Computer-implemented method for checking the indication of at least one radiological examination

Abstract

The invention relates to a computer-implemented method for checking the indication of at least one radiological examination with the following method steps: Receiving a text-based request for a radiological examination; Querying patient data (5) via a central information system (1); Checking the text-based request depending on the patient data (5) and selecting a suitable radiological examination depending on the verification of the text-based request using artificial intelligence; Transmitting a text-based request for the selected radiological examination to a radiological department (6) comprising several radiological devices (7); Checking the feasibility of the selected radiological examination on at least one of the radiological devices using artificial intelligence; Scheduling the selected radiological examination based on the text-based request and the feasibility of the selected radiological examination on at least one of the radiological devices using artificial intelligence. This provides a computer-implemented method that allows the indication for radiological examinations to be reviewed taking the overall situation into account, and the radiological examination to be scheduled reproducibly and with fewer errors, taking these aspects into account.

Description

[0001] The invention relates to a computer-implemented method for checking the indication of at least one radiological examination of a patient.

[0002] When a patient is referred to a radiology department for a radiological examination, the Radiological Request Form (RRF) is a central element of the radiological workflow, facilitating communication between referring physicians and radiologists. It enables radiologists to select the appropriate radiological examination and scan protocol. When selecting a radiological examination, the radiation dose, limitations of the various imaging modalities, and possible contrast phases are considered. Carefully considering these factors is time-consuming and error-prone. Furthermore, the consideration of these factors is subjective and rarely reproducible between different physicians.In order to use the resources of the radiology department more efficiently, RRF pre-sorting is required with regard to the appropriate radiological examination modality, body region and the need for contrast agents.

[0003] Conventional clinical decision support systems (CDS) can enable faster imaging. However, these systems are limited by the predefined, branched decision tree structures they are based on. In other words, the information used in decision-making is limited, preventing a decision from being made with the overall situation in mind.

[0004] Based on this, the object of the invention is to provide a computer-implemented method with which the indication of radiological examinations can be checked taking into account the overall situation and the radiological examination can be planned in a reproducible and less error-prone manner under these aspects.

[0005] This problem is solved by the subject matter of patent claim 1. Preferred developments can be found in the subclaims.

[0006] According to the invention, a computer-implemented method for checking the indication of at least one radiological examination of a patient is provided, comprising the following method steps: Receiving a text-based request for a radiological examination comprising a clinical question; Querying patient data via a central information system; Reviewing the text-based request based on the patient data and selecting a suitable radiological examination based on the review of the text-based request using artificial intelligence; Transmitting a text-based request for the selected radiological examination to a radiological department comprising multiple radiological devices; Checking the feasibility of the selected radiological examination on at least one of the radiological devices using artificial intelligence; Scheduling the selected radiological examination depending on the text-based request and the feasibility of the selected radiological examination on at least one of the radiological devices using artificial intelligence.

[0007] In a first step, the requester (e.g., a general practitioner, a non-radiology department within the hospital) submits a text-based request to the radiology department. This request is structured specifically into a medical history, a clinical question, and a test that the referring physician considers appropriate, and is submitted in text form.

[0008] In the approach presented here, an indication check is performed based on this request using artificial intelligence, thus determining the modality, body region, contrast agent application, and contrast agent phases. Furthermore, the necessity and its compatibility with the patient data are validated. If the artificial intelligence deems it necessary, the referring physician's text-based request is modified. In this process, medical history and relevant patient information, such as laboratory values, contraindications, and / or previous examinations, are obtained by accessing the central information system, so that a request can be generated based on the medical history and clinical question. A suitable radiological examination is then selected. This specifically means that a suitable radiological examination can be selected from many types of radiological examinations, such as CT, MRI, PET-CT, X-ray, etc.Based on the clinical question and patient data, an appropriate type of radiological examination is selected. The correct request is then forwarded to the radiology department.

[0009] It has been shown that such an AI-based pre-check of the text-based requirement significantly increases the likelihood that the subsequent human validation of the indication check will be carried out correctly. Even an incorrect pre-selection leads to a significant chain of errors throughout the entire radiological examination process. Pre-checking thus significantly increases the quality and reliability of the indication check and thus patient care. The pre-check is thus performed independently of subjective human intervention, whose decisions are subject to strong inter-individual fluctuations in their accuracy and are not always reproducible.

[0010] In a further step, the availability of the selected examination in the radiology department's equipment pool is checked using artificial intelligence. In this regard, a continuous exchange of information is preferably carried out so that the current availability of radiological equipment can be incorporated into the decision.

[0011] A key aspect of the invention is that all information relevant for indication assessment and examination planning is retrieved in real time and integrated into the decision-making process in its entirety at a central location. The subjective "human" component, or the radiologist, is isolated from this decision-making process, enabling a much more comprehensive overall assessment and achieving reliable and reproducible results of higher quality.

[0012] In this context, "artificial intelligence" refers specifically to a text-based generic network or a neural network model for machine learning that is trained to understand, generate, and respond to human-like text based on given inputs. It uses transformer architectures and is primarily used in areas such as text generation, text comprehension, and natural language processing. Preferably, the artificial intelligence is a large language model. A large language model, or LLM for short, is a language model characterized by its ability to generate non-specific texts. It is a computational linguistic probabilistic model that has learned statistical word and sentence sequence relationships from a large number of text documents through a computationally intensive training process.

[0013] According to a preferred development of the invention, the following further method steps are provided: Checking the scheduled selected radiological examination for contraindications depending on the patient data or predetermined conditions using artificial intelligence; If contraindications exist, issue a warning message

[0014] This provides an alarm function that automatically triggers an alert in the event of contraindications, such as a pacemaker during a scheduled MRI examination, or incompatibility with the patient data, such as advanced renal insufficiency during a planned contrast agent administration. This provides a review mechanism that allows the selected radiological examination to be individually tailored to the patient, including its feasibility.

[0015] According to a preferred development of the invention, the further method step is provided: Reporting the appointment of the selected radiological examination to an external referrer who submits the text-based request.

[0016] The selected, correct radiological examination and the corresponding appointment slot are reported back to the referring physician. This feedback is sent, for example, via email.

[0017] According to a preferred development of the invention, the further method step is provided: Determining device parameters of the at least one radiological device for the selected radiological examination depending on the patient data and the text-based request.

[0018] Depending on the patient data and the requirement, i.e., a clinical question, device parameters are determined using artificial intelligence. This allows device parameters to be set, such as the contrast agent dose, the correct tube settings for the X-ray tube and / or CT, the setting of the X-ray tube for hard beam technology (>100 kilovolts) for lung tissue examinations vs. soft beam technology (<100 kilovolts) for skeletal examinations, the reduction of the tube current in mA for a low-dose CT examination of the thorax, for example, in lung cancer screening, the correct coil selection for a specific body region for an MRI, and / or the determination of the contrast agent dose required on the day of the examination.In addition, ordering of necessary resources, such as contrast agents, can be automated, and / or the required energy capacity can be calculated and taken into account for downstream processes in the interest of sustainability. It has been shown that by determining device parameters using artificial intelligence, the tube current can be reduced from 80 mA to 40 mA for a screening examination specified in the patient's medical history.

[0019] According to a preferred development of the invention, the following further method steps are provided: Determining the availability of multiple image processing systems; Selecting a suitable image processing system depending on the selected radiological examination and / or the patient data and / or the text-based request using artificial intelligence.

[0020] Based on the previously requested and then performed examination, such as “X-ray of the skeleton, upper ankle joint”, the medical history, for example “twisting trauma”, and the clinical question, for example “fracture?”, artificial intelligence is used to select a specific, follow-up AI application available to the radiology department for further processing of the individually generated image data, such as an AI application for automated fracture detection.

[0021] According to a preferred development of the invention, the further method step is provided: Determining a suitable evaluation time of generated image data depending on a performed radiological examination and / or the patient data and / or the text-based request using artificial intelligence.

[0022] Based on the relevant parameters explained above, and in particular on previous examinations, the timing of the AI ​​analysis of the radiological examination is prioritized. This optimizes server utilization, as, for example, AI-supported analysis of radiological examinations with low priority and / or high server load are shifted to off-peak times. This allows existing capacity to be used more efficiently.

[0023] According to a preferred development of the invention, the further method step is provided: Checking a finding obtained by specialist personnel and / or using artificial intelligence for inconsistencies depending on the selected radiological examination and / or the patient data and / or the text-based request using artificial intelligence.

[0024] In particular, it is intended that specialist personnel, such as a radiologist with specialist qualifications, perform the final validation of a report. These generated reports are then subjected to an AI-based review. This includes, in particular, a semantic review, validation of the answers to all clinical questions, a consistency check of the report, and autocorrection. If inconsistencies are detected, the final release of the report is blocked.

[0025] It has been shown that this can lead to significant performance improvements, particularly with regard to error rate and coherence. Factors such as shift work, especially night shifts, or increased workload lead to significant differences in the content of the findings between different authors, or even between different evaluation periods for the same author. Due to these factors, in extreme cases, a result may not be reproducible, even by an identical author. Eliminating such subjective components from the evaluation therefore increases the quality of the evaluation or the report. This ultimately leads to increased patient safety and improved patient care.

[0026] Preferably, based on the findings in text form, for example, the CT findings describing an abnormal esophagus, supplementary imaging in accordance with current guidelines, imaging follow-up, further clinical diagnostics, and / or a suggestion for potential patient inclusion in a clinical study are determined and output using artificial intelligence. According to a preferred development of the invention, the following method step is provided: Carrying out coding and / or billing of the radiological examination depending on the radiological examination and / or the patient data and / or the text-based request and / or the collected findings using artificial intelligence.

[0027] In particular, automated pre-structuring or final coding, such as automatic ICD-10 coding, and billing of the service provided with the radiological examination and the materials used for this purpose, such as MRI contrast media, are enabled on the basis of the available text data in the anamnesis, clinical question and findings.

[0028] Preferably, artificial intelligence is used to create summaries and generate programming code to create overview graphics of the disease progression to date from the previous findings data.

[0029] According to the invention, a system for data processing comprising a central interface which establishes a connection between a central information system, an AI module and external interfaces is further provided, which system is designed to carry out the method according to one of the previous steps.

[0030] According to the invention, a computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method described above is further provided.

[0031] The invention is explained in more detail below using a preferred embodiment with reference to the drawings.

[0032] The drawing shows Fig. 1 A computer-implemented method according to a preferred embodiment of the invention in a flowchart.

[0033] In Fig.1 schematically shows a flow diagram illustrating the sequence of a computer-implemented method according to a preferred embodiment of the invention. At the center of the flow diagram is the central information system 1, which is shown here in the form of a hospital information system (HIS). This central information system 1 comprises the integrated AI module 2A, shown in the form of a text-based large language module (LLM). The requester 3 can be an external referring physician, such as a general practitioner, or a non-radiology department in a hospital. The patients being treated by the requester form the patient pool 4. Patient data 5 for each patient is stored in the central information system 1. The requester 3 requests a radiological examination for a specific patient.The AI ​​module then accesses the request and the patient data and selects a radiological examination suitable for the explicit request or performs a correction of the requested examination (indication check 8). This selection is transmitted to the radiology department 6. The AI ​​module 2B of the radiology department 6 accesses the radiology department's equipment park 7 and the respective equipment settings and can record which equipment is available and when. The AI ​​module 2B schedules the radiological examination 9 depending on the respective examination specified according to the indication check 8 and the patient data. It then provides feedback 10 to the requester 3.

[0034] During the radiological examination, image data 11 is generated. This image data 11 is evaluated using AI. For this purpose, the AI ​​module 2C selects the appropriate image evaluation software from a pool of available software products. Depending on the clinical question, a different evaluation may be required. This AI evaluation is used to create a finding 12. The AI ​​module 2D checks the finding 12 for its coherence, taking into account the originally stated requirement or clinical question, and can provide the requester 3 with direct feedback, for example, in the form of a further recommendation.

[0035] The report 12 is validated by a specialist 13 and has previously undergone an AI-supported semantic correction and / or consistency check. The AI ​​module 2E accesses the reviewed and validated report and can generate a coding and / or billing 14 for the radiological service.

[0036] In addition, the AI ​​module 2F - since it has access to all relevant data - can create a summary 15 of the disease progression to date.

[0037] The AI ​​modules 2A-2F are each to be understood as sub-modules of a unified module. A higher-level central interface (not shown) establishes a connection between the requester 3, the AI ​​modules 2A-2F, and the radiology department 6 and its equipment 7. List of reference symbols 1 central information system 2A-F AI module 3 requesters 4 Patient pool 5 Patient data 6 radiological department 7 Equipment park 8 Indication testing 9 Scheduling 10 Feedback 11 Image data 12 Findings 13 Validation 14 Billing 15 Summary

Claims

[1] Computer-implemented procedure for assessing the indication for at least one radiological examination of a patient, comprising the following procedural steps: Receiving a text-based request for a radiological examination encompassing a clinical question; Querying patient data (5) via a central information system (1); Verification of the text-based request based on patient data (5) and selection of a suitable radiological examination based on the verification of the text-based request using artificial intelligence; Transmitting a text-based request for the selected radiological examination to a radiology department (6) comprising several radiological devices (7); Verification of the feasibility of the selected radiological examination on at least one of the radiological devices using artificial intelligence; Scheduling the selected radiological examination depending on the text-based request and the feasibility of the selected radiological examination on at least one of the radiological devices using artificial intelligence. [2] Computer-implemented method according to claim 1, comprising the following further method steps: Checking the scheduled selected radiological examination for contraindications depending on the patient data (5) or on predetermined prerequisites using artificial intelligence; If contraindications exist, issue a warning message. [3] Computer-implemented method according to claim 1 or 2, comprising the following further method steps: Reporting the date of the selected radiological examination to an external requester who submits the text-based request (3). [4] Computer-implemented method according to one of the preceding claims, comprising the following further method step: Determining device parameters of at least one radiological device for the selected radiological examination depending on the patient data (5) and the text-based request. [5] Computer-implemented method according to one of the preceding claims, comprising the following further method steps: Determining the availability of multiple image processing systems; Selecting a suitable image processing system depending on the selected radiological examination and / or the patient data and / or the text-based requirement using artificial intelligence. [6] Computer-implemented method according to one of the preceding claims, comprising the following further method step: Determining a suitable evaluation time for generated image data depending on a radiological examination performed and / or the patient data and / or the text-based request using artificial intelligence. [7] Computer-implemented method according to one of the preceding claims, comprising the following further method step: Checking a finding obtained by specialist personnel and / or by means of artificial intelligence (12) for inconsistencies depending on the selected radiological examination and / or the patient data (5) and / or the text-based request using artificial intelligence. [8] Computer-implemented method according to one of the preceding claims, comprising the following further method step: Performing coding and / or billing (14) of the radiological examination depending on the radiological examination and / or the patient data (5) and / or the text-based request and / or the collected findings (12) using artificial intelligence. [9] Data processing system comprising a central interface that establishes a connection between a central information system (1), an AI module (2A-2F) and external interfaces designed to execute the procedure according to one of the previous steps. [10] Computer program product comprising instructions which, when the program is executed by a computer, cause it to execute the method according to any one of claims 1 to 8.

Images