time (TR), number of repetitions, invert time (TI), b-value, T1ρ preparation duration, T2 preparation duration, echo sequence length, echo interval, velocity encoding (VENC) value, and the count of average values, etc. The TR used in this specification can refer to the time interval between two repetitive and consecutive radio frequency pulses in a magnetic resonance pulse sequence (e.g., the time interval between two consecutive excitation radio frequency pulses in an SE pulse sequence, or the time interval between two consecutive 180° inversion pulses in an IR pulse sequence). The number of repetitions can refer to the count of repetitions in the magnetic resonance pulse sequence. TI can refer to the time interval between a 180° inversion pulse and a subsequent 90° excitation pulse in an IR pulse sequence. The b-value can refer to a factor reflecting the intensity and duration of the diffusion sensitization gradient in a DWI pulse sequence. T1ρ preparation duration can refer to the duration of the spin-locked pulse in a T1ρ preparation pulse sequence. T2-preparation duration can refer to the duration of the T2 preparation pulse in a T2 preparation pulse sequence.

[0072] The processor 206 includes an intelligent agent, which contains a long-term memory knowledge base and a short-term memory knowledge base. The long-term memory knowledge base stores information that is retained for extended periods, and this information can be stored and retrieved using a vector database. In this embodiment, the long-term memory knowledge base stores one or more combinations of prior scanning protocols, parameter tuning templates, parameter tuning rules, and parameter tuning priorities. The short-term memory knowledge base includes context information generated and temporarily stored during the execution of subtasks, and cleared after the task is completed. In this embodiment, the short-term memory knowledge base is used to store one or more combinations of real-time user feedback, the real-time status of the detected object, and the real-time status information of the medical imaging scanning system.

[0073] The processor 206 may include a function module or a function call module. The function module or function call module can perform functions such as artifact detection, motion monitoring, protocol preparation, SAR estimation, PNS estimation, patient cooperation estimation, and physiological signal capture to complete the target sub-task. That is, the execution of the target sub-task is affected by artifacts, whether there is motion, whether the protocol preparation is completed, SAR value, PNS value, patient cooperation, and physiological signals.

[0074] The current scanning protocol includes multiple sets of recommended scanning parameters. The processor 206 generates an image quality estimate for each set of recommended parameters, with different image quality estimates corresponding to different image quality metrics. Medical image quality metrics can include signal-to-noise ratio (SNR), contrast ratio, resolution, artifacts, and the corresponding scan time. For example, one set of recommended parameters might reflect optimal SNR, while another set might reflect optimal resolution. Alternatively, one set might reflect optimal contrast, and yet another set might reflect the shortest scan time. Users can choose the appropriate image quality metric based on their preferred image quality. Optionally, the initial and recommended scanning parameters of the current scanning protocol can be simultaneously displayed on the user interface 204. Users can observe which parameters of the scanning protocol have changed.

[0075] Processor 206 also updates the long-term memory knowledge base based on one or more combinations of recommended scan parameters of the current scanning protocol, decision trees, and real-time user feedback. For example, processor 206 can update the long-term memory knowledge base based on recommended scan parameters or decision trees of the current scanning protocol to verify the validity of scan parameters, perform iterative optimization of the model, or extend protocol compatibility.

[0076] Example 3

[0077] This application discloses a method for constructing an intelligent agent in a medical imaging scanning system. Using an MR system as an example, the constructed intelligent agent enables intelligent adjustment of magnetic resonance imaging protocol parameters, improving the quality and efficiency of magnetic resonance imaging. The construction process of the intelligent agent is as follows: Figure 3 Specifically, it may include:

[0078] Step S301: Initially construct an intelligent agent (protocol agent) based on a large language model. For example, a pre-trained large language model is selected as the core component of the protocol agent. This large language model has a large number of parameters, adopts a Transformer architecture, and is pre-trained on a certain amount of text data, including medical literature, magnetic resonance physics textbooks, and clinical cases, using a self-supervised learning method. After pre-training, fine-tuning is performed using self-built magnetic resonance protocol parameter tuning dialogue data, enabling the model to understand magnetic resonance imaging terminology and protocol parameter adjustment knowledge. Based on a large language model, the intelligent agent, after specific optimization, possesses the following capabilities: natural language understanding ability, enabling it to understand the needs expressed in natural language by users such as radiologists and clinicians, including scan sites, target lesions, and image quality requirements; multi-turn dialogue ability, supporting continuous multi-turn dialogues with users, clarifying needs, explaining parameter adjustment principles, and providing suggestions during the dialogue process; logical reasoning ability, enabling it to reason about the impact of different parameter adjustments on factors such as image quality and scan time based on the principles of magnetic resonance physics; knowledge application ability, enabling it to apply professional knowledge of magnetic resonance imaging, such as K-space sampling principles, radio frequency pulse sequences, and gradient pulse sequence characteristics, to guide parameter adjustments; and learning optimization ability, continuously optimizing parameter tuning strategies by recording user feedback and parameter tuning results.

[0079] In this embodiment, the agent's large language model adopts a Mixture of Experts architecture, comprising 16 expert networks. Each expert network specializes in handling specific types of magnetic resonance sequence parameter adjustments, such as T1-weighted imaging, T2-weighted imaging, diffusion-weighted imaging, and functional magnetic resonance imaging. A gating network dynamically selects the most suitable combination of expert networks for the current task, improving the accuracy of parameter adjustments.

[0080] Step S303: Construct the agent's memory system. The agent's memory system can be divided into short-term memory and long-term memory, used to store and manage various types of information required for magnetic resonance protocol adjustment.

[0081] Optionally, short-term memory employs an attention-based vector database to store information from the current session. The short-term memory can include user input requests and feedback, such as "wanting to reduce scan time while maintaining image quality" or "the patient cannot remain still for extended periods." It can also include the patient's current status information, including age, weight, presence of implants, and claustrophobia. Furthermore, it can store the current system status of the medical imaging scanning system, including magnet temperature, gradient coil status, and radio frequency coil selection. Finally, it can contain the history of scan parameters adjusted during the current session. In this embodiment, short-term memory uses a decay mechanism based on importance scoring, retaining important information for a longer period, with a maximum retention time of 24 hours after the end of a single session.

[0082] Optionally, long-term memory can be a combination of a hierarchical knowledge graph and a relational database. The long-term memory component could be a basic MRI protocol library, containing detailed descriptions of at least several commonly used MRI scanning protocols. Each protocol includes scanning parameters, their default values, ranges, and physical meanings. The long-term memory component could also include a list and description of scanning parameters, detailing the physical meaning of each parameter, adjustment methods, and its relationship with other scans. Furthermore, the long-term memory component could contain parameter tuning templates and examples of these templates. Different templates would be set for parameters with different clinical needs, such as "T2 sequence adjustment template to reduce motion artifacts" or "diffusion-weighted imaging adjustment template to improve the detection rate of small lesions." The long-term memory component could also contain parameter tuning rules and priorities, such as "prioritizing shorter scan times while ensuring diagnostic quality" or "prioritizing lower SAR values ​​for pediatric patients." Finally, the long-term memory component could be a magnetic resonance physics knowledge base, containing fundamental physical principles of MRI, characteristics of various sequences, image contrast formation mechanisms, and related physical limitations and calculation formulas. The long-term memory can include equipment parameters and limitations, such as hardware parameters, software version limitations, and gradient performance limits for different models of MRI equipment. Another long-term memory component is historical user feedback, which records past parameter tuning results and user feedback for continuous optimization of the tuning strategy.

[0083] In this embodiment, long-term memory uses a graph database to store the relationships between knowledge points, and uses a knowledge graph to represent and reason about complex knowledge. The long-term memory content is updated regularly, incorporating the latest magnetic resonance imaging research findings and clinical practice experience.

[0084] Step S305: Configure the agent's toolset. The agent is equipped with a variety of specialized tools to assist in decision-making regarding protocol parameter adjustments.

[0085] Optionally, the agent can be equipped with an artifact detection tool. This tool employs a convolutional neural network architecture and can automatically classify the type and severity of artifacts in MR images. Trained on labeled MR sample images, this tool can identify common artifacts such as motion artifacts, folding artifacts, chemical shift artifacts, and magnetic susceptibility artifacts. The artifact detection results are used to guide scanning parameter adjustments; for example, when motion artifacts are detected, the MR system may suggest shortening the acquisition time or using anti-motion techniques.

[0086] Optionally, the agent can be equipped with a motion monitoring tool that combines optical sensors and real-time image analysis technology to monitor the patient's movements during the scan. This tool can detect minute movements and generate motion reports based on the amplitude and frequency of those movements. The motion monitoring data is used to assess the suitability of the current scanning protocol, recommending adjustments to a faster acquisition sequence or enabling motion correction techniques if necessary.

[0087] Optionally, the agent can be equipped with a protocol preparation tool that automatically selects the most suitable basic protocol from a protocol library as a starting point for adjustments based on the currently selected scan site and clinical needs. This tool employs a hybrid approach based on rules and similarity calculations, considering factors such as clinical objectives, patient characteristics, and device status to select the most suitable protocol template from the protocol library.

[0088] Optionally, the agent can be equipped with a SAR estimation tool that predicts SAR values ​​during the scan based on the patient's weight, height, age, and the scanning parameters of the selected scan sequence. This tool uses a combination of finite element analysis and statistical models to predict the SAR values ​​corresponding to the scanning parameters. When the predicted SAR value approaches the safety limit (e.g., 4 W / kg for adults and 3 W / kg for children), the MR system will automatically suggest adjusting the scanning parameters to reduce the SAR value.

[0089] Optionally, the agent can be equipped with a PNS estimation tool that predicts the degree of peripheral nerve stimulation that may occur during the scan based on gradient parameters and patient body size. This tool uses a bioelectromagnetic model combined with a predictive algorithm built from clinical data to predict the probability of the patient experiencing nerve stimulation under different gradient parameters. When the PNS risk exceeds a moderate level, the MR system will suggest adjusting the gradient parameters or selecting an alternative sequence.

[0090] Optionally, the agent can be equipped with a patient compliance estimation tool. This tool assesses a patient's ability to remain still during the scan by analyzing information such as age, physical condition, and previous scanning history. Using machine learning algorithms, the tool categorizes patient compliance into high, medium, and low levels, providing a reference for protocol selection. For patients with low compliance, the MR system will prioritize recommending fast imaging sequences or motion-insensitive sequences.

[0091] Optionally, the agent can be equipped with a physiological signal capture tool that collects physiological signals such as the patient's respiration and heartbeat using MRI-compatible sensors to trigger scans or post-processing corrections. The physiological signal capture tool supports real-time respiratory gating, ECG gating, and pulse gating. The physiological signal data can be used to optimize sequence parameters, such as adjusting respiratory gating parameters based on respiratory rate.

[0092] The aforementioned tools are integrated with the agent's large language model through standardized APIs. The invocation of these tools is automatically decided by the large language model based on context and requirements. The execution results of the tools are returned to the large language model in the form of structured data for subsequent parameter adjustment decisions.

[0093] Step S307: Configure the system permissions for the intelligent agent. System permissions can be hierarchical, used to perform different levels of protocol parameter adjustment operations.

[0094] Optionally, the system permission is query permission. This query permission allows the protocol agent to query information such as the current status of the MR system, loaded protocol parameters, hardware configuration, and software version. This permission can be executed without user confirmation and is used to obtain basic system information to provide a basis for parameter adjustment.

[0095] Optionally, the system permission is parameter calculation permission. This permission allows the protocol agent to calculate the theoretical effects of different parameter combinations, such as signal-to-noise ratio, contrast, spatial resolution, and scan time, based on the physical model and system constraints. This permission can be executed without user confirmation and is used to generate parameter adjustment suggestions.

[0096] Optionally, the system permission is the parameter modification suggestion permission. This permission allows the agent to generate parameter modification suggestions and display them to the user in a visual manner, showing information such as a comparison before and after parameter adjustment, and expected changes in effect. This permission requires user review and confirmation before proceeding to the next execution stage. Figure 7 As shown, the displayed information includes three different sets of parameters, each corresponding to a different optimization metric.

[0097] Optionally, the system permission is parameter modification execution permission. This permission allows the agent to apply the adjusted parameters to the MR system after user confirmation. For example, the parameter modification execution permission adopts a dual confirmation mechanism: first, the agent generates a parameter modification plan, and then the user authorizes execution by issuing an explicit "confirmation" operation through the interaction interface.

[0098] Optionally, the system permission is scan initiation permission. Scan initiation permission allows the protocol agent to start the scan sequence with one click after completing parameter adjustments and obtaining user confirmation. This permission employs the highest level of security measures. In addition to user confirmation, the system also needs to verify that all security parameters (such as SAR values, gradient limits, etc.) are within the security range.

[0099] In this embodiment, the system permission management can adopt a role-based access control model. Users of different levels, such as chief physicians, general physicians, and technicians, can grant different levels of permissions to the intelligent agent, and all permission operations will be recorded in the system log to ensure that the operations are traceable.

[0100] Step S309: Optimize the agent intermittently or continuously.

[0101] Optionally, the agent achieves continuous learning and optimization through a user feedback learning mechanism: recording user feedback on the results of each scan parameter adjustment, including satisfaction ratings and textual evaluations. For example, the MR system uses a reinforcement learning algorithm to adjust the weights of the scan parameter optimization strategy based on user feedback. High-satisfaction parameter tuning strategies receive positive reinforcement, increasing the probability of being selected in similar situations; conversely, low-satisfaction parameter tuning strategies receive negative reinforcement, decreasing the probability of being selected in similar situations.

[0102] Optionally, the agent achieves continuous learning and optimization through a case library accumulation mechanism: successful scan parameter adjustment cases are stored in a long-term memory case library, including initial needs, patient characteristics, adjusted parameters, and final results. When encountering similar situations, the system can retrieve reference cases from the case library, improving parameter adjustment efficiency.

[0103] Optionally, the agent achieves continuous learning and optimization through a periodic knowledge update mechanism: regularly extracting the latest magnetic resonance imaging knowledge and technological advancements from medical literature databases and publicly available technologies from equipment manufacturers to update the knowledge base in long-term memory. The knowledge update process may include two stages: automated literature analysis and manual review, to ensure the accuracy and reliability of the incorporated knowledge.

[0104] Optionally, the agent can continuously learn and optimize through an A / B testing optimization mechanism: for common parameter tuning scenarios, the system periodically performs A / B tests to compare the effects of different parameter tuning strategies. For example, for the requirement of "reducing motion artifacts in FSE scanning", the MR system may test the effects of two strategies to shorten the scan time, such as "increasing the echo train length" and "partial Fourier acquisition", and optimize the default strategy based on the test results.

[0105] Optionally, the intelligent agent can achieve continuous learning and optimization through a multi-center data sharing mechanism: While protecting privacy, protocol agents from different medical institutions can share anonymized parameter tuning experience and performance data to accelerate the learning process. For example, data sharing adopts a federated learning framework, where each institution's raw data remains locally, and only model parameter updates are exchanged.

[0106] Optionally, the intelligent agent can achieve continuous learning and optimization through an expert review mechanism: MRI experts are periodically invited to review the system's parameter tuning strategies and knowledge base, providing professional guidance and corrections. For example, expert feedback can directly update the system's core knowledge base and parameter tuning rules through a privileged interface.

[0107] In this embodiment, through the above continuous learning mechanism, the protocol agent can continuously accumulate experience, optimize parameter tuning strategies, adapt to different clinical needs and equipment environments, and provide increasingly accurate magnetic resonance protocol parameter adjustment services.

[0108] Example 4

[0109] This application discloses a scanning control method for a medical imaging scanning system. Taking an MR system as an example, when a user needs to adjust the current scanning protocol, they can directly interact with the intelligent agent of the medical imaging scanning system using natural language to communicate their requirements for the scanning protocol. The intelligent agent can intelligently plan a parameter tuning strategy based on the user's needs, utilizing relevant MRI knowledge, standards, and tools. After user confirmation, it can automatically apply the new protocol to address the user's need to optimize scanning parameters. Specifically, as shown in Figure 4, this may include:

[0110] Step S401: Receive natural language input from the user. This input represents the user's adjustment requests for the current scanning protocol. These requests may include improving the quality of the scanned images corresponding to the current protocol, increasing scanning speed, meeting clinical requirements, and correcting user errors. In this embodiment, the current scanning protocol is the T2_FSE sequence, the scanned area is the spine, and the parameters are set to {TE, 100ms; TR, 2700ms; relapse balance, 0; bandwidth, 500Hz}. After executing the current scanning protocol, the user inputs the problem "cerebrospinal fluid not lit" to improve the scan results.

[0111] Step S403: Determine the target task based on the user's natural language input. This step corresponds to the agent's needs understanding stage, where the medical imaging scanning system receives the user's input needs through the user interaction interface 204. In this embodiment, based on the aforementioned information, the target task can be determined as "What to do if the cerebrospinal fluid is not visible during a T2_FSE scan of the spine."

[0112] Step S405: The target task is planned into multiple target sub-tasks, and the multiple target sub-tasks are sequentially combined to form a decision tree.

[0113] Before planning the target task, the agent activates relevant tools to collect information about the current environment and the patient. For example, the agent activates a patient cooperation estimation tool to assess the patient's ability to remain still; the agent activates a system query tool to obtain the current status of the MRI equipment and available protocols. The agent activates a protocol preparation tool to select suitable basic protocols from the protocol library. All the aforementioned collected information is stored in the agent's short-term memory as a basis for adjusting scan parameters.

[0114] In this embodiment, the database is searched according to the target task to obtain the following search-related information:

[0115] Information 1: The general reversion balance for FSE sequences is 0.5 or higher. Increasing TR, having more than 1 segment, and using flow compensation are all solutions for cerebrospinal fluid that is too dark.

[0116] Information 2, B1 will consistently highlight the signal if the uniformity correction option is set to advanced;

[0117] Information 3: Enabling retrospective homogenization will affect the contrast of highlighted areas, especially cerebrospinal fluid.

[0118] Based on the above information, it can be determined that adjusting the settings for relapse-return balance, TR, number of segments, flow compensation, and homogenization correction can improve signal recovery. However, retrospective homogenization needs to be disabled. Further, the target sub-tasks extracted from the retrieved information are: increasing "relapse-return balance" to 0.5 or higher to enhance long T1 tissue signals; appropriately extending TR to over 3000ms; increasing the number of segments to 3 or more to improve signal recovery intensity; enabling flow compensation to alleviate cerebrospinal fluid flow voids; checking the "homogenization" correction settings; if it is B1, switching to advanced mode; and disabling retrospective homogenization if it exists.

[0119] Step S407: Based on the decision tree, the target sub-tasks are executed sequentially. If the intermediate results corresponding to one or more target sub-tasks do not meet the set threshold, the decision tree is adjusted or the target task is replanned to generate recommended scanning parameters for the current scanning protocol. In this embodiment, the recommended scanning parameters for the current scanning protocol are set to TR of 3000ms and drive-back balance of 1.

[0120] Step S409: Control the medical imaging scanning system to execute the current scanning protocol according to the recommended scanning parameters to complete the imaging scan of the target object. The intelligent agent can use scan initiation permissions to control the MR system to execute the current scanning protocol according to the recommended scanning parameters. During the scan, motion monitoring tools and physiological signal capture tools continuously monitor the patient's status and provide real-time feedback when necessary.

[0121] Example 5

[0122] This application discloses a scanning control method for a medical imaging scanning system. Taking an MR system as an example, when a user wants to improve image quality, they can directly interact with the intelligent agent of the medical imaging scanning system using natural language to communicate their needs for the scanning protocol. The intelligent agent can intelligently plan parameter tuning strategies based on the user's needs, utilizing relevant MRI knowledge, standards, and tools. After user confirmation, it can automatically apply the new protocol to address the user's need to optimize scanning parameters, such as... Figure 5 Specifically, it may include:

[0123] Step S501: Receive natural language input from the user. This input represents the user's desired image quality adjustment for the current scanning protocol. The user's input is obtained through voice, text, or by scanning the user interaction interface 204. Specifically, the desired image quality adjustment is that the image results do not meet requirements, and the user wants to further improve the image quality but does not know how to adjust the parameters. In this embodiment, the current scanning protocol is the T1_FSE sequence, the scanning site is the head, and the parameters of the current scanning protocol are set to {TE, 10ms; TR, 90ms; back-drive balance, 0; bandwidth, 200Hz}. After executing the current scanning protocol, the user inputs the question "The chemical shift is too large, how to adjust it?" in text, hoping to improve the scanning results.

[0124] Step S503: The user's natural language input is fed into the medical imaging scanning system to generate recommended scanning parameters for the current scanning protocol. Specifically, the medical imaging scanning system includes an agent that can determine the target task based on the user's natural language input; the agent can divide the target task into multiple target sub-tasks and sequentially combine these sub-tasks to form a decision tree; the agent can execute the target sub-tasks sequentially according to the decision tree, and if the intermediate results corresponding to one or more target sub-tasks do not meet a set threshold, the agent can adjust the decision tree or re-plan the target task to generate recommended scanning parameters for the current scanning protocol.

[0125] like Figure 6 The diagram shown is a block diagram of the intelligent agent included in the medical imaging scanning system of this application embodiment. The intelligent agent has multiple functions such as memory, tool or tool recall, action, and planning.

[0126] The agent's memory function can be divided into short-term memory and long-term memory. The agent can store and manage various types of information required for adjusting magnetic resonance imaging (MRI) protocols. Optionally, short-term memory uses an attention-based vector database to store information from the current session. Short-term memory can include user input requests and feedback, such as "wanting to reduce scan time while maintaining image quality" or "the patient cannot remain still for a long time." Short-term memory can also include the current patient's status information, including age, weight, presence of implants, and claustrophobia. Furthermore, short-term memory can contain the current system status of the medical imaging scanning system, including magnet temperature, gradient coil status, and radiofrequency coil selection. Additionally, short-term memory can include the history of scan parameters adjusted in the current session. Optionally, long-term memory uses a combination of a hierarchical knowledge graph and a relational database. Long-term memory can include an MRI basic protocol library containing detailed descriptions of at least several commonly used MRI scanning protocols, each including scan parameters, their default values, value ranges, and physical meanings. Long-term memory can also include a list and description of scan parameters, detailing the physical meaning of each parameter, adjustment methods, and their relationships with other scans. The long-term memory component can also include parameter tuning templates and examples of these templates. Different templates are set for parameters with different clinical needs, such as "T2 sequence adjustment template to reduce motion artifacts" and "diffusion-weighted imaging adjustment template to improve the detection rate of small lesions." The long-term memory component can also contain parameter tuning rules and priorities, such as "prioritizing shortening scan time while ensuring diagnostic quality" and "prioritizing reducing SAR values ​​for pediatric patients." The long-term memory component can also be a magnetic resonance imaging (MRI) physics knowledge base, containing fundamental physical principles of MRI, characteristics of various sequences, image contrast formation mechanisms, and related physical limitations and calculation formulas. Furthermore, the long-term memory component can include equipment parameters and limitations, including hardware parameters, software version limitations, and gradient performance limits for different MRI models. Finally, the long-term memory component can also contain historical user feedback, recording historical parameter tuning results and user feedback for continuous optimization of parameter tuning strategies.

[0127] The agent's planning process involves subtask decomposition, constructing thought chains (decision chains), constructing thought trees (decision trees), evaluating task feasibility, and reflecting on knowledge based on short-term and long-term memory. The agent's planning first decomposes multiple subtasks, including patient assessment, system assessment, and determining the existence of available protocols. Based on these subtasks, thought chains and thought trees are constructed. Task feasibility is then evaluated based on these chains and trees, and reflection is performed based on knowledge from short-term and long-term memory if the task becomes unfeasible. Based on the assessed patient resting ability, suitable basic protocols are selected from the protocol library; it is determined whether the current MR system can execute the basic protocols; parameters are adjusted according to user needs and with reference to knowledge in long-term and short-term memory; and the adjusted parameters are checked against SAR and PNS thresholds. If any of these conditions are not met, the decision tree is replanned.

[0128] In this embodiment, the target task can be determined as "how to adjust the chemical shift when the T1_FSE scan head is too large" according to the aforementioned step S501. A database search is performed based on the target task, yielding the following relevant search information:

[0129] Information 1: During head scanning, set the scene to the chemical displacement direction of the layer where the head can change direction;

[0130] Information 2: Bandwidth, readout fat displacement direction, radio frequency mode, and scene all affect chemical displacement. Increasing bandwidth can reduce displacement in the readout direction, adjusting the radio frequency mode can reduce displacement in the layer direction, and setting the scene is also effective. However, increasing bandwidth will reduce the signal-to-noise ratio, and switching the radio frequency mode will increase the SAR value.

[0131] Information 3 suggests that increasing bandwidth, adjusting the radio frequency mode, changing the direction of fat readout displacement, and setting the scene to the head will reduce chemical artifacts.

[0132] Based on the above information, we can determine the bandwidth, reduce chemical shift in the readout direction; adjust the RF mode to fast to reduce layer-direction shift; check if the scene is the head and optimize layer-direction shift; adjust the readout fat shift direction to change the offset direction. Further, several target sub-tasks extracted from the retrieved information are: increase the bandwidth; adjust the RF mode to "fast"; confirm the scene is the head; adjust the "readout fat shift direction" parameter. The recommended scanning parameter settings for the current scanning protocol are: bandwidth, 250Hz (per pixel); RF mode, fast; scene, head.

[0133] Step S505: Display the recommended scanning parameters of the current scanning protocol in a visual manner. The recommended scanning parameters of the current scanning protocol can be displayed through the user interaction interface 204. The visualization of the recommended scanning parameters can be a comparison table before and after the scanning parameters are adjusted, the expected changes in image quality, the expected changes in scanning time, the safety parameter evaluation results, etc.

[0134] Optionally, after reviewing the recommended scanning parameters of the current scanning protocol, the user can suggest modifications or confirm directly. If the user suggests modifications, such as "wanting to further shorten the scanning time," the system will return to the parameter planning stage and generate new recommended scanning parameters.

[0135] Step S507: In response to the user's confirmation of recommended scanning parameters for the current scanning protocol via the user interaction interface, the medical imaging scanning system is controlled to execute the current scanning protocol. In this embodiment, the agent is equipped with the system permissions required to perform parameter tuning. After user confirmation, the agent initiates its action function, automatically applying the parameter tuning results, allowing the user to start executing the adjusted current scanning protocol with a single click.

[0136] In this embodiment, the medical imaging scanning system uses a large language model as its understanding and planning engine. It possesses the ability to understand user needs, communicate with users, and perform logical reasoning for protocol tuning. Based on relevant knowledge and tools, it optimizes protocol parameters and can intelligently learn and optimize tuning strategies based on user feedback. The intelligent agent has short-term and long-term memory (knowledge base) required for protocol tuning, including general descriptions of all basic MRI protocols, parameter lists and descriptions, tuning templates and examples, tuning rules and priorities, basic MRI physics knowledge (physical limitations and related formulas), equipment parameters and hardware / software limitations, user input and feedback, patient status, system status, and system information. The medical imaging scanning system can automatically apply the tuning results, allowing users to start scanning with the adjusted protocol with a single click. When users need to adjust the protocol, they can directly interact with the protocol intelligent agent through natural language to communicate their needs for the scanning protocol. The system intelligently plans tuning strategies and automatically applies the new protocol after user confirmation, addressing the user's need to optimize scanning parameters.

[0137] Example 6

[0138] This application discloses a schematic diagram of the display interface of a user interface 204 for a medical imaging scanning system. The display interface shows the recommended scanning parameters for the current scanning protocol, along with three sets of recommended parameters. Clicking on the simulation performance of each set of parameters allows for a comparison between the recommended parameter values ​​and the original scanning protocol parameter values. The scanning protocol is t2_flair_fse_tra, and the scanning site is the head. The intelligent agent unit 104 generates the scanning time, resolution, and motion resistance corresponding to each set of recommended parameters. The first set of recommended parameters corresponds to a scanning time closest to the original scanning protocol (1 minute 30 seconds) and has the best overall performance. The second set of recommended parameters has the highest resolution and the longest scanning time (2 minutes 30 seconds). The third set of recommended parameters has the best motion resistance and a relatively long scanning time (2 minutes 30 seconds). The first set of recommended parameters, which is the best recommendation displayed on the display interface, clearly shows the pixel resolution and signal-to-noise ratio, facilitating user understanding of the details of the best recommendation.

[0139] The medical imaging scanning system of this application embodiment supports one-click scanning by the user. When the user needs to adjust the protocol, they can directly interact with the protocol agent in the form of natural language to communicate their needs for the scanning protocol and intelligently plan parameter adjustment strategies. The recommended scanning parameters of the current scanning protocol are clearly displayed on the display interface; the recommended scanning parameters include multiple sets, providing the user with a variety of choices.

[0140] It will be understood by those skilled in the art that various aspects of this specification can be described and illustrated in several patentable ways, including any new and useful combination of processes, machines, products, or substances, or any new and useful improvements thereof. Therefore, various aspects of this specification can be implemented entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of software and hardware, which are generally referred to herein as “modules,” “units,” “components,” “devices,” or “systems.” Furthermore, various aspects of this specification can take the form of a computer program product embodied in one or more computer-readable media having computer-readable program code.

[0141] Furthermore, unless expressly stated in the claims, the order of processing elements and sequences, the use of numbers and letters, or other names described in this specification are not intended to limit the order of the processes and methods described herein. Although various examples have been discussed in the foregoing disclosure of some embodiments of the invention that are currently considered useful, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments; rather, the claims are intended to cover all modifications and equivalent combinations that conform to the spirit and scope of the embodiments described herein. For example, while the implementation of the various components described above can be embodied in a hardware device, it can also be implemented as a purely software solution, such as an installation on an existing server or mobile device.

[0142] Similarly, it should be noted that, in order to simplify the description disclosed herein and thus aid in the understanding of one or more embodiments of the invention, multiple features may sometimes be grouped into a single embodiment, drawing, or description thereof in the foregoing description of the embodiments. However, the method described herein should not be construed as reflecting an intention that the claimed object to be scanned requires more features than expressly recited in each claim. In fact, the embodiments contain fewer features than all the features of the individual embodiments disclosed above.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A server suitable for medical imaging scanning systems, characterized in that, The server includes: An interaction unit is used to receive natural language input from a user, the natural language input representing the user's adjustment needs for the current scanning protocol; An intelligent agent unit is configured to determine a target task based on the user's natural language input, and to plan the target task into multiple target sub-tasks, wherein the multiple target sub-tasks are sequentially combined to form a decision tree; according to the decision tree, the target sub-tasks are executed sequentially, and in response to the intermediate result corresponding to at least one target sub-task not meeting a set threshold, the decision tree is adjusted or the target task is re-planned to generate recommended scanning parameters for the current scanning protocol.

2. The server according to claim 1, characterized in that, The intelligent agent includes a long-term memory knowledge base and a short-term memory knowledge base. The long-term memory knowledge base is used to store at least one of the following: prior scanning protocol, parameter tuning template, parameter tuning rules, and parameter tuning priority. The short-term memory knowledge base is used to store at least one of the following: the user's real-time feedback, the real-time status of the detected object, and the real-time status information of the medical imaging scanning system.

3. The server according to claim 2, characterized in that, The server also includes: An update unit is used to update the long-term memory knowledge base based on at least one of the recommended scanning parameters of the current scanning protocol, the decision tree, and the user's real-time feedback.

4. The server according to claim 2, characterized in that, The server also includes a tool unit capable of performing at least one of artifact detection, motion monitoring, protocol preparation, SAR estimation, PNS estimation, patient compliance estimation, and physiological signal capture. The intelligent agent unit can call the tool unit to complete the target sub-task.

5. The server according to claim 1, characterized in that, The recommended scanning parameters of the current scanning protocol include a first set of recommended parameters and a second set of recommended parameters. The intelligent agent unit is also used to generate a first image quality estimate corresponding to the first set of recommended parameters and a second image quality estimate corresponding to the second set of recommended parameters. The image quality indicators corresponding to the first image quality estimate and the second image quality estimate are different.

6. The server according to claim 1, characterized in that, The interactive unit simultaneously displays the initial scanning parameters and the recommended scanning parameters of the current scanning protocol.

7. A medical imaging scanning system, characterized in that, include: The scanner is used by the user to scan the object being inspected. The user interaction interface is configured to receive natural language input from the user, which represents the user's adjustment requirements for the current scanning protocol; processor; A non-transient computer-readable storage medium operatively communicatively with the processor, wherein the non-transient computer-readable storage medium contains one or more programming instructions that, when executed, cause the processor to: The target task is determined based on the user's natural language input, and the target task is planned into multiple target sub-tasks. The sequential combination of the multiple target sub-tasks forms a decision tree. According to the decision tree, the target sub-tasks are executed sequentially. In response to the intermediate result corresponding to at least one target sub-task not meeting a set threshold, the decision tree is adjusted or the target task is re-planned to generate recommended scanning parameters for the current scanning protocol. The user interaction interface is controlled to display the recommended scanning parameters for the current scanning protocol, and in response to the user's confirmation operation on the user interaction interface, the scanner is controlled to execute the current scanning protocol according to the recommended scanning parameters.

8. The medical imaging scanning system according to claim 7, characterized in that, The processor includes an intelligent agent, which includes a long-term memory knowledge base and a short-term memory knowledge base. The long-term memory knowledge base is used to store at least one of the following: prior scanning protocol, parameter tuning template, parameter tuning rules, and parameter tuning priority. The short-term memory knowledge base is used to store at least one of the following: real-time feedback from the user, real-time status of the detected object, and real-time status information of the medical imaging scanning system.

9. A scanning control method for a medical imaging scanning system, characterized in that, The method includes: Receive natural language input from the user, the natural language input representing the user's adjustment needs for the current scanning protocol; The target task is determined based on the user's natural language input; The target task is divided into multiple target sub-tasks, and the multiple target sub-tasks are sequentially combined to form a decision tree; According to the decision tree, the target sub-tasks are executed sequentially, and in response to the intermediate results corresponding to at least one target sub-task not meeting the set threshold, the decision tree is adjusted or the target task is re-planned to generate recommended scanning parameters for the current scanning protocol. The medical imaging scanning system is controlled to execute the current scanning protocol according to the recommended scanning parameters in order to complete the imaging scan of the object to be detected.

10. The scanning control method for the medical imaging scanning system according to claim 9, characterized in that, The adjustment requirements for the current scanning protocol include: Improve the quality of the scanned images corresponding to the current scanning protocol, increase the scanning speed, meet clinical requirements, or correct the user operation error, at least one of the following:

Images