Techniques for providing control signals relating to a workflow for radiological devices, such as medical scanners

Abstract

The present invention relates to techniques for providing control signals regarding workflows for radiology devices, such as medical scanners, within a fleet of radiology devices. Technologies include a method having the steps of receiving (S104-A) first input data relating to a request for medical imaging of a patient, and receiving (S104-B) second input data indicative of availability of at least one medical scanner in a fleet of medical scanners. The received (S104-A) first input data and the received (S104-B) second input data are processed (S108). Processing (S108) includes selecting available medical scanners from the fleet. The processing is performed by means of generative artificial intelligence for creating a workflow of medical imaging on request (S108). A control signal indicative of a workflow for performing medical imaging by means of the selected medical scanner is provided (S110).

Description

Technical Field

[0001] The present invention relates to techniques for providing control signals indicating workflows for performing medical imaging post-processing, image analysis, and subsequent diagnosis with the aid of a plurality of radiology devices in a radiology device fleet, in particular including medical scanners. In particular, the techniques include methods, computing devices, systems including computing devices, computer program products, and computer-readable storage media. Background Art

[0002] Independent of grammatical term usage, individuals with either male or female identities are included in the term.

[0003] Running a radiology department or an independent radiology company is a challenging task. Typically, a "fleet" of several imaging systems including computed tomography (CT), mCT (molecular CT), magnetic resonance (MR), mMR (molecular MR) and X-ray equipment, SPECT, PET are used to perform medical imaging. In addition, digital imaging and communications in medicine (DICOM) routers, advanced visualization workplaces, artificial intelligence (AI) algorithms, and reading and reporting software need to be optimally set up to support the staff in performing diagnostic tasks. Surgical throughput, patient experience (and / or patient satisfaction), and high-quality results for the referral party are the most common optimization criteria. The referral party of the radiology entity decides to send and possibly resend the patient. Radiologists must provide high-quality results in a timely manner, such as to optimize productivity, asset usage, and obtain updated referrals.

[0004] Conventionally, radiology operations require a large number of site, user and indication specific configurations. Many systems are usually optimized separately and independently. Radiology staff (e.g., imaging technology experts, referred to as "technicians", radiologists and / or IT staff) usually collect patient data from different systems, and select imaging protocols, configure subsequent steps, such as reading tools, viewing and report templates. This conventional method is both time-consuming and error-prone. This usually does not produce a patient-specific optimal process that is optimized for the patient's clinical situation and suitable for answering the diagnostic tasks ordered and / or requested by the referral party for medical imaging.

[0005] Therefore, imaging procedures are often planned and executed based on data attributes that appear symptomatic for an action, but are not connected to an overall root cause or goal. Why the scan is needed in the first place and what goal the exam is intended to achieve are often unknown. Local rule sets implement incidental information around indications, reasons for the exam, and expected results. This is not necessarily optimal for the exam results. An example is that if the technical expert (referred to as the technician) at the scanner ensures that the DICOM series description contains the substring "LCAD", then lung computer-aided detection (referred to as "lung CAD") is performed by default. Therefore, a cost- and time-intensive process will be executed solely based on the request received without considering additional requirements. While this is a typical example, the full automation of the radiology department relies primarily or entirely on such rules that need to be independently configured in various radiology devices.

[0006] Conventional local processing rule sets are not optimized for the capabilities of available imaging devices. Resources are often not used optimally (e.g., optimally, faster CT scanners should be used for patients with arrhythmias and configured to prioritize speed over slower dual energy scanning modes).

[0007] Furthermore, local (and / or manual) configuration of, for example, workflows, tools, and / or reports is prone to errors and disruptions. Furthermore, conventional radiology operations typically involve extensive maintenance work by technicians and secondary personnel to address these process shortcomings.

[0008] "Fleet management" is usually focused on first collecting information about the operations performed on the scanner, and secondly uploading and downloading scan protocols to the scanner fleet. For all parameters (e.g., with technical parameters such as slices, magnetic fields, gradient strengths, rotation times, and software, including, for example, licenses, and in particular, the same scanning modality), the downloading of scan protocols is usually limited to the same (or quite similar) scanner model. Usually only the structure of the protocol name and the output data structure to DICOM are transmitted across the entire fleet. The transmission of the output structure is usually used to standardize the output of the scanner. For AI / Advanced Visualization (AI / AV) workstations, there is no comprehensive management solution on the market. In addition, there is no comprehensive fleet management for operators of scanner fleets that use multiple picture archiving and communication systems (PACS) at different locations.

[0009] Additionally, there are scanners with "exam companion" software that guides the technical specialist ("technician") by asking questions about the patient, such as "are there any bypasses?" Conventional exam companion software adjusts the scan protocol for the patient being scanned based on the answers. This helps translate the patient condition or exam goal into technical parameterizations for the scanner. The processing of question and answer pairs is typically a manual and error-prone process. It typically requires the radiology staff to find the corresponding information about the patient in existing documentation or by questioning the patient, and to manually and correctly enter the information. Summary of the invention

[0010] Therefore, the object of the present invention is to provide a solution for automating the optimal parameterization of radiology equipment involved in a diagnostic task, including medical scanners (in particular, including multiple scanning modalities), AI functions (e.g., Software as a Medical Device, SaMD, as defined by the International Medical Device Regulators Forum IMDRF), advanced visualization (AV), and / or PACS and / or report configurations and templates. The optimal parameterization of the radiology equipment includes the optimal selection of scan protocols and / or scan parameters based on the request of the referring party. Alternatively or additionally, a technique is needed for improving the quality of the requested medical imaging data affected by the conditions of the medical scanner fleet, the AI / AV equipment, the workplace, in particular, in view of their availability and / or technical specifications. Further alternatively or additionally, the object of the present invention is to minimize errors and avoid wear and / or increase maintenance intervals of equipment within a radiology equipment fleet (e.g., including medical scanners).

[0011] The object is solved by a method for providing control signals indicating a workflow for performing medical imaging (and optionally post-processing) by means of radiology devices (in particular, including medical scanners) in a radiology device fleet, by a computing device, by a system, by a computer program (and / or a computer program product) and by a computer-readable storage medium according to an embodiment. Advantageous aspects, features and embodiments and advantages are described in the embodiments and in the following description.

[0012] Hereinafter, the solution according to the present invention is described with respect to the claimed method and with respect to the claimed computing device. The definitions, features, advantages and / or alternative embodiments herein may be assigned to other claimed objects (e.g., systems, computer programs or computer program products), and other claimed objects (e.g., systems, computer programs or computer program products) may be assigned to the definitions, features, advantages and / or alternative embodiments herein. In other words, the features described or claimed in the context of the method may be used to improve the embodiments of the computing device and the system. In this case, the functional features of the method are embodied by the structural units of the computing device (and / or system), respectively, and vice versa.

[0013] Regarding the method aspect, a (particularly computer-implemented) method for providing a control signal indicating a workflow for performing medical imaging by means of a medical scanner in a radiology equipment fleet (particularly including a medical scanner). The method comprises a step of receiving first input data related to a request for medical imaging of a patient. The method further comprises a step of receiving second input data indicating the availability of at least one radiology equipment (particularly at least one medical scanner) within the radiology equipment fleet. The method further comprises a step of processing the received first input data and the received second input data. The processing comprises selecting at least one available radiology equipment (particularly a medical scanner) from the fleet. In the case of selecting more than one available radiology equipment (particularly a medical scanner), a time sequence in which the radiology equipment (particularly a medical scanner) is to be applied or used can be defined in the control signal. The processing is performed by means of a generative artificial intelligence (e.g., comprising a neural network, in particular, comprising a transformer architecture) and / or a large language model (LLM) for creating a workflow for medical imaging according to the request. The method further comprises the step of providing a control signal indicating a workflow for performing medical imaging by means of the selected radiology device, in particular the selected medical scanner.

[0014] Radiology equipment may include computing devices used within the workflow of medical imaging, post-processing, and diagnosis and treatment based on medical imaging. Alternatively or additionally, radiology equipment may include any equipment (medical or non-medical, such as, for example, a DICOM router) within a radiology department that is involved in the workflow of medical imaging, post-processing, and diagnosis and treatment based on medical imaging.

[0015] A radiology equipment cluster may, for example, include one or more medical or non-medical devices, one or more medical scanners, one or more DICOM routers, and / or one or more devices configured to perform medical imaging post-processing and / or to perform AI / AV functions.

[0016] The radiology equipment fleet may in particular comprise a fleet of (optionally: different types of) medical scanners.

[0017] Through the technology of the present invention, the workload balancing of radiology equipment (particularly, including medical scanners), and in particular, the workload balancing of medical scanners within a radiology equipment fleet, AI computing, AV and / or reading workplace can be significantly improved. Alternatively or additionally, the time period from request to execution of medical imaging (also expressed as examination and / or radiology process) can be improved. This may be particularly beneficial for emergency cases and / or emergencies, such as after an accident and / or in cases where a suspected serious illness needs to be treated in a timely manner. Alternatively or additionally, the quality of the requested medical imaging data (also expressed as medical images) can be improved, for example, by selecting the most suitable medical scanner (particularly, including its availability, its technical specifications, equipment and / or technical capabilities), the required AI / AV evaluation, the best visualization (e.g., including hanging protocols), the most suitable report template for a specific request within the fleet, and / or by optimizing the workflow of the selected radiology equipment (e.g., the selected medical scanner). Still further alternatively or additionally, through the present techniques, (e.g., transfer-related and / or human) errors (e.g., due to the medical image not corresponding to the request), premature deterioration and / or maintenance instances (and / or costs caused, for example, due to incorrect selection of the configuration of the medical scanner) can be avoided (and / or at least mitigated).

[0018] A radiology equipment fleet may include a plurality of radiology equipment operated (particularly jointly) by a fleet operator. A fleet operator may include a medical institution (particularly a hospital and / or a collection of hospitals in a predetermined local area), a radiology department and / or a radiology practice.

[0019] The second input data indicating the availability of at least one radiology device (in particular, at least one medical scanner) may be received passively (e.g., without a request by the generative AI and / or a computing device comprising the generative AI) and / or actively (e.g., in response to a request by the generative AI and / or with the aid of a measurement initiated by the generative AI and / or a computing device comprising the generative AI). Alternatively or additionally, the second input data indicating the availability of at least one radiology device (in particular, at least one medical scanner) may be received based on a push mechanism, a pull mechanism, or a combination thereof. For example, in the event of a change in availability (e.g., if the medical scanner is immediately available due to a medical emergency), the at least one radiology device (in particular, at least one medical scanner) may provide the second input data indicating its availability. The pull mechanism may include a request for, for example, periodically receiving second input data indicating the availability of at least one radiology device (particularly, at least one medical scanner) if first input data relating to a request for medical imaging of a patient is received (e.g., at a computing device comprising a generative AI) and / or if the generative AI for creating a workflow for medical imaging identifies at least one radiology device (particularly, at least one medical scanner) as (particularly, ideally) suitable for the request for medical imaging of the patient (e.g., due to its technical specifications and / or equipment).

[0020] For medical scanners, availability is the relevant optimization criterion. The subsequent paragraphs may also apply to optimization of concurrent use and / or availability of workplaces and / or licenses.

[0021] Selecting at least one available medical scanner (e.g. from a fleet and / or from a plurality of medical scanners) may include (particularly automatically and / or digitally) matching and / or assigning the available medical scanners to the patient to be examined. Alternatively or additionally, the selecting step may be performed algorithmically.

[0022] The selection step can process time conditions and / or location conditions so that the matching and / or assignment of "medical scanner-patient" takes into account the time period when the medical scanner is available and / or the patient is to be examined, and / or takes into account the location of the medical scanner and / or the patient (for example, by processing location-based information, such as GPS signals and / or indoor navigation such as ultra-wideband, UWB, tags).

[0023] Through the present techniques, improved image quality (leading to improved diagnosis) and improved patient outcomes in view of the requests received may be achieved.

[0024] The first input data may include the reason for the request and / or the organ and / or anatomical structure to be imaged. Alternatively or additionally, the first input data may include data related to the patient's health, such as age, height, weight and / or recently measured vital signs (e.g., blood pressure and / or heart rate). Further alternatively or additionally, the first input data may include laboratory (short: lab) values, previous surgeries (and / or amputations) and / or the presence of implants (e.g., pacemakers, stents, artificial joints and / or dental implants). Alternatively or additionally, the first input data may include existing examinations and / or the amount of existing examinations.

[0025] The request may include a preference and / or requirement for a predetermined medical imaging modality and / or an assessment to be performed.

[0026] The patient may be a human (and / or a person) or an animal.

[0027] The second input data may include an indication of when a medical scanner and / or another limited-time radiology device is available or unavailable or a certain (e.g., limited) function is available. This may also include the availability of required human experts such as technicians or radiologists. For example, a musculoskeletal specialist is typically only available during office hours.

[0028] The availability of a medical scanner (and / or any medical scanner within a fleet) may include: the medical scanner being in technically good condition, no previously scheduled medical imaging in the same time slot, and / or (in particular, specially trained) operators (and / or technicians, in short: technicians) being scheduled on site.

[0029] Alternatively or additionally, the radiology equipment (particularly the medical scanner) may be unavailable in case of a technical failure, lack of equipment, (e.g. scheduled) maintenance, no (particularly specially trained) operators available and / or the radiology equipment (particularly the medical scanner) is scheduled for other (e.g. urgent) requests, particularly those relating to another patient.

[0030] Either the first input data and the second input data may be received as raw data (not pre-processed, e.g., originating from different sources) and / or in a predetermined different format, e.g., text data, and / or in a different syntax. Alternatively or additionally, the first input data and / or the second input data may be received in a pre-processed form, e.g., in order to coordinate and unify different data sources and convert them into a common form that can be processed by the generative artificial intelligence.

[0031] Alternatively or additionally, any of the first input data and the second input data may be received in a different format (e.g., a format conventionally used to store and transmit the respective data). The (e.g., first and / or second) input data may be pre-processed to be available as pre-processed data (and / or data in a predetermined format, such as text data) for (particularly, further) processing.

[0032] A medical scanner (simply referred to as scanner) may also be referred to as a medical imaging device. Alternatively or additionally, the medical scanner may include a device for computed tomography (CT), a device for magnetic resonance tomography (MRT, also referred to as magnetic resonance imaging, MRI, or referred to as MR), a device for X-ray imaging (also referred to as radiography), an ultrasound (US) device, a device for single photon emission tomography (SPECT) and / or a device for positron emission tomography (PET).

[0033] Radiological devices (particularly, medical devices) in radiology may alternatively or additionally include AI functionality (e.g., as software, particularly, implemented on dedicated hardware and / or cloud-implemented), which can be used for example for automatic identification of lung nodules (LungCAD), automatic identification of pneumothorax and / or automatic assessment of cortical thickness in the brain.

[0034] Alternatively or additionally, the radiology device (particularly, medical) device may include an advanced visualization system that can perform automatic actions such as reconstruction and / or provide and / or supply interactive functions, for example to assess arteriosclerosis. If appropriate preprocessing has been automatically performed on the advanced visualization system, the interactive functions can be provided (and / or supplied) to the user (e.g., radiology staff) in a performance-optimized manner.

[0035] Timely and adequate triggering of preprocessing of the advanced visualization system may include functionality that may be automatically optimized within the scope of the present invention.

[0036] Further alternatively or additionally, diagnostic reading software and / or workplaces (eg, including hardware and / or software) may be included in a fleet of radiology equipment (particularly medical equipment) that may be optimized within the scope of the present invention, eg with respect to hanging protocols.

[0037] Still further alternatively or additionally, a radiology device (e.g., which is not classified as a medical device in most legislations) may include a DICOM router and / or a reporting system. The behavior and / or configuration of the DICOM router and / or the reporting system may also be optimized accordingly (e.g., which data needs to be routed to which workplace or which report template is used, in particular, for post-processing of acquired medical imaging data).

[0038] The radiology equipment fleet may include at least two different medical scanners.Alternatively or additionally, the radiology equipment fleet (in particular including medical scanners) may include at least two medical scanners using different medical imaging modalities (eg at least CT and MRT).

[0039] Two medical scanners using the same medical imaging modality may differ in their technical specifications and / or equipment (e.g., in terms of magnetic fields and / or field gradients for an MRT scanner, and / or the ability and / or equipment (e.g., the presence of coils) to perform specific functions such as dual energy scanning for a CT scanner.

[0040] Alternatively or additionally, radiology devices may differ by services provided (eg, post-processing, reconstruction, and / or AI / AV functionality), available licenses, processing speed, and / or location.

[0041] A medical scanner (and / or medical scanners within a fleet of radiology equipment) can be associated with a medical facility such as a hospital (and / or an association of hospitals, for example, in a town, region or other predefined area) or an outpatient facility such as a physician's office (and / or an alliance of physicians' offices) (in particular, a radiology department).

[0042] Receiving the first and / or second input data may be performed in an automatic manner by means of a computer via an electronic interface.The receiving may be regarded as digital receiving.

[0043] Providing a control signal involves providing a preferred intermediate result. The result including the control signal may be used to control a medical scanner for medical imaging, in particular during an acquisition phase (and / or an application phase). Providing is to be regarded as a digital providing, which may be performed by means of an interface, in particular a control signal providing interface.

[0044] The processing may be performed using a large language model (LLM), deep learning (DL), a base model, a generative model, and / or generative intelligence, particularly generative artificial intelligence (AI).

[0045] Generative AI (also referred to as: GenAI) may include a neural network (NN), for example, including a transformer architecture.

[0046] Generative AI may be able to generate text, images, and / or other media using a generative model.Alternatively or additionally, generative AI may learn the patterns and structure of its input training data and then generate new data with similar characteristics.

[0047] Generative AI may include deep learning (DL) and / or unsupervised and / or preferably self-supervised learning.

[0048] Generative AI can be unimodal or multimodal. For example, if all received (e.g., first and second) input data and optionally technical specifications and / or generally applicable instructions are preprocessed and normalized to a specific (particularly, text) format, then the generative AI can be unimodal. Alternatively, the generative AI is multimodal and can include the generative AI being trained on various (particularly, text) formats such as natural language text and / or programming language text.

[0049] Generative AI may include a base model (also referred to as a basic model). The base model may include a large machine learning (ML) model trained on a large amount of data (e.g., through self-supervised learning and / or semi-supervised learning) so that it can be adapted to a wide range of (e.g., downstream and / or subsequent) tasks.

[0050] Examples of base models include pre-trained language models (LMs) including Bidirectional Encoder Representations from Transformers (BERT), as described by J. Devlin et al. in arxiv:1810.04805[cs.CL][1], which is incorporated herein by reference. Additional examples of base models include GPT base models, such as the "GPT-n" series.

[0051] Alternatively or additionally, the generative AI may include an attention-based NN as described in A. Vaswani et al., “Attention is all you need,” arXiv:1706.03762[cs.CL][2], which is incorporated herein by reference.

[0052] Alternatively or additionally, the generative AI may include a transformer and / or a NN having a transformer architecture. A brief overview of the transformer model is provided at https: / / huggingface.co / docs / transformers / model_summary [3], which is incorporated herein by reference.

[0053] The transformer architecture may include a deep learning (DL) architecture including a particularly parallel and / or multi-headed attention mechanism.

[0054] The transformer architecture may include a bidirectional autoregressive transformer (BART), for example, as described by Lewis et al. in arXiv:1910.13461 [], which is incorporated herein by reference. Alternatively or additionally, the transformer architecture may include at least one decoder, and preferably at least one encoder. Further alternatively or additionally, the transformer architecture may include a BERT, as described by Rogers et al. in arXiv:2002.12327 [5], which is incorporated herein by reference.

[0055] The transformer architecture may be trained using historical (also denoted as real) and / or synthetic input data sets (in particular, comprising first input data relating to requests for medical imaging and second input data indicating the availability of one or more medical scanners within a fleet) having already associated or matched workflows for which associations or matches have been found to meet predefinable quality criteria as ground truth.

[0056] A neural network (NN) including a transformer architecture may include an LLM.

[0057] A NN including a transformer architecture can perform natural language processing (NLP) of medical documents. For example, processing of the first input data can include extracting relevant information for medical imaging from the received (and / or pre-processed) first input data related to the patient.

[0058] Alternatively or additionally, the generative AI may include a generative adversarial network (GAN), in particular, a generative adversarial transformer, as described by DA Hudson and CL Zitnick in arXiv:2103.01209 [cs.CV] [6], which is incorporated herein by reference.

[0059] Any generative AI may face the challenge of achieving training parallelism (e.g., compared to sequential training), low-cost inference, and / or strong performance. The challenge of achieving all three goals simultaneously can be referred to as the “impossible triangle” (e.g., as described by Y. Sun et al. in arXiv:2307.08621[cs.CL], [7]). Figure 2 ), where the linear transformer achieves training parallelism and low-cost inference but not string inference, where the recurrent neural network (RNN) achieves low-cost inference and strong performance but fails in training parallelism, and where the transformer achieves training parallelism and strong performance but fails in low-cost inference.

[0060] Still further alternatively or additionally, the generative AI may include a retention network (RetNet), as described by Y. Sun et al. in arXiv:2307.08621[cs.CL][7] and by S. Chandra in https: / / medium.com / ai-fusion-labs / retenBve-networks-retnetexplained-the-mu ch-awaited-transformers-killer-is-here-6c17e3e8add8[8], both of which are incorporated herein by reference.

[0061] RetNet can, for example, have better language modeling performance than any (linear and / or nonlinear) transformer or RNN, achieving better language modeling performance with (e.g., 3.4×) lower memory consumption, (e.g., 8.4×) higher throughput, and (e.g., 15.6×) lower latency.

[0062] In embodiments of generative AI, a local attention mechanism may be included. The local attention mechanism may enable a (e.g., very) large trained NN to operate with local focus. Alternatively or additionally, the generative AI may be post-trained (e.g., downstream) and / or fine-tuned (e.g., continuously).

[0063] In embodiments, a radiology device (and / or computing device) configured to perform the method may be trained and tested by a manufacturer and then shipped to its deployment location (e.g., a radiology department). The radiology device (and / or computing device) need not be configured to learn (and / or perform post-training) at the deployment site.

[0064] In other embodiments, the radiology device (and / or computing device) can be configured to continuously fine-tune during operation. For example, through user feedback and / or by observing workflows and / or reports. For example, if a user continuously overrides an automatically generated workflow, a previously trained radiology device (and / or computing device) can be optimized.

[0065] A workflow (also denoted as a schedule and / or a process workflow) may include at the scanner the selection of scan parameters (also: scan parameters; briefly: parameters), configuration of the medical scanner, a scan protocol (also denoted briefly as a protocol), a sequence plan and / or a scan sequence to be performed on the medical scanner. Thus, a workflow may include technical features (instructions and / or parameters and / or other features) for performing medical imaging.

[0066] Alternatively or additionally, the workflow may include AI functionality that is configured to perform steps of the diagnostic process, and / or it may be configured not to perform to save processing time, and / or it may be configured to save medical personnel time in evaluating the created data (e.g., the created data needs to be analyzed).

[0067] Further alternatively or additionally, the workflow may include AV functions, in particular AV functions for pre-processing the acquired medical imaging data.

[0068] Still further alternatively or additionally, the workflow may include visualizing data within the reading process, and / or selecting which data to display (eg, simultaneously) from available existing data.

[0069] Still further alternatively or additionally, the workflow may include providing a report template to use, and / or which report to fill out based on provided guidelines.

[0070] Radiology equipment can be selected and workflows created so that the requested medical imaging data are optimized in terms of quality, appropriateness of reason and / or timeliness of acquisition, in particular, all requests for all patients within a certain time period are optimized in view of the radiology equipment and / or available personnel for the fleet. Thus, the selection of radiology equipment (e.g., medical scanners) from the fleet can be performed at a superordinate level for all radiology equipment in the fleet (e.g., of a certain type, e.g., all medical scanners) and, for example, for all requests in common, and not just for one specific radiology equipment. For example, in the case of a suspected injury after an accident, the nearest suitable medical scanner can be selected in order to avoid further transportation, mechanical shocks or vibrations and / or time delays.

[0071] The control signal may be configured to enable the selected medical scanner to execute (particularly, the complete) workflow.Alternatively or additionally, the control signal may be provided in a format that the medical scanner can import.

[0072] For example, the steps of processing the received first and second input data and providing control signals may be repeated if the availability of the medical scanner changes, if the request is modified, and / or if further data related to the patient is received (e.g., laboratory data of the patient and / or previously acquired medical images, in particular including preferably hanging films, are received after the request is received based on the first input data).

[0073] A hang may refer to a hang protocol, particularly according to the DICOM standard, and may include a spatial arrangement of multiple medical views and / or multiple medical images (e.g., various cross-sectional views of organs and / or anatomical structures such as the head and / or views showing selected anatomical structures such as soft tissue, e.g., organs and / or anatomical structures of the brain or bone structures) on a computer screen. A hang (protocol) may refer to a set of viewing instructions that determines the layout and display of images. It allows a user to initiate viewing or study of images based on configurable attributes such as, for example, modality type, number of images, availability of comparison images, and / or many other user-defined criteria. Predefined hang protocols allow standardized viewing settings for specific images or exam types, thereby improving quality and efficiency.

[0074] The method may further comprise the step of receiving a technical specification of a radiology device (e.g., a medical scanner), and preferably receiving a technical specification of any radiology device (e.g., any medical scanner) within a fleet of radiology devices (e.g., including medical scanners). Optionally, the technical specification may comprise the presence and / or specification of particularly detachable equipment (e.g., one or more coils).

[0075] Technical specifications may, for example, include the size of the opening (eg, for CT), the maximum weight capacity (eg, of a patient table), the availability of scanning protocols, and / or the rotation time (eg, of a detector of a medical scanner).

[0076] Alternatively or additionally, the technical specifications may include magnetic field strengths and / or gradient field strengths (also: gradient strengths) for the MRT.

[0077] Further alternatively or additionally, the technical specifications may include the capability to image (eg, multiple) slices, the capability for dual energy medical imaging, the capability for photon counting, and / or the capability for physiological triggering.

[0078] "Physiological triggering" may refer to when medical image acquisition is performed relative to the patient's cardiac cycle and / or respiration. For example, cardiac triggering may be used in MRT and / or CT. Due to the fact that MRT is generally slower than CT, respiratory triggering and gating may be used there alternatively or additionally.

[0079] Gating may include taking a measurement (and / or, in particular, medical imaging, data acquisition) and acquiring, classifying or discarding data. Alternatively or additionally, triggering may include (in particular, medical imaging) data acquisition being performed (and / or occurring) triggered by, for example, the correct cardiac phase.

[0080] Still further alternatively or additionally, the technical specifications may include a list of acceptable scan parameters (e.g., including intervals and / or a set of values ​​for each scan parameter), possible configurations of the medical scanner, available scan protocols, available sequence plans, and / or available scan sequences.

[0081] The (particularly detachable) equipment (also: tool) may comprise one or more coils (eg radio frequency RF coils for MRT).

[0082] The technical specifications of any of the further radiology devices (eg a device comprising an AV function, an AI function and / or a DICOM router) may include a (particularly data) interface and / or software installed on the respective radiology device.

[0083] The technical specifications may be received (e.g. only once) after deployment of the radiology device (e.g. medical scanner). Alternatively or additionally, the technical specifications may be updated after maintenance and / or upgrading (e.g. software upgrading and / or updating and / or expansion of its equipment) of the radiology device, in particular the medical scanner. The technical specifications may be received continuously or repeatedly, in particular after deployment, and / or more than once and / or may change over time.

[0084] A feedback channel is provided for transferring a signal from a radiology device (e.g., a medical scanner, also referred to as: scanner) back to a computing device, which is configured to provide a control signal for the radiology device (e.g., medical scanner). Thus, a closed-loop control for controlling a radiology device (e.g., a medical scanner) in a fleet of radiology devices (in particular, including a medical scanner) is provided. In particular, second input data and / or optionally technical specifications of the radiology device (e.g., medical scanner) can be fed back to the computing device, in particular during processing. This allows the control signal to be dynamically adjusted to the actual situation at a specific radiology device (e.g., scanner). For example, if, for example, a workflow represented in a control signal has been established and an emergency is to be examined on the specific scanner in question that is assigned to a certain patient examination, a new schedule and / or a new workflow must be determined taking into account the specific emergency and the assignment (e.g., indicating that the original scanner is no longer available and another scanner needs to be assigned). Another example involves a crash or failure of a specific medical scanner and the transmission of information back to the computing device responsible for providing the control signal.

[0085] Alternatively or additionally, the method may further comprise the step of receiving generally applicable instructions related to medical imaging of the medical scanner fleet. The generally applicable instructions may in particular comprise one or more medical guidelines and / or one or more standard operating procedures (SOPs).

[0086] The generally applicable instructions may be cluster-specific (eg, depending on the medical imaging modality within the cluster) and / or specific to an operator of the cluster of medical scanners.

[0087] The received first input data relating to the request for medical imaging of the patient may comprise a reason for requesting the medical imaging, in particular comprising a reason for medical imaging of a suspected medical condition and / or a suspected lesion.

[0088] The cause may also be expressed as a medical problem.Alternatively or additionally, the cause may include an event, for example, the patient experienced an accident.

[0089] The first input data and / or the reason for the request may include an indication of the transportability of the patient (which may be reduced, for example, in case of expected concussion, severe injury and / or significant blood loss).

[0090] Alternatively or additionally, the first input data received in relation to the request for medical imaging of the patient may include one or more anatomical structures and / or one or more organs to be captured by the medical imaging.

[0091] The one or more anatomical structures (and / or organs) to be imaged may, for example, include the head region (in particular, including the brain, for example in case of a suspected aneurysm in the region), the chest (e.g., in particular, the heart and / or lungs) and / or the joint region (e.g., at the knees and / or hips).

[0092] Further alternatively or additionally, the first input data received in connection with the request for medical imaging of the patient may comprise information about the patient's general health state, in particular in view of the presence of a bypass, amputation, stent, pacemaker and / or any (in particular, further) implants.

[0093] Information regarding the patient's health status may be received or obtained from an electronic health record (EHR) and / or electronic medical record (EMR), laboratory results, and / or recorded vital signs.

[0094] Still further alternatively or additionally, the received first input data relating to the request for medical imaging of the patient may comprise existing medical images (existing examination) of the patient, in particular relating to the request for selecting a medical scanner according to quality requirements for providing medical images.

[0095] The existing medical image may, for example, indicate one or more requested views (and / or hangers) with respect to the current medical imaging. By comparing the existing medical image with the current medical imaging, lesions may be tracked and / or the success of the surgical procedure may be controlled.

[0096] Alternatively or additionally, existing medical images may indicate the location of bypasses, implants (eg, stents and / or pacemakers), and / or amputations.

[0097] At least a part of the first input data related to the request for medical imaging of the patient may be received by means of a user interface (UI), in particular a graphical user interface (GUI).

[0098] The request may be issued (and / or entered) by a health care professional (eg, a general practitioner and / or a specialist such as an internist, surgeon, orthopedist, and / or oncologist). The issuer of the request may also be denoted as a referring party.

[0099] The request may be received, for example, from a computer of the referring party.Alternatively or additionally, the UI (particularly the GUI) may be arranged at a location associated with the medical scanner fleet.

[0100] Processing may include selecting at least one medical scanner (or several medical scanners operating in sequence) based on the technical specifications of the medical scanners.Alternatively or additionally, processing may include determining a set of scan parameters for the workflow.

[0101] The processing may take into account previous medical images of patients with similar characteristics (also denoted prior examinations), for example in view of the equipment used, the measurements performed, the range of scan parameters used and / or the views selected.

[0102] Creation of workflows may also be based on historical data.Alternatively or additionally, a generative AI (e.g., implemented by a NN that includes, in particular, a transformer architecture) may be trained to select a medical scanner and (in particular, allowed for the medical scanner) scanning parameters based on historical data.

[0103] Scan parameters may also be denoted as operating parameters (and / or as scan settings, also simply settings). Alternatively or additionally, scan parameters may parameterize a workflow (eg, scan parameters may include a number of slices, particularly a CT scan, and / or a scan protocol, particularly an MRT scan).

[0104] The processing may alternatively or additionally comprise selecting a medical scanner and / or determining (particularly optimizing) a set of scanning parameters for a workflow according to a temporal sequence (particularly constraints on availability) of a (e.g. previously created) workflow of the medical scanner (and / or any one of the medical scanners within the fleet, particularly deemed a priori suitable for the request). Further alternatively or additionally, the processing may comprise selecting a medical scanner and / or determining (particularly optimizing) a set of scanning parameters for a workflow according to location constraints, e.g. prioritizing short transfer paths and / or short transfer times for patients with medical emergencies.

[0105] In some embodiments, processing can include selecting the medical scanner based on the received second input data, even if the medical scanner is occupied by other workflows, if the request for medical imaging includes a priority that is higher than the priority of other workflows (and / or if the other workflows have not yet been assigned a priority).

[0106] Generative AI can include NNs that include attention mechanisms.

[0107] Alternatively or additionally, the generative AI may include a NN that includes a transformer architecture.

[0108] Further alternatively or additionally, the generative AI may include a preserved network.

[0109] Still further alternatively or additionally, the generative AI may include a generative adversarial network (GAN), in particular a generative adversarial transformer.

[0110] Even further alternatively or additionally, the generative AI may include an LLM.

[0111] Generative AI can be trained (and / or pre-trained) using unsupervised, self-supervised, and / or semi-supervised learning.

[0112] The method may further comprise the step of preprocessing received first input data relating to a request for medical imaging of a patient. Alternatively or additionally, the method may further comprise the step of preprocessing received second input data indicating the availability of at least one medical scanner within the fleet. Further alternatively or additionally, the method may further comprise the step of preprocessing received technical specifications of the medical scanner. Further alternatively or additionally, the method may further comprise the step of preprocessing received generally applicable instructions.

[0113] The pre-processing step may be performed, in particular by a computing device, after receiving (eg, first and / or second) input data (eg, collectively and / or after completing an iterative reception of, in particular, the first input data).

[0114] Alternatively or additionally, first input data relating to a request for medical imaging of a patient may first (and / or separately) be pre-processed before being received, in particular by a computing device, for processing by a generative AI (e.g., including a NN).

[0115] Pre-processing may include structuring (e.g., first and / or second) input data, technical specifications and / or generally applicable instructions, and / or converting (e.g., first and / or second) input data, technical specifications and / or generally applicable instructions into pre-processed data and / or a general data format (particularly readable by a generative AI such as a NN that performs the processing). Alternatively or additionally, structuring may include sorting, for example, findings by organ and / or anatomical structure.

[0116] Alternatively or additionally, preprocessing may include normalizing the (e.g., first and / or second) input data, technical specifications and / or generally applicable instructions, in particular making them available in a common data format (e.g., from the raw data). Alternatively or additionally, normalizing the (e.g., first) input data and / or generally applicable instructions may include converting natural language text (e.g., text of a request, medical guideline and / or SOP) into text (and / or data) according to a predetermined ontology, such as the Systematic Nomenclature of Medicine (SNOMED) that is particularly suitable for processing by generative AI.

[0117] Preprocessing may be performed using one or more further NNs (e.g., one for each input data, e.g., depending on the first input data or the second input data and / or depending on the data type within the first input data including text and / or image information in a predetermined format) that include, for example, a transformer architecture. The (e.g., transformer) architecture of the further NNs used for preprocessing may be independent of (and / or different from) the (e.g., transformer) architecture of the NN used for processing.

[0118] In an alternative embodiment, the (e.g., transformer) architecture of the NN used for processing and the further NN used for pre-processing may be (e.g., at least partially) the same.

[0119] The method may further comprise the step of acquiring, by the selected medical scanner, medical imaging data based on the provided control signal, in particular in accordance with the request for medical imaging.

[0120] The acquired medical imaging data may also be stored in a (eg, Digital Imaging and Communications in Medicine, DICOM, and / or Picture Archiving and Communication System, PACS) database.

[0121] Alternatively or additionally, the method may further comprise the step of post-processing the acquired medical imaging data.

[0122] Post-processing may include reconstructing a medical image from the acquired medical imaging data.Alternatively or additionally, post-processing may include preparing (also denoted as pre-processing) the acquired medical imaging data for further use, such as for segmentation, lung CAD and / or diagnostic purposes.

[0123] Post-processing may include, for example, selecting one or more views and / or tiles of views according to a request included in the first input data.

[0124] Alternatively or additionally, post-processing may include using AI / AV (Advanced Visualization) software on the acquired medical imaging data and / or analyzing the acquired medical imaging data for reasons included in the request. For example, a segmentation (e.g. of organs and / or anatomical structures) may be performed, and / or the extension of a lesion may be measured (particularly automatically).

[0125] The AI / AV software (e.g., including syngo.via, also referred to as via, and / or AI Rad Companion, AIRC) may include analysis modules and / or analysis functions. Alternatively or additionally, the AI / AV software may be configured to analyze, evaluate, interactively visualize, and / or automatically visualize (e.g., acquired) medical imaging data.

[0126] The AI / AV software may, for example, be configured for 3D rendering of diffusion imaging and / or perfusion imaging related to the acquired medical imaging data.

[0127] By the AI / AV software, a visualization may be provided to a user (e.g., a radiology staff member) (e.g., in a first iterative step). The user may confirm or reject the provided visualization. Alternatively or additionally, the user may provide instructions and / or indications for a modified visualization, which are provided by the AI / AV software (e.g., in a second iterative step).

[0128] Still alternatively or additionally, post-processing may include preparing and / or transmitting a (particularly textual) report regarding the medical imaging data and / or findings therefrom.

[0129] Post-processing may utilize, for example, one or more further NNs including a transformer architecture. The (e.g., transformer) architecture of the further NNs (e.g., one for each output data) may be independent of the (e.g., transformer) architecture of the NN performing the processing and / or the (e.g., transformer) architecture of the one or more further NNs used for pre-processing.

[0130] Post-processing can be performed locally at a medical scanner where medical imaging data is acquired by a computing device (e.g., a workstation), at a computing device (e.g., a workstation) including a display, externally (e.g., at a central computing location), cloud-based, and / or any combination thereof.

[0131] The method may further include the step of displaying the created workflow using a UI, particularly a GUI. Alternatively or additionally, the method may further include the step of displaying (e.g., on a UI, particularly a GUI) the acquired and / or post-processed medical image data. In one embodiment, the display (e.g., UI, particularly GUI) for the created workflow and the acquired and / or post-processed medical image data may be independent of each other.

[0132] By means of the created workflow and / or display of the medical imaging data, a (particularly human) operator (and / or technician) of the medical scanner can verify the correctness of the created workflow and the medical imaging data acquired thereby. Thereby, if deviations from the request are detected, adjustments and / or corrections can be performed in a timely manner. Alternatively or additionally, scheduling another appointment for further medical imaging of the patient can be avoided.

[0133] Medical imaging may be performed with the aid of medical imaging modalities such as CT, MRI, X-ray imaging (also denoted radiography), ultrasound / US, SPECT and / or PET and / or for molecular devices like mCT, mMRI.

[0134] The medical scanner fleet may include at least two medical scanners associated with different medical imaging modalities. The fleet may, for example, include at least one CT scanner and at least one MRT scanner.

[0135] Depending on the reason for requesting medical imaging, processing by the generative AI (e.g., a NN that specifically includes a transformer architecture) may include selecting a medical imaging modality.

[0136] The method can be used to schedule multiple patients across a fleet of medical scanners.

[0137] The present technology can be used to streamline and / or simplify the scheduling of all medical imaging requests and / or medical scanner fleets. For example, a combined schedule of fleets can be iteratively provided in response to received requests for medical imaging.

[0138] With respect to the device aspect, a computing device for providing a control signal is provided, the control signal indicating a workflow for performing medical imaging with the aid of a medical scanner in a radiology equipment fleet (particularly, including a medical scanner). The computing device includes a first input data receiving interface, which is configured to receive first input data related to a request for medical imaging of a patient. The computing device also includes a second input data receiving interface, which is configured to receive second input data indicating the availability of at least one radiology device (particularly at least one medical scanner) within the radio equipment fleet. The computing device also includes a processing unit, which is configured to process the received first input data and the received second input data. The processing includes selecting at least one available radiology device (particularly a medical scanner) from the fleet. The processing is performed with the aid of a generative AI (e.g., a NN and / or LLM including a transformer architecture) for creating a workflow for medical imaging according to the request. The computing device also includes a control signal providing interface, which is configured to provide a control signal indicating a workflow for performing medical imaging with the aid of the selected radiology device (particularly the selected medical scanner).

[0139] Optionally, the computing device may include a technical specification receiving interface configured to receive technical specifications of the medical scanner, preferably for receiving technical specifications of any radio device within the medical scanner fleet (and / or in particular for any medical scanner). Optionally, the technical specifications include the presence and / or specifications of in particular detachable equipment (e.g. one or more coils).

[0140] Alternatively or additionally, the computing device may include a generally applicable instruction receiving interface configured to receive generally applicable instructions related to medical imaging performed by a radiology device medical scanner fleet, particularly including one or more medical guidelines and / or one or more SOPs.

[0141] Further alternatively or additionally, the computing device may include a pre-processing unit. The pre-processing unit may be configured to pre-process received first input data related to a request for medical imaging of a patient. Alternatively or additionally, the pre-processing unit may be configured to pre-process received second input data indicating availability of at least one medical scanner within the fleet. Further alternatively or additionally, the pre-processing unit may be configured to pre-process received technical specifications of the medical scanner. Still further alternatively or additionally, the pre-processing unit may be configured to pre-process received generally applicable instructions.

[0142] Further alternatively or additionally, the computing device may include a medical imaging data acquisition interface configured to receive medical imaging data acquired by the selected medical scanner based on the provided control signal.

[0143] Still further alternatively or additionally, the computing device may include a post-processing unit configured for post-processing the acquired medical imaging data.

[0144] Alternatively or additionally, the computing device may include a workflow display interface configured to display the created workflow using a UI (particularly a GUI). Also alternatively or additionally, the computing device may include a medical imaging data display interface configured to display (e.g., on a UI, particularly a GUI) the acquired and / or post-processed medical image data. The created workflow and the display (e.g., UI, particularly GUI) of the acquired and / or post-processed medical image data may be independent of each other.

[0145] The computing device may be configured to perform any of the steps or include any of the features described in the context of the method aspects.

[0146] Regarding the system aspect, a system for providing control signals indicating a workflow for performing medical imaging with the aid of a medical scanner in a cluster of radiology devices (particularly including a medical scanner) is provided. The system includes a computing device according to the device aspect. The system also includes a cluster of radiology devices (particularly including a medical scanner). Each radiology device in the cluster (particularly each medical scanner) may include a control signal receiving interface configured to receive the control signal from a control signal providing interface of the computing device.

[0147] The system may be cloud-based.

[0148] Regarding a further aspect, a computer program product is provided. The computer program product comprises a program element which, when loaded into a memory of a computing device, causes the computing device to perform the steps of a method for providing a control signal indicative of a workflow for performing medical imaging with the aid of a medical scanner in a fleet of radiology devices, in particular including a medical scanner, according to a method aspect.

[0149] Regarding yet another aspect, a computer readable medium having program elements stored thereon is provided. The program elements can be read and executed by a computing device so that when the program elements are executed by the computing device, according to a method aspect, the steps of a method for providing a control signal are performed, the control signal indicating a workflow for performing medical imaging with the aid of a medical scanner in a fleet of radiology equipment (particularly, including medical scanners).

[0150] The above-mentioned characteristics, features and advantages of the present invention and the manner of achieving them will become clearer and more easily understood from the following description and embodiments which will be described in more detail in the context of the accompanying drawings.

[0151] The following description does not limit the invention to the embodiments included. In different figures, the same components or parts may be marked with the same reference numerals. In general, the drawings are not to scale.

[0152] It should be understood that a preferred embodiment of the present invention may also be any combination of the above embodiments.

[0153] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS

[0154] Figure 1 is a flow chart of a method for providing a control signal indicative of a workflow for performing medical imaging by means of a medical scanner in a medical scanner fleet according to a preferred embodiment of the present invention;

[0155] Figure 2 is an overview of the structure and architecture of a computing device for providing control signals indicative of a workflow for performing medical imaging with the aid of medical scanners in a fleet of medical scanners according to a preferred embodiment of the present invention;

[0156] Figure 3 Schematically shows the steps for executing Figure 1 details of the types of data provided by the method;

[0157] Figure 4A and Figure 4B Schematically shows Figure 4A An example of a previous medical image hang-up in Figure 1 The method pre-processes, processes and / or post-processes the information according to the medical imaging request, thereby obtaining Figure 4B Similar medical image clips in

[0158] Figure 5 The example shows Figure 1 The method uses technical specifications and parameters for performing medical imaging with a medical scanner from a medical scanner fleet; and

[0159] Figure 6 is a schematic diagram of a closed-loop control with a backchannel from a medical scanner to a computing device.

[0160] Any reference signs in the detailed description should not be construed as limiting the scope. DETAILED DESCRIPTION

[0161] Figure 1 An exemplary flow chart of a (particularly computer-implemented) method for providing control signals indicative of a workflow for performing medical imaging by means of medical scanners in a fleet of medical scanners is schematically shown. The method is generally referred to by reference numeral 100 .

[0162] exist Figure 1 and Figures 2 to 6 In any further embodiments of the present invention, the medical scanner cluster may be included in the radiology equipment cluster.

[0163] The method 100 includes a step S104 -A of receiving first input data related to a request for medical imaging of a patient.

[0164] The method 100 further comprises a step S104 -B of receiving second input data indicating the availability of at least one medical scanner within the medical scanner fleet.

[0165] The method 100 further comprises a step S108 of processing the first input data received S104-A and the second input data received S104-B. The processing S108 comprises selecting an available medical scanner from the fleet. The processing S108 is performed with the aid of a generative artificial intelligence (e.g., an AI comprising a neural network and / or a large language model LLM including a transformer architecture) for creating a workflow for medical imaging according to the request.

[0166] The method 100 further comprises a step S110 of providing a control signal indicative of a workflow for performing medical imaging by means of the selected medical scanner.

[0167] Optionally, the method 100 comprises a step S102-A of receiving technical specifications of the medical scanner (and / or preferably for any medical scanner within the medical scanner fleet). Optionally, the technical specifications comprise in particular the presence and / or specifications of detachable equipment (e.g. one or more coils).

[0168] Additionally optionally, the method 100 comprises a step S102 -B of receiving generally applicable instructions related to medical imaging of the medical scanner fleet, in particular comprising one or more medical guidelines and / or one or more standard operating procedures (SOPs).

[0169] Still further optionally, the method 100 comprises a step of preprocessing S106. Preprocessing S106 may comprise preprocessing first input data received S104-A relating to a request for medical imaging of a patient. Alternatively or additionally, preprocessing S106 may comprise preprocessing second input data received S104-B indicating the availability of at least one medical scanner within the fleet. Still further alternatively or additionally, preprocessing S106 may comprise preprocessing technical specifications of the medical scanner received S102-A. Still further alternatively or additionally, preprocessing S106 may comprise preprocessing generally applicable instructions received S102-B.

[0170] The method 100 may further include a step S112 of acquiring, by the selected medical scanner, medical imaging data based on the provided S110 control signal.

[0171] Alternatively or additionally, the method 100 may comprise a step S114 of post-processing the acquired S112 medical imaging data.

[0172] The method 100 may comprise a step S109 of displaying the created workflow using a user interface (UI), in particular a graphical user interface (GUI). Alternatively or additionally, the method 100 may comprise a step S115 of displaying the acquired S112 and / or post-processed S114 medical image data, for example using the same or another UI (in particular the same or another GUI).

[0173] Figure 2 An exemplary architecture of a computing device for performing medical imaging with the aid of medical scanners in a medical scanner fleet is schematically illustrated.The computing device is generally referred to by reference numeral 200.

[0174] The computing device 200 includes a first input data receiving interface 204 -A configured to receive first input data related to a request for medical imaging of a patient.

[0175] The computing device 200 further includes a second input data receiving interface 204 -B configured to receive second input data indicating availability of at least one medical scanner within the medical scanner fleet.

[0176] The computing device 200 further comprises a processing unit 208 configured to process the received first input data and the received second input data. The processing comprises selecting an available medical scanner from the fleet. The processing is performed with the aid of a generative AI (e.g., by a NN including in particular a transformer architecture and / or an LLM) for creating a workflow for medical imaging according to the request.

[0177] The computing device 200 further comprises a control signal providing interface 210 configured for providing a control signal indicative of a workflow for performing medical imaging by means of the selected medical scanner.

[0178] Optionally, the computing device 200 includes a technical specification receiving interface 202-A configured to receive technical specifications of a medical scanner, preferably technical specifications for any medical scanner within a fleet of medical scanners. Optionally, the technical specifications include the presence and / or specifications of, in particular, detachable equipment (e.g., one or more coils).

[0179] Alternatively, the computing device 200 includes a generally applicable instruction receiving interface 202-B configured to receive generally applicable instructions related to medical imaging performed by a medical scanner fleet, particularly including one or more medical guidelines and / or one or more SOPs.

[0180] Still further alternatively, the computing device 200 may include a pre-processing unit 208. The pre-processing unit 208 may be configured to pre-process received first input data related to a request for medical imaging of a patient. Alternatively or additionally, the pre-processing unit 208 may be configured to pre-process received second input data indicating availability of at least one medical scanner within the fleet. Still further alternatively or additionally, the pre-processing unit 208 may be configured to pre-process received technical specifications of the medical scanner. Still further alternatively or additionally, the pre-processing unit 208 may be configured to pre-process received generally applicable instructions.

[0181] The computing device 200 may include a medical imaging data acquisition interface 212 configured to receive medical imaging data acquired by the selected medical scanner based on the provided control signal.

[0182] The computing device 200 may further include a post-processing unit 214 configured to post-process the acquired medical imaging data.

[0183] The computing device 200 may further include a workflow display interface 209 configured to provide the created workflow for display using a UI, particularly a GUI. Alternatively or additionally, the computing device 200 may include a medical imaging data display interface 215 configured to provide acquired and / or post-processed medical image data for display, for example, on the same or another UI, particularly the same or another GUI.

[0184] A first input data receiving interface 204 -A, a second input data receiving interface 204 -B, an optional technical specification receiving interface 202 -A, an optional generally applicable instruction receiving interface 202 -B and / or an optional medical imaging data acquisition interface 212 may be included in the input interface 216 .

[0185] A control signal providing interface 210 , an optional workflow display interface 209 and / or an optional medical imaging data display interface 215 may be included in the output interface 218 .

[0186] Alternatively or additionally, any of the interface 204 -A; the interface 204 -B; the interface 202 -A; the interface 202 -B; the interface 212 and / or the interface 210 ; the interface 209 ; the interface 215 may be included in the input-output interface 220 .

[0187] The processing unit 208 , the optional pre-processing unit 206 , and / or the optional post-processing unit 214 may be implemented by a processor 222 .

[0188] Any of the processing unit 208, the optional pre-processing unit 206, the optional post-processing unit 214, and / or the processor 222 may be implemented by a central processing unit (CPU) and / or a graphics processing unit (GPU).

[0189] Computing device 200 may also include memory 224 .

[0190] The computing device 200 may be configured to perform the method 100. For example, the interface 204-A, the interface 204-B, the processing unit 208, and the interface 210 may be configured to respectively perform step S104-A, step S104-B, step S108, and step S110.

[0191] The system may include a computing device 200 and a medical scanner cluster. Each medical scanner in the cluster may include a control signal receiving interface configured to receive a control signal from a control signal providing interface 210 of the computing device 200.

[0192] The one or more technical specifications and the one or more availability indications received by the technical specification receiving interface 202 -A and the second input data receiving interface 204 -B of the computing device 200 , respectively, may be associated with one or more (particularly each) medical scanner within the fleet.

[0193] The system may be configured to perform method 100 .

[0194] By means of the present technology (e.g., including method 100 and / or computing device 200), a fleet of medical scanners (and / or one or more radiology systems) can be configured and managed (also: controlled). Medical scanners may include, for example, MR devices, CT devices, and / or XR devices.

[0195] AI / AV software (e.g., syngo.via, via and / or AI Rad Companion, AIRC) and / or reading and reporting software (e.g., Carbon and / or Plaza) can be used in the context of (and / or in extension of) the present technology.

[0196] The specific clinical workflow for each patient and imaging procedure can (particularly according to the present technology) be optimized end-to-end from exam scheduling to the reading and reporting process.

[0197] Figure 3 An exemplary (e.g., radiology operation) system with information, reasoning, and action levels is shown. The action level can include events and / or plan deviations, such as equipment (particularly medical scanners 304 within a fleet) failing and / or patients not showing up. Information can be fed back to the information level, as illustrated by reference numeral 309, which in turn changes the plan at the reasoning level, as schematically depicted by reference numeral S104-B.

[0198] The (e.g., radiology operating) system can (particularly due to the present technology) automatically adapt to each of the planned (and / or current and / or future) patient surgical cases, is highly flexible to changing input information, and can handle events such as patient absence or failure of one or more medical scanners (also referred to as equipment failure).

[0199] exist Figure 3 , information levels with different types of received information and / or input data are schematically shown. Exemplarily, existing patient information is shown at reference numeral 302-1 (e.g., as part of the first input data received in step 104-A). Further exemplarily, technical specifications of a medical scanner are shown at reference numeral 302-2 (particularly according to receiving step 302-A). At reference numerals 302-3 and 302-4, site-specific information and operating procedure-specific information are further exemplarily shown, respectively (e.g., as received according to step 102-B). At reference numeral 302-5, information about the availability of one or more medical scanners in the fleet is exemplarily shown (e.g., received in step 104-B and / or updated when the availability changes, e.g., due to technical problems and / or unforeseen emergency situations).

[0200] At reference numeral 304, a medical scanner fleet is schematically depicted. At reference numeral 306, multiple devices (e.g., workstations) for performing AV functions (e.g., for performing any of step S109; step S115) are exemplarily shown. At reference numeral 308, multiple instances of devices with AI functions (e.g., for performing post-processing step S114) are schematically depicted.

[0201] Any of device 304; device 306; device 308 may be included in a radiology equipment fleet.

[0202] As schematically indicated by reference numerals 310 and 312 , the acquired medical imaging data may be provided to any of the device for performing AV functionality 306 and / or the device having AI functionality 308 .

[0203] As schematically indicated at reference numeral 314, devices 306 and 308 may work in combination (and / or may be combined in a joint device) to perform AI / AV functionality.

[0204] Any of the (e.g., incoming and / or acquired medical imaging) data forwarding may be routed through one or more DICOM routers (e.g., such as Figure 3 ).

[0205] Reasoning (e.g., in step S108 of method 100) may include selecting the appropriate scanner, configuring appropriate scanning parameters, standardizing the read output, triggering the appropriate AI / AV algorithm appropriate to the patient's clinical condition, reproducibly displaying the data in the appropriate view and / or hanging slide, providing the correct tools, and preparing and / or sending a report to the referrer. "Right" herein may be used synonymously with "suitable," "appropriate," and / or particularly "optimal for the request."

[0206] Natural language processing (NLP) techniques can automatically analyze medical documents, such as reports, including radiology reports and / or referral letters. One technical implementation is the large language model (LLM). Such LLMs have been shown to be effective in tasks such as text generation and language translation. LLMs have also been applied to the analysis of medical texts, including extracting information from electronic medical records (EMRs), in particular Miotto et al., 2016 [9], which is incorporated herein by reference.

[0207] The present technology (e.g., including method 100 and / or computing device 200) utilizes algorithms, AI and / or (e.g., preferably) LLM to analyze multiple information sources. Any combination of information sources is meaningful, such as Figure 3302-5. The information (e.g., 302-1; 302-2; 302-3; 302-4; 302-5) can be directly obtained using common algorithms based on AI and / or specifically LLM. LLM can be used directly for information (e.g., 302-1; 302-2; 302-3; 302-4; 302-5) and / or for previously generated embeddings of information sources. Alternatively or additionally, the information (e.g., 302-1; 302-2; 302-3; 302-4; 302-5) can be pre-processed, for example, for generating embeddings.

[0208] Existing patient information 302 - 1 may include, for example, referral orders, electronic health record (EHR) patient information, images, radiology reports, and / or laboratory reports.

[0209] Existing information 302-1, in particular reports and / or referrals, may be analyzed using the (particularly generated) AI 308 and / or LLM. Analysis of existing information 302-1 may provide an indication of a request for medical imaging and / or an examination.

[0210] Analysis of prior information 302-1 may provide equipment, tools, scanning parameters, and / or measurements for prior examinations, previous procedures, and / or for similar patients. Similar patients may be associated with similar reasons for medical imaging (e.g., similar intended diagnoses and / or similar incidents).

[0211] Alternatively or additionally, the prior information 302 - 1 may provide views used and / or created in prior examinations, previous procedures, and / or for similar patients.

[0212] Alternatively or additionally, prior information 302 - 1 may provide quantitative data related to the current examination obtained manually, semi-automatically, and / or automatically (eg, by using software tools such as, for example, AIRC, via, and / or syngo.via on current and prior data).

[0213] Alternatively or additionally, the prior information 302 - 1 may provide information for prior examinations, previous procedures, and / or view configurations for similar patients.

[0214] Alternatively or additionally, the existing information 302 - 1 may provide existing radiology reports for the same patient associated with the request, the existing radiology reports being related to the reason for the medical imaging examination and / or similar patients.

[0215] Alternatively or additionally, the existing information 302 - 1 may provide existing medical examinations, such as blood values, reasons for examinations and surgeries, particularly with respect to the patient to whom the request refers.

[0216] The existing information 302-1 can be output in a structured manner. The following example shows how to prompt an LLM (e.g., GPT4-32k) to create a JavaScript Object Notation (JSON) structure about the findings from an unstructured report:

[0217] Report:

[0218] Medical History: 56-year-old female presented with complaints of low back pain, post-motor vehicle accident status.

[0219] Technique: Multiple views of lumbosacral CR are obtained.

[0220] Findings:

[0221] The patient had S1 spondylosis, which basically means the body of the sixth lumbar vertebra. There was anterior wedging of L2. The fracture healed, involving the anterior superior endplate and body of L4. This was probably due to acute flexion of the lumbosacral spine. There was bilateral sacroiliac joint narrowing.

[0222] Print:

[0223] 1. The patient has a segmental abnormality. There are six vertebrae in the lumbar spine.

[0224] 2. L2 has anterior wedging and L4 has a deformity involving the superior anterior endplate, consistent with fracture union. Tip: Create a structured report. Sort findings by organ and anatomy. Output as JSON.

[0225] The output of the above example prompt related to reporting is as follows:

[0226]

[0227]

[0228]

[0229] The capabilities of the medical scanners and / or equipment (and / or more generally the assets) involved play a key role in defining the optimal workflow.

[0230] The capabilities of the MRT scanner (also denoted as a magnetic resonance imaging, MRI device) (e.g., based on the technical specifications received in step S102-A) may include hardware and software capabilities, such as magnetic field strength, gradient strength, one or more receiver channels, one or more available radio frequency (rf) coils, one or more available MRI sequences and / or protocols, and / or available software licenses.

[0231] The capabilities of the CT scanner (also denoted CT device) (e.g., based on the technical specifications received in step S102-A) may include hardware and software capabilities, such as (particularly multiple) slices, detector width, dual energy capability, multiple coils, photon counting capability, physiological triggering, available CT software and / or protocols.

[0232] AI / AV (e.g., on a medical scanner, locally and / or in the cloud) capabilities (e.g., for post-processing step S114) can include hardware and software capabilities, such as availability of lung or colon CAD, cerebral hemorrhage detection, coronary artery assessment, and / or the number of available seats and / or users.

[0233] Capabilities for reading and / or reporting (e.g., for post-processing step S114 and / or for displaying either of steps S109 and S115) may include hardware and software capabilities, such as available monitor configurations, one or more AI / AV tools, and / or rendering capabilities.

[0234] Alternatively, information about the asset may include availability (eg, instantaneous and / or time-limited).

[0235] Information about capabilities and / or availability may exist and / or be provided in a digitally structured manner. The assets involved (e.g., medical scanners, equipment, AI / AV, and / or reading and reporting tools within a fleet) may require an application programming interface (API) to bring their information to the inference level in a structured manner. The information may be standardized and placed into a regularly updated database. Real-time status information may be included in a preferred implementation, such as describing the current and future availability of assets (e.g., medical scanners, equipment, AI / AV, and / or reading and reporting tools within a fleet) for use.

[0236] For two MRT scanners, the standardized information can be exemplarily read as follows:

[0237]

[0238]

[0239]

[0240] Structured information may still require (and / or need) an ontology (eg, for radiology equipment and / or medical devices) that specifically includes technical specifications (eg, of medical scanners across multiple medical imaging modalities) to make capabilities comparable.

[0241] Referral orders must (and / or typically are) entered into the fleet's (and / or radiology entity's) calendar. Typically, an electronic health record (EHR) scheduling system (e.g., primarily in the U.S.), a traditional radiology information system (RIS, e.g., in both the U.S. and Europe), and / or (e.g., B2C) calendar (and / or scheduling) tools, such as Outlook, can be used to manage scheduling slots for patients and / or (e.g., typically also) personnel.

[0242] Field personnel typically interact directly with patients (e.g., in an outpatient setting) and / or with clinical departments (e.g., in an inpatient setting) to schedule appointments for medical imaging (and / or radiology procedures). Often, it may be easy to coordinate multiple (e.g., five) calendars to implement a procedure.

[0243] In a preferred embodiment, the present technology (eg, including method 100, computing device 200, and / or particularly radiology operating system) uses patient scheduling information to plan and coordinate various surgical workflows. Scheduling information can be precise time slots or general availability.

[0244] In another preferred embodiment, the present technology (e.g., including the method 100, the computing device 200, and / or the radiology operating system in particular) includes a back channel from the action layer to the scheduling system, so that events such as delayed scans and / or failures of medical scanners (and / or systems) can be relayed back to the scheduling system and the patient (e.g., as Figure 3 denoted by reference numeral S109 in the figure).

[0245] Traditional fleet management solutions usually provide scanner productivity analysis based on the created DICOM data. Some solutions (such as Team Collaboration Insights) also use log data. However, using DICOM data and / or using log data is a costly and cumbersome activity, as the log syntax is mostly non-standardized and manually created filters are required to import such data into (particularly site-specific) performance metrics.

[0246] Using algorithms, particularly generative AI (preferably LLM), it can be derived where which scan parameters are used, what the actual slot times are, and / or which image contrasts are created based on which scan parameters. Any of these types of information can be standardized and put into a regularly updated database.

[0247] The preferences of operators of a fleet of medical scanners can be expressed through behavioral patterns (e.g., of technicians and / or radiologists), SOPs and / or clinical (and / or legal) guidelines (short: guidelines). Clinical (and / or legal) guidelines are typically expressed as written documents. Although actual behavior is typically recorded in (particularly site-specific) performance indicators, information based on SOPs and guidelines is typically not available for automated workflow optimization. Manual configuration of radiology equipment is typically not performed. For example, conventionally, DICOM routers are not (particularly manually) configured for optimizing workflows. Alternatively or additionally, settings for performing (and / or when to perform) lung CAD are typically not configured. Also alternatively or additionally, conventionally, configurations for (particularly image) reconstruction are not (particularly manually) optimized.

[0248] Using generative AI (e.g., one or more LLMs), SOPs and / or guidelines can be interpreted from human language and converted (and / or standardized and / or normalized) into a machine-interpretable structure. Preferably, the conversion (also: standardization and / or normalization) is combined with a classification algorithm (e.g., an image basis model) that can classify the images. Such information (e.g., including classifications, standardized SOPs, and / or standardized guidelines) can, for example, include rules on how to view a specific case in a PACS. For example, a SOP can include the definition of a so-called "hanging" or "hanging protocol," such as Figure 4A and 4B exemplified in .

[0249] exist Figure 4A , prior (also: existing) medical imaging data (e.g., of a patient and / or from a CT scan) of a human abdomen is shown. At reference numeral 402-A, an axial view is provided. At reference numeral 404-A, a sagittal reconstruction view is provided, and at reference numeral 406-A, a coronal reconstruction view of the abdomen is provided. Any of views 402-A; 404-A; 406-A can be prepared with a window selected to display soft tissue.

[0250] For the case where no contrast agent was used when acquiring the medical imaging data, (particularly the previous hanging film and / or part thereof) can be converted into, for example:

[0251] Segment Col1, Row1 = {contrast = non-contrast}, {time = current}, {orientation = axial}, {window = soft tissue}

[0252] Segment Col1, Row2 = {Contrast = Non-contrast}, {Time = Current}, {Orientation = Sagittal}, {Window = Soft Tissue}.

[0253] Figure 4B With Figure 4AThe same view (e.g., orientation and particularly soft tissue displayed) of the same order and / or same hanging slide for medical imaging performed according to the present technology (e.g., by means of CT and / or for the same patient involved in the request) of the previous medical imaging data (e.g., obtained as the first input data in step 104-A). For example, axial, sagittal reconstruction and coronal reconstruction views are shown at reference numerals 402-B; 404-B and 406-B.

[0254] By providing the same hanging patch, visual inspection and / or diagnosis by medical practitioners (particularly referring parties) is improved.

[0255] The selection of views and / or panels can be done, for example, by Figure 3 Executed by one of the computers 306 in .

[0256] At the inference level, information level data can be combined to create and continually update the optimal workflow for each patient procedure.

[0257] In an embodiment, the MRT imaging process is configured and planned based on the referral order (particularly, included in the first input data received in step 104-A) and scheduling information (particularly, included in the second input data received in step 104-B) as well as additional patient information (particularly also included in the first input data received in step 104-A) such as age and MRT scan counter-indications, and the SOP of the radiology department (particularly, included in the generally applicable instructions in step S102-B).

[0258] In other embodiments, the CT imaging procedure is configured and planned on the best available (or good enough) scanner for the imaging task. Photon counting CT may be preferred due to advantages in resolution, dual source CT may be preferred in performing cardiac examinations, and / or wide bore scanners may be preferred for obese patients.

[0259] Including according to Figure 4A In another embodiment of the example of and / or 4B for reading a hanger, information about the clinical indication (and / or reason) and additional structured information about the scan are used to configure the reading screen content. The hanger is aligned closer to the actual needs of the radiologist for reading the individual case.

[0260] An exemplary workflow (and / or workflow plan) may be read in human language as follows:

[0261] Based on the patient's history and symptoms, CT scans of the following areas may be recommended:

[0262] 1. High-resolution CT (HRCT) scan of the chest:

[0263] This will help evaluate the lung parenchyma and airways for any underlying lung disease, infection, or other abnormalities that may be causing the cough and chest pain.

[0264] 2. Paranasal sinus CT scan:

[0265] Given the patient's history of chronic sinusitis, observation for any potential mucosal thickening, sinus obstruction, or inflammation would be helpful.

[0266] Anatomical structures to be examined:

[0267] 1. Upper and lower airways

[0268] 2. Lung parenchyma

[0269] 3. Mediastinum

[0270] 4. Paranasal sinuses

[0271] Automatic processing and image-based AI features for radiologists:

[0272] 1. Multiplanar reconstruction (MPR) - enables better visualization of various lung abnormalities, airway involvement, and sinus anatomy.

[0273] 2. Volume Rendering (VR) – Visualizes lung and airway structures in 3D format, allowing for easier assessment of their relationship to other chest structures.

[0274] 3. Minimum Intensity Projection (MinIP) - assesses airway patency and detects any mucosal thickening or obstruction in the paranasal sinuses.

[0275] Possible AI features:

[0276] 1. Automatically identify potential lung abnormalities such as lung nodules or infiltrates, as well as any airway abnormalities.

[0277] 2. Automatically measure bronchial wall thickness and bronchiectasis severity index (BSI) to determine whether bronchiectasis or chronic airway disease is present.

[0278] After being converted (particularly automatically) into a machine-readable format, the exemplary workflow (and / or workflow plan) may be read as follows:

[0279] {"CT Scans":[({"CT Scans":[)

[0280] {"Scan_Type":"Chest High-resolution CT(HRCT)scan","Purpose":"Evaluatelung parenchyma and airways for underlying diseases"},

[0281] { "Scan_Type":"Paranasal Sinuses CT scan", "Purpose":"Assess chronic sinusitis and potential inflammation"}],

[0282] "Anatomical_Structures_To_Examine":("Anatomical structures to examine":)

[0283] [ "Upper and lower airways", "Lung parenchyma", "Mediastinum", "Paranasal sinuses"],

[0284] "Auto-Processings":[

[0285] { "Type":"Multiplanar Reconstructions(MPR)", "Usage":"Better visualization of lung abnormalities and sinus anatomy"},

[0286] { "Type":"Volume Rendering(VR)", "Usage":"3D visualization of lungand airway structures"},

[0287] { "Type":"Minimum Intensity Projection(MinIP)", "Usage":"Assessairway patency and detect sinus obstructions"}],({"Type":"Minimum Intensity Projection(MinIP)","Usage":"Assessairway patency and detect sinus obstructions"}],)"AI_Features":[

[0288] { "Task":"Automatic identification of lung and density abnormalities","AI Job":"LungCAD,LungDensity"},({"Task":"Automatic identification of lung and density abnormalities","AI Job":"LungCAD,LungDensity"},)

[0289] { "Task":"Automated measurement of bronchial wall thickness","AIJob":"Airways"},]}({"Task":"Automated measurement of bronchial wall thickness","AIJob":"Airways"},]})

[0290] At the action level (e.g. Figure 3 10 ; S112 ), the current optimal workflow (and / or plan) for individual patient surgery (particularly according to each patient's requested medical imaging) can be executed in a flexible process.

[0291] All actions that follow the inference level plan may occasionally fail. For example, a radiology staff member may be pulled into an emergency, a patient may not show up, and / or a medical scanner (and / or equipment) may malfunction.

[0292] All faults and / or events can change the message level (especially in Figure 3 Based on the change in the information level (particularly the change in the availability of the medical scanner 304), the reasoning level (e.g., Figure 3 ) and action levels (e.g., Figure 3The reasoning level may be dynamically adjusted (as shown in the reference numerals S110; S112 in FIG. 1 ). Changes in the reasoning level may include, for example, changes in scanning and / or protocol recommendations, scheduling and / or use of specific assets (particularly different medical scanners and / or particularly detachable equipment). Changes in the reasoning level ensure that standardized operation, patient-specific adaptation and / or system (and / or medical scanners within the fleet) availability are always the best combination.

[0293] Figure 5 The parameterization of a (particularly specific) medical scanner is exemplarily shown in . For example, for a CT scanner, the routine for head (and in particular brain) imaging may allow for 25 slices. Figure 5 Any of the scan parameters illustrated in the example may be combined with only a subset of the additional scan parameters shown, and / or with Figure 5 Additional scanning parameter combinations not shown in FIG.

[0294] Since medical scanners traditionally cannot process (and / or do not accept) scan parameters in parameterization for documentation and / or human use (e.g. Figure 5 ), the scan parameters need to be converted (also: standardized and / or normalized) into a format that the medical scanner can process (and / or import). The conversion of the scan parameters can be achieved by a converter. The converter can be implemented by (e.g., another) generative AI (e.g., including a neural network, the neural network including a transformer architecture, and / or in particular an LLM), and the generative AI can be provided with a grammatical structure in particular according to the medical scanner selected for medical imaging. The conversion of the scan parameter format can help integrate legacy medical scanners (and / or devices).

[0295] In a preferred embodiment, the (particularly radiology operation) system can display (also: for example shown according to method step S109) the planned and actual workflow for each patient and an integrated overview of the use of the entire radiology assets (e.g. including medical scanners and / or particularly detachable equipment). Radiologists and technicians can access current and past individual workflows and a comprehensive performance overview.

[0296] In another preferred embodiment, the output of the (e.g., radiology operation) system and human interaction (e.g., using a UI, particularly a GUI) are performed (and / or completed) via a mobile device (e.g., a mobile phone, tablet, smart watch, smart glasses and / or smart contact lenses).

[0297] In a further preferred embodiment, the output of the (eg radiology operation) system and human interaction are performed (and / or accomplished) via an optical and / or touch unit at the medical scanner (eg using a UI, in particular a GUI, at the medical scanner).

[0298] Figure 6 Schematically represents closed-loop control of radiology equipment (e.g., medical scanner 304) in a radiology equipment fleet (particularly, including medical scanners) by providing a backflow of information from the medical scanner 304 to the computing device 200. The computing device 200 is configured to provide control signals via a control signal providing interface. The control signals indicate a workflow for performing medical imaging. The back channel can be used for transmission of signals indicating the actual status of the corresponding radiology equipment (e.g., medical scanner 304). For example, the back channel can be used to indicate occupancy of the corresponding medical scanner 304 due to an emergency, or a signal indicating a failure of the medical scanner 304 or other current conditions that are different from the scheduled or planned conditions.

[0299] The back channel signal may be processed during the inference phase.

[0300] The back channel signal may include second input data indicating the availability (in this case modified) of the radiology equipment (eg, the medical scanner 304).The back channel signal may be received on the scanner availability receiving interface 204-B.

[0301] The back channel signal may include technical specifications of the medical scanner 304. The back channel signal may be received on the technical specifications interface 202-A.

[0302] Back channel signals are transmitted from the medical scanner 304 to the computing device 200 and / or control signals are transmitted from the computing device 200 (as a result of processing) to the medical scanner 304 to control the medical scanner 304 .

[0303] Any of the embodiments may be combined with each other.

[0304] The present technology (e.g., including method 100 and / or computing device 200) can be instantiated in a cloud infrastructure. Thus, a widely distributed fleet of assets (e.g., medical scanners and / or particularly detachable equipment) can be addressed, managed, scheduled and / or controlled.

[0305] The inventive technique (e.g., comprising method 100, computing device 200 and / or in particular radiology operating system) combines multiple information sources (e.g., comprising first input data relating to a request for medical imaging, second input data indicating the availability of one or more medical scanners within a fleet, technical specifications of any medical scanner within the fleet and / or generally applicable instructions including, for example, guidelines and / or SOPs) and creates an optimal workflow (also: workflow plan). In addition, the inventive technique allows for the execution of dynamic, patient-adjusted and / or personalized workflows and parameterizations using available assets (e.g., medical scanners within a fleet and / or in particular detachable equipment). The inventive technique avoids the need for complex individual and local configuration of assets. The configuration itself can be based on, for example, human-readable documents, SOPs and / or guidelines, and can adapt to current practices and / or behaviors beyond existing existing data. Behavior can include selection of scan parameters and / or settings for the requested medical imaging.

[0306] The present technology may be directly visible and / or detectable in how to configure and indicate implementations, including, for example, data connections.

[0307] In any case not explicitly described, the various embodiments described with respect to the drawings or their various aspects and features may be combined or interchanged with each other without limiting or expanding the scope of the described invention, as long as such combination or interchange is meaningful and within the meaning of the invention. The advantages described with respect to a specific embodiment of the invention or with respect to a specific drawing are also advantages of other embodiments of the invention where applicable.

[0308] Citations

[0309] [1]Jacob Devlin and Ming-Wei Chang and Kenton Lee and KristinaToutanova:BERT:Pre-training of Deep Bidirectional Transformers for LanguageUnderstanding.arXiv:1810.04805v2[cs.CL].

[0310] [2]Ashish Vaswani and Noam Shazeer and Niki Parmar and JakobUszkoreit and Llion Jones and Aidan N.Gomez and Lukasz Kaiser and IlliaPolosukhin:Attention Is All You Need.arXiv:1706.03762v7[cs.CL](2023).

[0312] [3]The Transformer model family.

[0313] https: / / huggingface.co / docs / transformers / model_summary

[0314] [4]Lewis,M.,Liu,Y.,Goyal,N.,Ghazvininejad,M.,Mohamed,A.,Levy,O.,Stoyanov,V.and Zettlemoyer,L.,2019.Bart:Denoising sequence-to-sequence pre-training for natural language generation,translation,and comprehension.arXiv:1910.13461v1[cs.CL].

[0316] [5]Rogers,A.,Kovaleva,O.and Rumshisky,A.,2020:Primer in BERTology:What we know about how BERT works.arXiv:2002.12327v3[cs.CL].

[0318] [6]Drew A.Hudson and C.Lawrence Zitnick:Generative AdversarialTransformers.arXiv:2103.01209v4[cs.CL](2022).

[0319] [7]Yutao Sun and Li Dong and Shaohan Huang and Shuming Ma and YuqingXia and Jilong Xue and Jianyong Wang and Furu Wei:Retentive Network:ASuccessor to Transformer for Large Language Models.arXiv:2307.08621v4[cs.CL]

[0320] [8]Shantanu Chandra:Retentive Networks(RetNet)Explained:The much-awaited Transformers-killere is here.AI FUSION LABS.4 August2023.

[0321] https: / / medium.com / ai-fusion-labs / retentive-networks-retnet-explained-th e-much-awaited-transformers-killer-is-here-6c17e3e8add8

[0322] [9]Miotto,R.,Li,L.,Kidd,B.A.,et al.(2016).Deep patient:Anunsupervised representation to predict the future of patients from electronichealth records,Scientific Reports,DOI:10.1038 / srep26094;https: / / www.nature.com / articles / srep26094

Claims

1. A computer-implemented method (100) for providing a control signal indicating a workflow for performing medical imaging with a medical scanner in a fleet of medical scanners, the method (100) comprising the following steps: - receiving (S104-A) first input data relating to a request for medical imaging of a patient; - receiving (S104-B) second input data indicating the availability of at least one medical scanner within the medical scanner fleet; - processing (S108) the received (S104-A) first input data and the received (S104-B) second input data, wherein the processing (S108) comprises selecting an available medical scanner from the fleet, and wherein the processing (S108) is performed with the aid of generative artificial intelligence for creating a workflow for the medical imaging according to the request; and - providing (S110) a control signal indicative of said workflow for performing said medical imaging by means of said selected medical scanner.

2. The method (100) according to claim 1, further comprising at least one of the following method steps: - receiving (S102-A) technical specifications of the medical scanner, and preferably receiving (S102-A) technical specifications for any medical scanner within the medical scanner fleet, optionally wherein, The technical specifications include the presence and / or specification of, in particular, detachable equipment; and - Receiving (S102-B) generally applicable instructions related to medical imaging of said medical scanner fleet, in particular comprising one or more medical guidelines and / or one or more standard operating procedures SOP.

3. The method (100) according to any one of the preceding claims, wherein: The received (S104-A) first input data related to the request for medical imaging of the patient includes at least one of the following: - a reason for requesting said medical imaging, in particular including said medical imaging of a suspected medical condition and / or a suspected lesion; - one or more anatomical structures and / or one or more organs to be captured by said medical imaging; - Information about the general health status of the patient, in particular in view of the presence of a bypass, stent, pacemaker and / or any in particular further implants; as well as - Existing or existing medical images of said patient, in particular medical images relevant to said request.

4. The method (100) according to any one of the preceding claims, wherein: By means of a user interface UI, in particular a graphical user interface GUI, at least a part of the first input data related to the request for medical imaging of the patient is received (S104-A).

5. The method (100) according to any one of the preceding claims 2 to 4, wherein: Selecting the available medical scanner from the fleet is based on the technical specifications of the medical scanner and / or wherein processing (S108) includes determining a set of scanning parameters for the workflow.

6. The method (100) according to any one of the preceding claims, wherein: The generative artificial intelligence includes at least one of the following: -Neural networks including attention mechanisms; -Neural networks including transformer architectures; -Generative Adversarial Networks, in particular Generative Adversarial Transformers; -Retain network; as well as -Large Language Model LLM.

7. The method (100) according to any one of the preceding claims, further comprising the following method steps: - Preprocessing at least one of the following (S106): o first input data received (S104-A) relating to said request for medical imaging of said patient; o received (S104-B) second input data indicating the availability of at least one medical scanner within the fleet; o the received (S102-A) technical specifications of the medical scanner; and / or o the received (S102-B) generally applicable instructions.

8. The method (100) according to any one of the preceding claims, further comprising the following method steps: - Acquiring ( S112 ) medical imaging data by the selected medical scanner based on the provided ( S110 ) control signal.

9. The method (100) according to any one of the preceding claims, further comprising the following method steps: - Post-processing ( S114 ) the acquired ( S112 ) medical imaging data.

10. The method (100) according to any one of the preceding claims, further comprising at least one of the following method steps: - using a user interface UI, in particular a graphical user interface GUI, to display (S109) the created workflow; and / or - Displaying (S115) the acquired (S112) and / or post-processed (S114) medical image data, preferably using a UI, in particular said GUI.

11. The method (100) according to any one of the preceding claims, wherein: The medical scanner fleet includes at least two medical scanners associated with different medical imaging modalities.

12. Use of the method (100) according to any one of the preceding claims for scheduling a plurality of patients across the fleet of medical scanners.

13. A computing device (200) for providing a control signal indicating a workflow for performing medical imaging with a medical scanner in a medical scanner fleet, the computing device (200) comprising: - a first input data receiving interface (204-A) configured to receive first input data related to a request for medical imaging of a patient; - a second input data receiving interface (204-B) configured to receive second input data indicating the availability of at least one medical scanner within the medical scanner fleet; a processing unit (208) configured to process the received first input data and the received second input data, wherein the processing comprises selecting an available medical scanner from the fleet, and wherein the processing is performed with the aid of generative artificial intelligence for creating a workflow for the medical imaging according to the request; and - a control signal providing interface (210) configured for providing a control signal indicative of a workflow for performing said medical imaging by means of the selected medical scanner.

14. The computing device (200) according to the immediately preceding claim, further configured to perform any of the steps of any of the method claims 2 to 13 and / or to include any of the features of any of the method claims 2 to 13.

15. A system for providing a control signal indicating a workflow for performing medical imaging with a medical scanner in a fleet of medical scanners, the system comprising: - A computing device (200) according to claim 13 or 14; as well as - A medical scanner cluster, wherein each medical scanner (304) within the cluster comprises a control signal receiving interface, wherein the control signal receiving interface is configured to receive a control signal from the control signal providing interface (210) of the computing device (200).

16. A computer program product, comprising program elements which, when loaded into a memory of a computing device (200), cause the computing device (200) to perform the steps of a method for providing a control signal according to one of the preceding method claims, wherein the control signal indicates a workflow for performing medical imaging with the aid of a medical scanner in a medical scanner fleet.

17. A computer-readable medium having program elements stored thereon, the program elements being capable of being read and executed by a computing device (200) so that when the program elements are executed by the computing device (200), the steps of a method for providing a control signal according to one of the preceding method claims are performed, the control signal indicating a workflow for performing medical imaging with the aid of a medical scanner in a medical scanner fleet.

Images