Investigation system with one agent

Abstract

Investigation system (10), showing: - a medical imaging system (1) with a control unit (2), - an agent (3) trained in software and / or hardware-based operation to control the control device (2) and to communicate with the control device (2), - an interface (5) based on an open standard between the control device (2) and the agent (3).

Description

[0001] The invention relates to an examination system.

[0002] When controlling medical imaging systems for patient examinations, so-called software agents, also known as agents or softbots, can be used to support the operating personnel. Such a software agent comprises a computer program capable of a certain well-defined, independent, and self-dynamic autonomous behavior. This means that, depending on various states, a specific processing operation takes place without requiring an external start signal or any external control intervention during the process. In addition to software agents, there are also so-called technical agents. A technical agent is a definable hardware and / or software unit with defined objectives. A technical agent strives to achieve these objectives through autonomous behavior and interacts with its environment and other agents in the process.

[0003] Furthermore, in the context of artificial intelligence applications, there are software-based agents that possess the following properties: The agent operates autonomously, i.e., independently of user intervention. The agent exhibits cognitive abilities, i.e., it is capable of learning and learns based on previous decisions and observations. The agent is communicative and communicates its state as an effect on its environment. The agent is modally adaptive, i.e., it changes its own settings, particularly specific parameters and / or structures, based on its own state and the state of its environment. The agent is active and performs actions on its own initiative. Furthermore, the agent is reactive, i.e., it reacts to changes in the environment. The agent is robust, meaning it compensates for external and internal disturbances. The agent is "social," i.e., it communicates with other agents.However, simultaneous external interventions that may restrict autonomy or change decision criteria cannot be ruled out.

[0004] Artificial intelligence-based agents, so-called AI agents, facilitate the operation and control of external systems, particularly medical imaging systems. These AI agents are implemented as artificial neural networks and are trained with training data to exhibit behavior consistent with that data.

[0005] One specific type of such AI agent is a Large Language Model (LLM), which is characterized by its ability to generate text. It is a computational linguistic probability model that has learned statistical word and sentence sequence relationships from a large number of text documents through a computationally intensive training process. This process primarily utilizes so-called transformers, which form the basic architecture of pre-trained machine learning models, such as generative pre-trained transformers (GPTs).

[0006] Therefore, when we speak of an "agent", we mean a technical system that interacts with a medical technology system and other technical resources, as well as with a user of the medical technology system.

[0007] In order to allow the described agents to interact with a medical imaging system, an interface is needed through which the agent connects to the medical imaging system.

[0008] Today's medical imaging systems, especially CT systems (CT stands for "computed tomography"), are heavily reliant on the proprietary user interface, particularly the graphical user interface (GUI) and controls such as buttons and switches, of the respective manufacturer. This individualized design limits the possibility of using alternative or intelligent control methods. In particular, the integration of modern AI agents or large language models for the direct control of a medical imaging system, especially a CT system, is not currently envisioned. As a result, the potential for flexible, language-based, and context-adaptive interaction with the CT system or other medical imaging systems remains untapped.

[0009] It would therefore be desirable to control a medical imaging system, particularly a CT system, with any agent, especially an AI agent, and preferably an AI agent with LLM, regardless of a manufacturer-specific interface. Such a broad, universally applicable interface would also greatly simplify and facilitate automation, intuitive operation, and the intelligent application of assistance functions in everyday clinical practice.

[0010] Therefore, the task is to simplify and broaden the interaction of different software- and / or hardware-based agents with a medical imaging system, in particular a CT system.

[0011] This problem is solved by an investigation system according to claim 1.

[0012] The examination system according to the invention comprises a medical imaging system with a control unit. The examination system also includes a software-based and / or hardware-based agent for controlling and communicating with the control unit, and an open-standard-based interface between the control unit and the agent. It should be expressly noted that the examination system according to the invention may comprise multiple agents, but includes at least one.

[0013] The control unit enables the control of an imaging functionality within the medical imaging system.

[0014] As explained in the introduction, the agent preferably has the property of operating autonomously, i.e., independently of user intervention. As already explained, the agent preferably possesses cognitive abilities. That is, it is preferably capable of learning and learns based on previous decisions and observations. The agent is preferably communicative and communicates its states to the environment as effects on its surroundings. The agent is preferably modally adaptive, i.e., it changes its own settings, especially specific parameters and / or structures, based on its own states and the states of the environment. The agent is preferably active and performs actions on its own initiative. Furthermore, the agent is preferably reactive, i.e., it reacts to changes in the environment. The agent is preferably robust, i.e., it compensates for external and internal disturbances. In a technical sense, the agent is also preferably "social," i.e.,He communicates with other agents. However, simultaneous external interventions that might restrict autonomy or change decision criteria cannot be ruled out.

[0015] An open standard between the control unit and the agent enables communication between the control unit and one or more agents of any type or manufacturer. The diagnostic system thus includes an open, standardized communication layer that allows the use of any agent capable of communicating via the generally known, standardized protocol. Advantageously, the diagnostic system can utilize not only manufacturer-specific agents but also any agent capable of communicating via the standardized protocol. Advantageously, a user can select the most suitable agents from a pool that extends far beyond the potentially limited selection provided by the manufacturer of the medical imaging system. Advantageously, the user can also utilize the most innovative operating solutions and assistants, regardless of their manufacturer.Furthermore, due to standardized, open communication, the integration and coordination effort for integrating a new agent is reduced compared to conventional examination systems, and the interoperability between the medical imaging system and the agent is significantly increased compared to the conventional approach.

[0016] The dependent claims and the following description each contain particularly advantageous embodiments and further developments of the invention. Furthermore, within the scope of the invention, the various features of different embodiments and claims can also be combined to form new embodiments.

[0017] In one variant of the investigation system according to the invention, the interface is based on a manufacturer-independent MCP protocol.

[0018] MCP (an acronym for "Model Context Protocol") is an open standard introduced by the company Anthropic to standardize the integration and data exchange between artificial intelligence, particularly large language models, and external tools, systems, and data sources. MCP was released as an open-source project. MCP enables language models to communicate with the outside world and thus directly interact with software.

[0019] Before the introduction of MCP, developers had to create individual interfaces, called APls, for each combination of AI model or agent and external data source, resulting in a complex M*N integration problem (N and M are natural numbers, and they can be different). This meant generating M*N different individual interfaces for M agents and N external data sources, which was very costly. MCP reduces this effort to a single, open, standardized interface for all M agents and N data sources.

[0020] Preferably, the interface includes a server, ideally an MCP server, which communicates with the control unit via a proprietary, non-public interface. Thus, while the server communicates externally with the agent(s) via an open standard, it interacts with the control unit via its familiar proprietary, non-public interface. In this way, the safety-relevant and functional boundaries of the medical imaging system remain protected by the control of the medical imaging system's manufacturer. The overall system therefore preferably features an open, universal, or...a universal interface between the agent and the control unit and a proprietary interface between the control unit and one or more servers, preferably of the medical imaging system, with the above-mentioned advantages for the security and flexibility of the operation of the medical imaging system.

[0021] The interface is particularly well-suited, due to its open standard, for communicating with agents of any make or type. As mentioned previously, this greatly expands the selection of agents that can be used to operate a medical technology system.

[0022] The agent is particularly well-suited for facilitating communication between a user and the control unit. Advantageously, the agent can simplify the use of the medical imaging system by actively guiding the user through the examination process and ensuring the provision of the necessary support programs and resources.

[0023] The investigation system according to the invention is also preferably designed such that the interaction between the agent and the control unit is implemented by a client-server system with a client and the aforementioned server. The agent is preferably implemented as a host, and particularly preferably as an MCP host. In this variant, the host comprises or controls one or more clients. In the simplest variant, the agent comprises only a single client or is implemented as a client. Typically, the agent is configured as a host when multiple clients are integrated into the client-server system and are to be controlled and managed.

[0024] Preferably, the interface between the agent and the control system comprises one or more clients and one or more servers, with the client(s) configured to access the server(s). Advantageously, the client(s) access the server(s) via an interface based on an open standard. Because of this open standard, which is valid and available to all clients, compatibility problems between different clients and servers are avoided, making the overall system more flexible and easily scalable compared to conventional approaches.

[0025] The aforementioned client-server system, preferably designed as an MCP architecture, comprises an agent, which includes an application, preferably a K-application, that preferably acts as the MCP host, coordinating and managing one or more components, preferably MCP clients. The agent of the client-server system maintains a connection, optionally with the aid of components that are preferably MCP clients, to one or more servers in order to obtain context from the server or servers that the agent can use.

[0026] A server comprises a program that provides a client or agent with context, particularly functionalities. Context generally encompasses the communication between the client and the server, and specifically the services, particularly functionalities, that the server can provide to the client. The server is therefore primarily configured to provide functionalities, especially tools, resources, and prompts. Advantageously, the agent can directly access server resources during its interaction, thus effectively supporting the planning and execution of a patient examination.

[0027] In a preferred embodiment of the investigation system according to the invention, the server is therefore configured to provide at least one of the following functionalities: - a tool, - Resources, - a prompt.

[0028] The server therefore provides not only direct active functionalities, such as tools or prompts, but also passive functionalities, i.e., resources that primarily include data.

[0029] A tool comprises a helper program or utility. In particular, the tool preferably includes functions that the agent, preferably an LLM, can actively call, whereby the agent can decide, preferably based on user requests, when the functions are used. The protocol of the interface, based on an open standard, includes specific protocol operations for querying the available tools and calling a specific tool.

[0030] Preferably, the tool has a function of one of the following types: - writing data to a database, - calling an external application programming interface, - modifying files, - triggering a change of state of the control device and / or the medical imaging system, - triggering a function of the control device and / or the medical imaging system.

[0031] The tool used is therefore preferably designed to write to databases and / or call external APIs (short for "Application Programming Interfaces") and / or modify files.

[0032] Preferably, the tool used is configured to trigger a change of state in the control device and / or the medical imaging system. A state of the control device and / or the medical imaging system comprises an operating state of the control device and / or the medical imaging system, which preferably influences the imaging process of the medical imaging system. A change of state is defined as a change in the described state.

[0033] The change of state preferably comprises a change of state of one of the following types: - Change from an energy-saving state / standby state to an operational readiness state, - Change in the position of the patient bed (mechanical condition), - Change in the tilt angle of the gantry (mechanical condition), - Change in image capture state.

[0034] An image acquisition state includes the state of an ongoing image acquisition of the medical technology system and the state of an image acquisition of the medical technology system that has not yet started or been completed.

[0035] Furthermore, the tool used is preferably configured to trigger a function of the control unit and / or the medical imaging system.

[0036] The function of the control device and / or the medical imaging system preferably includes a function of one of the following types: - Move the patient (on the couch) to the starting position for image acquisition, - Start image capture, - Move the patient bed to a position from which the patient can comfortably get down, - Give the patient a verbal instruction (in particular: inhale, hold your breath, exhale).

[0037] The tool used can be configured specifically for registering patients, changing scan parameters, performing scans, and executing system functions. The agent controls the tool using AI models. Specifically, the tool enables the execution of actions. Each tool preferably defines a specific operation with typed inputs and outputs. The agent requests the execution of a tool based on the context. Tools are interface-defined functions that can be called, in particular, by LLMs (Learning Management Models). Tools may require user consent before execution to ensure that users retain control over the actions performed by the model.

[0038] Resources, particularly data resources, primarily consist of passive data sources that provide read-only access to contextual information relevant to a specific use case of the investigation system. Resources may contain information regarding a patient waiting list or system properties. Resources provide structured access to information that an agent, particularly an AI agent, can retrieve and provide as context to the model used by the agent, preferably an AI model. Resources provide data from files, APIs, databases, or other sources that an agent, particularly an AI agent, needs to understand the context. Applications or agents can directly retrieve this information and decide how to use it.The protocol of the interface based on an open standard preferably includes suitable protocol operations for listing available direct resources, querying so-called resource templates, retrieving resource contents, and monitoring resource changes.

[0039] A prompt comprises pre-built instruction templates that tell a model used by the agent how to interact with specific tools and resources. Prompts can be used to reschedule a patient, summarize meetings, or draft an email. Prompts offer reusable templates. They allow an MCP server author—the developer or vendor of the MCP server—to provide parameterized prompts for a specific topic or to demonstrate how to best utilize the MCP server.

[0040] Prompts comprise structured templates that define expected inputs and interaction patterns. They are user-driven and require explicit execution. Prompts can be context-sensitive and reference available resources and tools to create comprehensive workflows. The protocol of the open-standard-based interface includes suitable protocol operations for querying available prompts and retrieving prompt details. Prompts can also be advantageously implemented for combined speech and text input. In this scenario, an agent prompts the user to enter data via voice input, while simultaneously allowing the user to read the prompt and enter data via text input through a text prompt that specifically requests data entry.The advantage is that communication takes place via two different communication channels, which utilize different types of sensory perception, making the communication more reliable and robust compared to communication via only one communication channel.

[0041] In particular, the implementation of the client-server system enables the use of an AI agent or AI assistant that responds to textual or voice input and can assist in registering a patient for an examination. In one embodiment of the examination system according to the invention, the agent thus comprises an AI application, i.e., an application generated using artificial intelligence. An agent that uses such an AI application is also referred to simply as an "AI agent."

[0042] Artificial intelligence-based agents, so-called AI agents, facilitate the operation and control of external systems, particularly medical imaging systems. A specific and particularly preferred type of such an AI agent used according to the invention comprises a Large Language Model (abbreviated "LLM"), which is characterized by its text generation capability. It is a computational linguistic probability model that has learned statistical word and sentence sequence relationships from a large number of text documents through a computationally intensive training process. For this purpose, so-called transformers are used, which form the basic architecture of so-called pre-trained machine learning models, such as generative pre-trained transformers (abbreviated "GPT").

[0043] In one embodiment of the examination system according to the invention, the medical imaging system comprises a control unit of one of the following types: - a computed tomography system, - a magnetic resonance imaging system, - an X-ray imaging system, - an ultrasound system.

[0044] The flexible use of agents, especially AI agents, can make the operation of very different types of medical imaging systems and the preparation for examinations with these systems easier, more user-friendly, and more flexible. In particular, the integration of agents into the operation of these systems is simplified.

[0045] In a preferred embodiment of the investigation system according to the invention, the investigation system comprises a plurality of different agents and servers. The different agents access the different servers via a common and universal interface and can access the different tools, resources, or prompts of the different servers using only a single protocol. In this way, any agent can be integrated into the investigation system without having to generate separate software for bidirectional communication between the respective agent and the respective server, so that all components communicate using the same standard or protocol. Thus, when using N different agents and M different servers, N * M communication standards are no longer required.Protocols need to be created; instead, a single software interface is sufficient for all N * M possible uses.

[0046] The invention is explained in more detail below with reference to the accompanying figures and exemplary embodiments. The figures show: Fig. 1 a schematic representation of an investigation system according to an embodiment of the invention, Fig. 2 a schematic representation of an investigation system according to an alternative embodiment of the invention with an agent, wherein the agent uses a plurality of MCP clients as an MCP host and the investigation system provides a plurality of MCP servers, Fig. 3. A schematic representation of a conventional system with multiple agents using different communication standards and multiple servers. Fig. 4 a schematic representation of the principle of an investigation system according to an embodiment of the invention with a plurality of agents with a uniform communication standard and a plurality of servers, Fig. 5 a flowchart illustrating the functioning of an examination system according to an embodiment of the invention by reproducing a dialogue between an AI agent of the examination system and an MTRA, i.e. a medical-technical radiology assistant.

[0047] In Fig. Figure 1 is a schematic representation of an investigation system 10 according to an embodiment of the invention.

[0048] The in Fig. The specific examination system 10 shown comprises a CT system 1 with a control unit 2. The control unit 2 has in the Fig. In the embodiment shown in Figure 1, a software-based agent 3 is used to control the control device 2 and to communicate with the control device 2.

[0049] The software-based Agent 3, for example, is stored as an AI system on a separate computer in the hospital, which acts as the host.

[0050] Part of the investigation system 10 is also an interface 5 based on an open standard, i.e., a universal interface, in this embodiment in the form of an MCP interface, between the control unit 2 and the agent 3. The MCP interface comprises an MCP protocol, on the basis of which communication takes place between the control unit 2 and the software-based agent 3. Furthermore, the MCP interface includes an MCP server 2a (left in the image), which provides resources, tools, and prompts, and an MCP client 4 (right in the image) as a component of the agent 3 (which in this configuration is also referred to as the MCP host), which maintains a communication connection via the interface 5 with the MCP server 2a of the control unit 2 and receives context or data from the MCP server 2a, which the agent 3 can use to execute functions.

[0051] In Fig. Figure 2 shows a schematic representation of a medical examination system 10 according to an alternative embodiment of the invention. In the Fig. In the medical technology examination system 10 shown in Figure 2, the agent 3, i.e., a computer application, has a plurality of different MCP clients 4. The agent 3 is configured as an MCP host. The MCP clients 4 each communicate with their assigned MCP servers 2a, which are each part of the control unit of a CT system 1. The same universal interface 5 is used for the different MCP clients 4 and MCP servers 2a. However, it should be noted that communication within the control unit of the CT system takes place via a proprietary interface. That is, communication between the MCP clients 4 and the control unit takes place via a universal interface, and communication between the control unit and the servers 2a it comprises takes place via a proprietary interface.Therefore, it is not necessary to use different interfaces for different MCP clients 4 and different MCP servers 2a. This is possible because all MCP clients 4 and MCP servers 2a use one and the same MCP protocol, and thus no specific interfaces are required for different MCP clients 4 or MCP servers 2a. For example, a first MCP server 2a (far left in . Fig. 2) a first patient database as a resource, to which a first MCP client 4 (far left in Fig. 2) can access.

[0052] The first MCP client 4, for example, has the function of querying a patient waiting list, controlled by agent 3, and accesses the first patient database to perform this function. This database provides the first MCP client 4 with read-only access to information for this context or specific task. A second MCP client 4 (in Fig. 2 (in the second position from the left) has the function, controlled by agent 3, to query so-called log files as resources, and accesses a second database containing these log files for this purpose (see in Fig. 2 in the second position from the left).

[0053] A third MCP client 4 (in Fig. 2 (in the third position from the left) has the function, controlled by agent 3, of maintaining a connection with a tool that is configured to plan an examination in conjunction with agent 3, register patients for this purpose, and adapt scan parameters to an individual patient and their patient data. For this purpose, the third MCP client 4 communicates with a tool for planning an examination, which is located in an MCP server 2a (in) responsible for this task. Fig. 2 in the third position from the left) is stored and communicates with the third MCP client 4 via the universal interface 5.

[0054] A fourth MCP client 4 (in Fig. 2 (shown on the far right) has a function with a tool for controlling a recording and executing system functions. For this purpose, the fourth MCP client 4 accesses an MCP server 2a, which has such a tool (in Fig. 2 (shown on the far right) for the execution of active functions of the CT system. The Kl-Agent 3, as the MCP host, coordinates and manages the various MCP clients 4, which maintain a connection to the respective MCP server 2a and receive context from the respective MCP server 2a, which the agent 3 can use for the respective task.

[0055] Agent 3 requests the execution of a specific tool – mediated via the appropriate MCP client 4, based on the respective context. This is for communication within the context of Fig. The MCP protocol used in the embodiment shown in Figure 2 now has different methods for using the described tools. For example, there is a method to query the available tools or to execute a specific tool. The tools enable the agents 3, which are located in the Fig. The embodiment shown in 2 includes AI applications that perform actions on behalf of a user.

[0056] In Fig. Figure 3 shows a schematic representation of a conventional system 30 with multiple agents 3 and servers 2a. In this conventional system 30, the agents 3 access a server 2a via an individual protocol. If an agent 3 wants to access a different server 2a, it must use the individual protocol of that server 2a. Conversely, if each server 2a wants to communicate with a specific agent 3 of a specific make, it must use a protocol specific to that agent, resulting in a total of 3 * 4, or 12, different protocol combinations.

[0057] In Fig. Figure 4 shows a schematic representation of the principle of an investigation system 10 according to the invention with a plurality of agents 3 and servers 2a. The different agents 3 access the different servers 2a via a universal interface 5 and can access the different tools, resources, or prompts of the different servers 2a using only a single protocol. In this way, any agents 3 can be integrated into the system without having to generate separate software for bidirectional communication between agent 3 and server 2a, so that all components communicate using the same standard or protocol. Thus, when using N different agents 3 and M (N and M are natural numbers) different servers 2a, N * M communication standards are no longer required.Protocols need to be created; instead, a single software interface 5 is sufficient for all N * M uses.

[0058] In Fig. Figure 5 shows a flowchart 500 which illustrates the functioning of an examination system according to an embodiment of the invention by reproducing a dialogue between an AI agent 3 of the examination system and an MTRA, i.e. a medical-technical radiology assistant.

[0059] In the first dialogue step 5.l, the MTRA initially informs the AI ​​agent 3 via voice input of the following: "Schedule an examination for Peter Meier, born on April 26, 1999, for next Tuesday. He is complaining of stomach pains."

[0060] AI Agent 3 now analyzes the words spoken by the medical technical radiology assistant (MTRA) and accesses different servers 2a with different databases for this purpose. For example, the first server 2a contains a database of people's names and their genders. AI Agent 3 then queries the database using client 4 to determine the likely gender of a person named "Peter Meier." The first database, or server 2a, then informs AI Agent 3, via client 4, that "Peter Meier" is a man. Therefore, AI Agent 3 assumes that "Peter Meier" is likely a male person. Consequently, in the second dialogue step 5.II, AI Agent 3 asks the MTRA: "Peter Meier sounds like a man's name to me. Would you please confirm this?"

[0061] Furthermore, the AI ​​agent 3, using a client 4, queries a second database to determine what type of examination might be appropriate for the symptom "stomach pain". The second database, or rather a server 2a containing the second database, informs the AI ​​agent 3, via client 4, that an abdominal examination would be appropriate in connection with the aforementioned symptom "stomach pain".

[0062] In the third dialogue step 5.III, the MTRA responds to the question from AI agent 3 regarding confirmation of the patient's presumed gender: "That's correct".

[0063] Now, AI agent 3, using client 4, queries server 2a, which contains a third database with a schedule for an abdominal CT scan. Server 2a, using client 4, informs AI agent 3 of an available appointment for "next Tuesday at 9:00 AM".

[0064] In the fourth dialogue step 5.IV, the AI ​​agent 3 therefore informs the MTRA: "There is an appointment available next Tuesday at 9:00 am".

[0065] The radiographer now asks the patient, Peter Meier, if he can come for an examination next Tuesday at 9:00 a.m. Mr. Meier replies to the radiographer that he unfortunately only has time next Tuesday afternoon. Therefore, the radiographer answers AI agent 3 in the fifth dialogue step (5.V): "Mr. Meier can only come in the afternoon."

[0066] As part of the sixth dialogue step 5.VI, AI Agent 3, using its client 4, queries server 2a, which contains the third database for scheduling an abdominal CT scan, to inquire about an available appointment next Tuesday afternoon. Server 2a, via client 4, informs AI Agent 3 that an appointment is available at 2 PM. Therefore, in the sixth dialogue step 5.VI, AI Agent 3 asks the radiographer: "2 PM?"

[0067] The MTRA responds to this in the seventh dialogue step 5.VII: “Yes”.

[0068] Finally, in the eighth dialogue step 5.VIII, the AI ​​agent 3 confirms the scheduling of the appointment for the abdominal examination of the patient, Mr. Peter Meier, for the agreed date: "I have registered Mr. Meier as a patient and scheduled an examination of the abdominal cavity for Tuesday at 2 p.m."

[0069] Finally, it should be noted once again that the methods and devices described above are merely preferred embodiments of the invention and that the invention can be varied by a person skilled in the art without departing from the scope of the invention, insofar as it is defined by the claims. For the sake of completeness, it should also be noted that the use of the indefinite articles "a" or "an" does not preclude the possibility that the features in question may be present multiple times. Likewise, the term "unit" does not preclude the possibility that it consists of several components, which may also be spatially distributed. Regardless of the grammatical gender of a particular term, persons of male, female, or other gender identities are included. QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited non-patent literature

[0000] Plan an investigation for Peter Meier, born on 26.4.1999

[0059] Mr. Peter Meier, for the agreed appointment: “I have registered Mr. Meier as a patient and scheduled an examination of the abdomen for Tuesday at 2 p.m.

[0068]

Claims

[1] Investigation system (10), comprising: - a medical imaging system (1) with a control unit (2), - an agent (3) trained in software and / or hardware-based operation to control the control device (2) and to communicate with the control device (2), - an interface (5) based on an open standard between the control device (2) and the agent (3). [2] Investigation system according to claim 1, wherein the interface (5) is based on a manufacturer-independent model context protocol. [3] Investigation system according to claim 1 or 2, wherein the interface (5) comprises a server (2a), preferably a Model Context Protocol server, which communicates with the control device (2) via a proprietary, non-public interface. [4] Investigation system according to one of the preceding claims, wherein the interface (5) is set up to communicate with an agent (3) of any make or type due to the open standard. [5] Investigation system according to one of the preceding claims, wherein the agent (3) is configured to communicate between a user and the control device (2). [6] Investigation system according to one of the preceding claims, wherein the interaction between the agent (3) and the control device (2) is implemented by a client-server system comprising a client component (4) and the server (2a). [7] Investigation system according to claim 6, wherein the agent (3) is implemented as an MCP host and / or comprises a client. [8] Investigation system according to claim 6 or 7, wherein the interface (5) between the control device (2) and the agent (3) comprises the client component (4) and the server (2a), wherein the client component (4) is configured to be controlled by the agent (3), preferably the MCP host, and to access the server (2a). [9] Investigation system according to any one of claims 6 to 8, wherein the server (2a) is configured to provide at least one of the following functions: - a tool, encompassing a helper program or utility, - a resource, especially a data resource, - a prompt. [10] Investigation system according to claim 9, wherein the server (2a) is configured to provide a tool, and the tool has a function of one of the following types of functions: - writing data to a database, - calling an external application programming interface, - modifying files, - triggering a change of state of the control device (2) and / or the medical imaging system (1), - triggering a function of the control device (2) and / or the medical imaging system (1). [11] Investigation system according to claim 9 or 10, wherein the server (2a) is configured to provide a resource as a function, and the resource comprises a passive data source that provides read-only access to information relevant to a specific use case of the investigation system (10). [12] Investigation system according to one of claims 9 to 11, wherein the server (2a) is configured to provide a prompt as a function, and the prompt has a function comprising a pre-made instruction template that tells the agent (3) how to work with a specific tool and / or a specific resource. [13] Investigation system according to any one of claims 6 to 12, wherein the agent (3) comprises one of the following functions: - receiving a user request, - querying information from the server (2a) via available tools or resources, - the active invocation of a tool or resource based on a user request and based on information from the server (2a) about available tools or resources. [14] Investigation system according to one of the preceding claims, wherein the agent (3) comprises an application operating on the basis of artificial intelligence. [15] Examination system according to one of the preceding claims, wherein the medical imaging system (1) comprises a control device (2) of one of the following types: - a computed tomography system, - a magnetic resonance imaging system, - an X-ray imaging system, - an ultrasound system.

Images