Smart scan recommendations for MRI

Abstract

The present invention relates to intelligent scan recommendation for magnetic resonance imaging. A scan recommendation method is disclosed, wherein evaluation data including classification data (CIA, CMA) is received, wherein the classification data (CIA, CMA) classifies abnormalities (IA, MA) detected in input data (ID, IM, IHRI, UII) regarding the medical condition of a patient. Based on the evaluation data, an automatic determination of the next recommended MR protocol (NRP) to be performed based on the MR scan of the patient is performed. In addition, a personal scan preparation method is disclosed. In addition, a medical imaging method is described. A scan workflow execution method is also described. In order to perform the above method, a scan recommendation system (45), a personal scan preparation system (40), a medical imaging system (50) and a scan workflow execution system (60) are also described.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to U.S. Provisional Application No. 63 / 013,825, filed on April 22, 2020, the entire disclosure of which is incorporated herein by reference in its entirety. Technical Field

[0003] The present disclosure relates to techniques for developing workflows for magnetic resonance imaging (MRI), and more particularly to techniques for recommending MR protocols using various data sources that can be part of an automated process. Background Art

[0004] Magnetic resonance imaging (MRI) is a medical imaging technique that provides diagnostically relevant information about biological tissues. Its diagnostic relevance derives from the information power and versatility of this imaging modality, which provides differentiated and precise structural information in a non-invasive manner. Magnetic resonance imaging is based on the controlled manipulation of nuclear spins inside the patient's body and the subsequent detection of the nuclear spin responses. Spatially correlated encoding of the spin responses allows reconstruction of the patient's structural and functional composition for diagnostic reading by radiologists.

[0005] Conventionally, to generate MR images for diagnosis, an MRI technician performing an MRI scan selects a sequential arrangement of MR protocols from a library of MRI scanner imaging protocols, which may be based on, for example, the relevant body part, the suspected disease (e.g., clinical indication by a referring physician or radiologist), patient-specific requirements (e.g., implants, motion-related non-compliance), and the preferences of the clinical institution and radiologist who will later perform the reading of the MR images for diagnosis. In addition, during the scanning process, the technician will typically perform a check to ensure that the image quality of the recorded scan allows for a diagnosis. The technician can then make protocol modifications, for example, to adapt the field of view to the patient's body and others, and if necessary, a rescan will be performed.

[0006] This classic MRI scanning workflow requires MRI technicians to undergo intensive modality-specific training and acquire many years of scanning experience. Today, the number of MRI scans performed is increasing, while there is a shortage of qualified personnel, which leads to high operating costs. Even for well-trained technicians, the large number of manual steps required to optimally perform MRI scans and the increased time pressure lead to a significant risk of operating errors. Even in well-known hospitals, there may be times when poorly trained technicians have to operate the scanner, for example at night or on weekends.

[0007] In US 9,928,589 B2, a device and method for supporting the acquisition of multi-parameter images are described. The device includes: a disease selector configured to select a suspected disease of a patient based on patient information; and an image selector configured to determine a set of imaging conditions for a multi-parameter magnetic resonance image corresponding to the suspected disease based on a multi-parameter magnetic resonance imaging model. However, it is usually difficult to determine the patient's disease before recording a medical image.

[0008] In US 2018 / 0 101 644 A1, a method for providing confidence information about a diagnosis of a diagnostic system based on medical image data is described. The diagnostic system is designed as an algorithm that uses the medical image data and further patient-specific data as its input. The measure of confidence is determined by executing a determination algorithm that has the following as its input parameters: characteristic values ​​for the diagnostic system, such as the sensitivity or specificity of the diagnostic system; and / or image content from the medical image data. In addition, it can be proposed to further acquire additional image data based on the confidence information. The additional medical image data can be acquired such that the data exhibits a changed image contrast relative to the already existing medical image data. Summary of the invention

[0009] Likewise, as of today, most MRI scans are performed manually, requiring the technician to select the desired protocol from a library of protocols. In order to automate the MR imaging workflow, applications have been designed that contain sequentially arranged protocols that are specialized for imaging of specific anatomical regions and / or suspected diseases. Specifically, the steps of image slice planning have been automated through pattern matching algorithms. However, as of today, there is no software that can enable an untrained technician to perform an entire MRI scan with a high success rate while taking into account the patient's specific needs or findings.

[0010] It is therefore an object of the present invention to find a solution for preparing and performing an MRI scan without the need for qualified operators.

[0011] The above problems are solved by: a scan recommendation method according to the present invention, a personal scan preparation method according to the present invention, a medical imaging method according to the present invention, a scan workflow execution method according to the present invention, a scan recommendation system according to the present invention, a personal scan preparation system according to the present invention, a medical imaging system according to the present invention, and a scan workflow execution system according to the present invention.

[0012] The scan recommendation method according to the present invention comprises a step of receiving evaluation data including classification data. The classification data classifies abnormalities detected based on input data about the medical condition of the patient. Preferably, the abnormalities are detected in medical image data of the patient. The classification comprises the type of identification of the detected abnormality, and probability values ​​for different possible types of abnormalities are assigned to the detected abnormalities. For example, the abnormalities may be abnormalities detected in input data such as medical image data, or based on medical data about different types of patients, they may include information about the abnormalities.

[0013] The abnormality detected in the image data includes a detected image abnormality or a unique image feature, such as a high signal area, a low signal area, abnormal tissue, a midline shift, a vascular stenosis, a plaque, demyelination or a mass effect. As further described below, the abnormality can also be a detected image metric abnormality or a distinguishing feature. Then, based on the evaluation data, the next recommended MR protocol is automatically determined, which is intended to be performed together with the MR scan of the patient. As described below, the evaluation is not limited to the evaluation of the image data. Therefore, multiple different types of input data about the patient's condition can be analyzed to determine the evaluation data. In addition, multiple data groups of a single special type of input data about the patient's condition can be evaluated. The scan recommendation can be based on receiving multiple evaluation data of different types. The detection of abnormalities can be implemented based on a machine learning method (e.g., based on a neural network such as a convolutional neural network), which takes input data such as image data as input and provides the probability of the input data being normal and abnormal as output. The classification of the detected abnormalities can also be implemented using a machine learning method (e.g., a neural network-based technique or more specifically using a convolutional neural network). If the probability value of the detected abnormality for the received input data such as image data is higher than a certain customizable threshold ξ d , then as described above, input data such as image data assigned to the detected abnormality is used as input for the subsequent classification step. Scan recommendations can also be implemented based on machine learning-based techniques, in particular neural network techniques as described below. In the simplest variant, the evaluation data simply includes classification data. However, the evaluation data can also additionally include different types of data. The method according to the present invention allows trained or untrained technicians to perform a highly personalized MRI scanning workflow for each patient, thereby generating the optimal MRI image set required for diagnosing the patient's condition. The method can provide a basis for a fully autonomously optimized MR imaging workflow, thereby also allowing it to be manually or electronically customized by the operator as needed. The MR scanner operator can choose to run the MR scanning workflow in fully automatic, fully manual or hybrid mode.

[0014] The personal scan preparation method according to the present invention comprises the following steps: receiving input data about the medical condition of the patient, such as medical image data. Determining the evaluation data based on the input data by detecting anomalies in the input data, determining classification data based on the detected anomalies as evaluation data, and performing the scan recommendation method according to the present invention. Advantageously, the classification of the anomalies can be used to adapt the MR protocol to special kinds of anomalies. For example, a special contrast is selected for the recommended MR protocol. In addition, the personal scan preparation method shares the advantages of the scan recommendation method.

[0015] The medical imaging method according to the present invention comprises the following steps: determining a next recommended protocol based on the scan recommendation method according to the present invention, and sending the next recommended protocol to the MR scanning system, and the MR scanning system performing MR scanning based on the received protocol. The medical imaging method shares the advantages of the scan recommendation method according to the present invention.

[0016] The scanning workflow execution method according to the present invention includes the steps of executing the medical imaging method according to the present invention. After the MR scan of the medical imaging method, an abnormality is detected in the reconstructed MR image data, and classification data is determined based on the detected abnormality. Then, the above three steps are iteratively performed in consideration of the determined classification data until the termination criteria are obtained. In other words, the next recommended protocol is first determined based on the scan recommendation method of the present invention or the personal scan preparation method according to the present invention. Then, the next recommended protocol is sent to the MR scanning system, and the MR scanning is performed by the MR scanning system based on the recommended protocol. In addition, iterative protocol determination and measurement are performed, which includes the following steps, controlling the MR scan, evaluating the reconstructed MR image data, and iteratively improving the input data for the next recommended protocol based on the classification of the abnormality of the reconstructed MR image data. For example, the operator can first record a standard MR protocol group, and adopt a scan recommendation method based on the recorded first image for additional MR scanning, which can be additionally indicated for the current patient based on imaging information. Advantageously, a complete automatic workflow for optimized MR scanning of a patient is achieved, wherein the protocol for MR imaging and therefore the MR imaging itself are optimized in each iterative cycle. It has to be mentioned that all the methods described above - i.e. the scan recommendation method, the personal scan preparation method, the medical imaging method and the scan workflow execution method - can also be performed without an operator sitting in front of the scanner console. Thus, all the described methods can be performed next to the MR scanning system or even without a console room, i.e. fully automated, or with a monitoring operator sitting somewhere else and remotely controlled. Advantageously, a skilled operator can perform the investigation of the patient and the monitoring of the medical imaging procedure remotely and thus have access to professional skills when the MR system is available.

[0017] The scan recommendation system according to the present invention comprises at least one interface, preferably a plurality of input interfaces, for receiving evaluation data. The evaluation data comprises classification data, wherein the classification data classifies abnormalities detected based on the input data, wherein the input data relates to a medical condition of the patient. Furthermore, the scan recommendation system according to the present invention comprises a recommendation unit for automatically determining a next recommended MR protocol to be performed based on the MR scan, and determining a next recommended MR protocol to be performed based on the MR scan of the patient based on the received evaluation data.

[0018] In other words, the scan recommendation system can be configured to fully and automatically recommend the next most diagnostically insightful MR protocol for measurement. The evaluation data for the recommendation system can be based on a complete set or subset of inputs, which include image data (historical or current), image-derived metrics such as geometric properties of tissues and cavities (such as vessel wall thickness, cross-sections and ventricular volumes, etc.), electronic health record information, and / or user interface inputs. The scan recommendation system according to the present invention shares the advantages of the scan recommendation method according to the present invention.

[0019] The personal scan preparation system according to the present invention comprises an input interface for receiving input data about the medical condition of the patient. In addition, the personal scan preparation system comprises an evaluation unit and a scan recommendation system according to the present invention, the evaluation unit comprising: a detection unit for detecting anomalies in the input data; a classification unit for determining classification data as evaluation data based on the detected anomalies. The personal scan preparation system shares the advantages of the personal scan preparation method according to the present invention.

[0020] The medical imaging system according to the present invention comprises an MR scanning system and a scan recommendation system or a personal scan preparation system according to the present invention, which is coupled to the MR scanning system to determine a next recommended protocol and sends the next recommended protocol to the MR scanning system for performing an MR scan. The medical imaging system according to the present invention shares the advantages of the medical imaging method according to the present invention.

[0021] The scanning workflow execution system comprises a medical imaging system according to the present invention. In addition, the scanning workflow execution system comprises an abnormality detection unit for detecting abnormalities in the reconstructed MR image data. In addition, the scanning workflow execution system comprises a classification unit for determining classification data based on the detected abnormalities. The scanning workflow execution system further comprises an iteration unit for determining whether to iteratively perform the three steps of medical imaging, abnormality detection and abnormality classification based on the determined classification data until a termination criterion is obtained. The scanning workflow execution system shares the advantages of the scanning workflow execution method according to the present invention.

[0022] The basic components of the scan recommendation system according to the present invention, the personal scan preparation system according to the present invention, the medical imaging system according to the present invention, and the scan workflow execution system according to the present invention can be designed in the form of software components for the most part. This is particularly applicable to the recommendation unit of the scan recommendation system, which may include, for example, a fully connected neural network or a random forest, an evaluation unit, a detection unit, and a classification unit of the personal scan preparation system, and an abnormality detection unit and an iteration unit of the scan workflow execution system, as well as a part of the input interface. However, in principle, some of these components can also be implemented in the form of software-supported hardware such as FPGA, especially when it comes to particularly fast calculations. Similarly, the required interfaces, for example, if it is just a matter of transmitting data from other software components, can be designed as software interfaces. However, they can also be designed as hardware-based interfaces controlled by suitable software. In addition, some parts of the above-mentioned components can be distributed and stored in a local or regional or global electronic network or a combination of a network and software, especially in a cloud system.

[0023] A largely software-based implementation has the advantage that a medical imaging system, including an already used magnetic resonance imaging system, can be easily adapted by means of a software update so as to operate in the manner according to the invention. In this respect, the object is also achieved by a corresponding computer program product having a computer program, which can be directly loaded into a storage device, for example a control device of a medical imaging system, the computer program product having a program part so as to carry out all the steps of the method according to the invention when the program is executed in a medical imaging system, in particular in a control device. In addition to the computer program, such a computer program product may contain additional components, for example documentation and / or additional components including hardware components for using the software, for example hardware keys (dongles etc.).

[0024] For transport to and / or storage on or in the medical imaging system, a computer-readable medium is used, such as a memory stick, a hard disk or some other transportable or permanently installed data carrier, on which program parts of a computer program are stored which can be read in and executed by a computer unit of the medical imaging system. The computer unit may comprise, for example, one or more cooperating microprocessors etc. for this purpose. The computer program may also be obtained by downloading from the applicant's central online store or another internet-based source or a cloud-based source.

[0025] The dependent claims and the following description each contain particularly advantageous embodiments and improvements of the present invention. Specifically, a claim of one claim category may also be further improved similarly to a dependent claim of another claim category. In addition, within the scope of the present invention, various features of different exemplary embodiments and claims may also be combined to form new exemplary embodiments.

[0026] In a variation of the scan recommendation method according to the present invention, the automatic scan recommendation is based on using at least one of the following software types or a combination of software and hardware types:

[0027] - A machine learning model, preferably comprising at least one of the following:

[0028] - Shallow neural network models,

[0029] - Deep neural network models,

[0030] - Convolutional Neural Networks,

[0031] - fully connected neural networks,

[0032] - Support Vector Machines,

[0033] - Decision trees,

[0034] - Random Forest,

[0035] - Expert systems,

[0036] -Classic model algorithm.

[0037] For example, a scan recommendation method can be performed using a trained model (e.g., a fully connected neural network or a random forest). The trained model receives the above-mentioned evaluation data, which, for example, includes classified abnormalities. Therefore, the trained model is used to determine the next recommended MR protocol to be performed based on the patient's MR scan. The mentioned input data, such as the patient's image data, can be all kinds of information about the patient, which may affect the imaging procedure to be planned. The mentioned input data can be used directly as input for the trained model, or can undergo one or several preprocessing steps, such as by a machine learning or deep learning model, before they are fed to the trained model.

[0038] Preferably, in the scanning recommendation method according to the present invention, the evaluation data includes a first possible scanning workflow direction, which is determined by mapping the subject of the extracted keywords in the input data about the patient's medical condition to the probability of the next recommended protocol. The input data can be, for example, written information, that is, text data about suspected diagnosis, symptoms, indications involving symptoms, or body parts (symptoms are located at the body parts, and indications or suspected diagnoses involve the body parts). The input data may include electronic health record information, which includes at least one of suspected diagnosis, suspected disease, relevant observations of technicians, instructions from referring clinicians, body parts, patient data, or MR protocols. Therefore, the electronic health record information can add additional information about the patient's health status and can be used to modify the MR imaging protocol. In addition, the protocol can be modified by direct input by the user. Therefore, the input data may also include data entered by the user. Keyword extraction can be performed based on an expert system or a machine learning method or a deep learning method or a combination thereof. The extracted keywords can be determined based on data preprocessing such as lemmatization or stem extraction and bag of words generation. The topic is determined by a topic detection algorithm or a generated topic model. Topic detection can be achieved by unsupervised topic detection with latent Dirichlet allocation trained. Latent Dirichlet allocation can be performed using: classical machine learning methods such as hierarchical Bayesian models trained with variational inference as described in D. Blei et al. JMLR 3, 2003; or neural variational inference as described in Y. Miao et al. Proc. ICML, 2017. The mapping of topics to possible scanning workflow directions (i.e., possible recommended next scanning protocol groups) can be based on expert knowledge, customer preferences, or additional specified machine learning algorithms or deep learning algorithms. For each of the topics, a recommended next scanning protocol group with a probability value is assigned.

[0039] In addition, a second possible scanning workflow direction can also be determined by the scanning recommendation method according to the present invention, wherein the classification data included in the evaluation data is mapped to a probability value of the next recommended protocol. This mapping can be implemented using a mapping model such as a random forest or a fully connected neural network. Then, a final recommendation step is performed based on the second possible scanning workflow direction and optionally additionally based on the first possible scanning workflow direction. The final recommendation step can be implemented using a weighted average of the first possible scanning workflow direction and the second possible scanning workflow direction or using a fully connected neural network.

[0040] In a variation of the personal scan preparation method according to the present invention, the input data includes image data, and the step of detecting anomalies includes detecting anomalies in the image data. As mentioned above, the detection and classification of anomalies in the image data can be implemented using a neural network, such as a convolutional neural network. The anomaly can be a direct anomaly of the image data or an anomaly of a metric related to the image data. Therefore, the input data may also include image-derived metrics based on the image data, and the step of detecting anomalies may include detecting anomalies in the image-derived metrics. The image-derived metrics include geometric properties of tissues and cavities, which may be compared with standard data from a cohort. Such derived metrics include vessel wall thickness, cross-sections of blood vessels and ventricular volumes, gray mass volumes in the brain, and white mass volumes in the brain.

[0041] Also preferably, in the personal scan preparation method, the input data may also include electronic health record information and / or user interface input data, and the step of determining the evaluation data may include the following steps: extracting keywords from the electronic health record information and / or user interface input, determining the subject based on the extracted keywords, and determining the first possible scan workflow direction by mapping the subject to the probability of the next recommended protocol. Keywords may include, for example, information about at least one of a suspected diagnosis, a suspected disease, a technician's relevant observation, an instruction from a referring clinician, a body part, patient data, or an MR protocol. Keyword extraction may be performed based on an expert system or a machine learning method or a deep learning method or a combination thereof. Scan recommendations may be implemented by one or several classical and or machine learning algorithms. As mentioned above, electronic health record information may add additional information about the patient's health status and may be used to modify the MR imaging protocol. In addition, the MR imaging protocol may be modified by direct input by the user. Using keyword extraction enables automatic analysis of electronic health record information and / or user interface input about MR protocol-related content. Advantageously, information about previous medical examinations may also be integrated into scan recommendations based on written text or direct protocol data entered by the user. In this way, the database for automatically determining the next recommended MR protocol is expanded.

[0042] As mentioned above, different types of information can be combined as a basis for automatically determining the next recommended protocol. The input data listed above may include patient information, user information, and hospital information, which can be used directly as input to the scan recommendation system, or can undergo one or several pre-processing steps (e.g., through machine learning or deep learning models) before being fed to the scan recommendation system.

[0043] In a variation of the scanning recommendation system according to the present invention, the recommendation unit includes at least one of the following:

[0044] - A machine learning model comprising at least one of the following:

[0045] - Shallow neural network models,

[0046] - Deep neural network models,

[0047] - Convolutional Neural Networks,

[0048] - fully connected neural networks,

[0049] - Decision trees,

[0050] - Random Forest,

[0051] - Support Vector Machines,

[0052] - Expert systems,

[0053] -Classic model algorithm.

[0054] The described variants of the scan recommendation system share the advantages of the corresponding variants of the scan recommendation method.

[0055] The scan recommendation system according to the present invention may further include an input interface for receiving evaluation data, the evaluation data including a first possible scan workflow direction determined by mapping a topic based on the extracted keywords about the patient's medical condition to a probability value of a next recommendation protocol. The described variants of the scan recommendation system share the advantages of the corresponding variants of the scan recommendation method.

[0056] In an alternative variant of the scanning recommendation system according to the present invention, the recommendation unit includes: a mapping unit for determining a second possible scanning workflow direction by automatically mapping the classification data to a probability value of the next recommended protocol; and a final protocol recommendation unit for finally recommending the next recommended protocol based on the second possible scanning workflow direction and optionally additionally based on the first possible scanning workflow direction. The described variants of the scanning recommendation system share the advantages of the corresponding variants of the scanning recommendation method.

[0057] In a variation of the personal scan preparation system according to the invention, the input interface is configured to receive input data comprising medical image data, and the detection unit is configured to detect anomalies in the image data. The input interface may also be configured to receive input data comprising image-derived metrics based on the image data, and the detection unit may be configured to detect anomalies in the image-derived metrics. The variation of the personal scan preparation system shares the advantages of the corresponding variation of the personal scan preparation method.

[0058] In another variant of the personal scan preparation system according to the present invention, the input interface is configured to receive input data including electronic health record information and / or user interface input data. In this variant, the evaluation unit includes: an extraction unit for extracting keywords from the electronic health record information and / or the user interface input; a topic discovery unit for determining topics based on the extracted keywords; and a mapping unit for determining a first possible scan workflow direction as evaluation data by mapping the topic to a probability value of the next recommended protocol. The described variants of the personal scan preparation system share the advantages of the corresponding variants of the personal scan preparation method according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The above and other features and advantages of the present disclosure will be more apparent to those skilled in the art based on the detailed description of preferred embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0060] Figure 1 A flowchart of a scanning preparation method according to an embodiment of the present invention is shown.

[0061] Figure 2 Shows Figure 1 A flowchart of specific steps of the method shown,

[0062] Figure 3 A flowchart of a scanning workflow execution method according to an embodiment of the present invention is shown.

[0063] Figure 4 A block diagram showing a personal scan preparation system with a scan recommendation system according to an embodiment of the present invention is shown,

[0064] Figure 5 It is shown that the Figure 4 A block diagram of a medical imaging system of a personal scan preparation system is depicted,

[0065] Figure 6 A block diagram showing a scanning workflow execution system according to the present invention is shown.

[0066] Figure 7 An example of abnormal classification for image data is shown in detail.

[0067] Figure 8 An example of metric anomaly classification is shown in detail.

[0068] Fig. 9 An example of keyword analysis based on electronic health record input data or user interface input data is shown in detail.

[0069] Fig.10The scan recommendation function according to the embodiment of the present invention is shown in detail. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solutions and advantages of the present disclosure more obvious, the present disclosure will be further described in detail below by way of implementation examples.

[0071] Figure 1 A flowchart 100 of a scan preparation method for recommending a workflow for medical imaging of a patient according to an embodiment of the present invention is shown. Figure 1 As shown, in step 1.I of the scan preparation method, image data ID from the patient to be examined is obtained. The image data ID can be obtained by an initial scan for creating a T1-map and a T2-map. In addition, the image data ID can be received from a historical image previously obtained from the patient. In step 1.II, the image data ID is directly used as the first input data for determining the evaluation data. The image data about the image abnormality IA is analyzed. The detection of abnormal IA in the image data ID can be implemented using a convolutional neural network. In step 1.III, the image data ID is first subjected to a pre-diagnosis analysis step, in which some metrics IM of the image data ID are derived. For example, some pre-diagnosis analysis data, such as geometric properties of tissues and cavities, such as vascular wall thickness, cross-section and ventricular volume, etc., are detected and measured, and compared with standard data from a cohort. In addition, in step 1.IV, these image metrics IM are used as second input data, and abnormality detection is performed on them using a fully connected neural network, in which some abnormalities MA of the metrics IM can be detected. In step 1.V, the image abnormality IA is further classified using the convolutional neural network mentioned above. In step 1.VI, the detected anomaly MA of the metric IM is classified using, for example, a fully connected neural network, a decision tree or a random forest. Therefore, in steps 1.V and 1.VI, the evaluation data including the classification data CIA, CMA for the image anomaly IA and the metric anomaly MA are provided to a later step 1.XI, which is the actual scan recommendation step. The classification data CIA for the image anomaly IA includes a probability value group P(i) for the probability of the presence of different image anomalies IA in the image data ID input to the scan preparation method in step 1.I. The classification data CMA for the metric anomaly MA includes a probability value group P(j) for the probability of the presence of different metric anomalies IM, which can also be understood as input data provided to the scan preparation method in step 1.IV.

[0072] In other words, initial image data IDs are recorded to provide a first measurement step benchmark, and these image data IDs are subsequently input for anomaly detection and anomaly classification. A quality assurance step after each measurement can be employed to ensure adequate image quality, and a rescan is triggered if necessary (e.g., due to patient motion). Anomaly detection and anomaly classification can be implemented, for example, by a convolutional neural network, which can be implemented as a stand-alone. If anomaly detection / classification is applied, an embodiment advantageously includes the use of a quantitative MR imaging protocol, for example, for T1-maps and T2-maps for measuring the initial input images, to provide standardized image input for anomaly detection and classification.

[0073] In step 1.VII, some electronic health record information IHRI is used as third input data, which includes data from various sources such as objects (e.g. suspected diagnosis, reference indications, etc.) and / or other data sources (e.g. RIS, patient data, etc.). In step 1.VIII, some additional information UII - also referred to as user interface input UII - includes data indicating such as suspected diagnosis, reference indications, one or more body parts, desired MR protocol, etc. obtained as fourth input data for the personal scan preparation method. In step 1.IX, the input electronic health record information IHRI is processed using, for example, natural language processing (NLP) to perform keyword extraction. In this way, extracted keywords EK-IHRI of the input electronic health record information IHRI are obtained. Further continuing the personal scan preparation method, in step 1.X, the user interface input UII is processed using, for example, NLP to additionally or alternatively extract keywords EK-UII. As described above and in more detail later, the extracted keywords EK-IHRI, EK-UII are then assigned to different topics, and the topics are then mapped to probability value groups P(n), P(o), which represent possible scanning workflow directions PDI, PDU. Then, the possible scanning workflow directions PDI, PDU are provided as the third and fourth evaluation data to the automatic scanning recommendation method of step 1.XI. In step 1.XI, as further discussed herein, automatic scanning recommendations according to an embodiment of the present invention are performed based on a combination of various evaluation values ​​CIA, CMA, PDI, PDU to identify the next recommended protocol. Automatic scanning recommendations can be implemented based on classical or machine learning systems such as fully connected neural networks (FCNN). However, scanning recommendations are not limited to this implementation, and can be implemented according to any suitable type of classical algorithm, expert system or machine learning design that can process input, such as according to a classical, training or learning process, to recommend protocols based on a specific type of input.

[0074] exist Figure 2 In the figure, it is depicted that Figure 1Flowchart 100a of step 1.I shown. In step 1.Ia, T1-map measurements are performed based on the initial MR scan of the patient, and a T1-map T1-M is acquired. Compared to the main imaging process of the examination, the initial scan of the patient can be completed at a low resolution and does not take much time. In addition, in step 1.Ib, T2-map measurements are performed based on the initial MR scan of the patient, and a T2-map T2-M is acquired. In step 1.Ic, a quality analysis is performed on the acquired data T1-M, T2-M, and some quality-certified image data T1-MA, T2-MA are provided for further processing and analysis. In step 1.Id, the acquired data T1-MA, T2-MA - included by the image data ID mentioned in the context of step 1.I - are used to generate a synthetic contrast image SCA. In addition, as described above in Figure 1 As discussed in detail in the context of , in steps 1.II and 1.III, the acquired data T1-MA, T2-MA are also used as image data ID.

[0075] exist Figure 3 In the embodiment of the present invention, a flowchart 300 is depicted, which shows a scanning workflow execution method including the above-mentioned personal scanning preparation method, which includes scanning recommendations for scanning workflows and additional iterative protocol determination and measurement processes, which can be run until the relevant scanning abort criteria combination is met. Flowchart 300 starts from step 3.I, which includes the complete personal scanning preparation method, as in step 1.I to step 1.XI. Figure 1 Therefore, in step 3.I, perform the following Figure 1 The method described in and creates the next recommended protocol NRP based on the input data ID, IM, IHRI, UII, the input data ID, IM, IHRI, UII as in Figure 1 The next recommended protocol NRP is put into the scan queue. Furthermore, in step 3.II, some additional manually or electronically selected protocols EP to be scanned are added to the scan queue due to, for example, clinical criteria, health insurance guidelines, radiologist or technician preferences. Then in step 3.III, the next scan protocol NSP is sent from the scan queue to the MR scanning unit. In step 3.IV, the actual scanning process for examining the patient and acquiring raw data is performed as a basis for reconstructing image data MR-ID with high resolution and quality according to the patient. In step 3.V, the reconstructed image data MR-ID is then analyzed with respect to potential abnormalities MR-IA in the image data MR-ID. In case some abnormalities MR-IA are detected, this is Figure 3In the case where no abnormal MR-IA is detected, the process continues to step 3.VII. Figure 3 In step 3.VI, it is determined whether any additional RI items remain in the queue, which means that the protocol remains in the queue, which must be processed. If it is Figure 3 In the case where there is no additional item RI left in the queue, the process continues to step 3.III. Figure 3 denoted by "n" in the figure, the process proceeds to step 3.IX, i.e., the scan is completed and the reconstructed image data MR-ID is sent to the medical staff. If the process proceeds to step 3.VII, the detected abnormal MR-IA is classified, for example, using a fully connected neural network. In step 3.VIII, it is determined whether the predicted probabilities P(i) of the class i of the classified abnormal data MR-CIA are all subject to the adjustable probability threshold θ c Limit. If Figure 3 In this case, denoted by "j" in , the process continues with step 3.IX. When at least one of the probabilities P(i) of class i exceeds the adjustable probability threshold θ c In this case, Figure 3 denoted by "n" in the figure, the process proceeds to step 3.X, where it is checked whether any scan abort flags requiring the completion of the scan workflow are raised. The abort flag can be raised based on the abort criteria ACR, which can be reaching a time limit, reaching a maximum number of scan protocols, or no scan protocols remaining in the scan queue. If the abort criteria ACR is not met, this Figure 3 In the case where the abnormal MR-CIA detected is taken into account in the scan recommendation for the next MR scan, the process continues with step 3.I. In the case where at least one of the abort criteria ACR is met, this Figure 3 Indicated by "j" in the figure, the process ends with step 3.IX.

[0076] exist Figure 4 , a block diagram is shown, which shows a personal scan preparation system 40 according to an embodiment of the present invention. Figure 4The various blocks or modules shown and described may be implemented as any suitable number of hardware circuits, processors, and / or software (e.g., algorithms executed). The personal scan preparation system 40 includes an image data input interface 41 that receives the above-mentioned image data ID, for example, from an initial scan. In addition, the scan recommendation system 40 includes a metric data input interface 42 that receives the image metric IM. In addition, the personal scan preparation system 40 includes an anomaly detection unit 43, which is configured to detect some anomalies IA in the image data ID and some anomalies MA in the image metric IM. The image anomalies IA and the metric anomalies MA are sent to an anomaly classification unit 44, which is configured to classify the detected anomalies IA, MA, and send the detected classifications CIA, CMA to the scan recommendation unit 45.

[0077] In one variation, the anomaly detection unit 43 may perform anomaly detection using, for example, a convolutional neural network configured to process the initial image ID as input and provide the probability of the input image ID being normal (i.e., healthy) versus abnormal (i.e., requiring further imaging) as output. If the probability of abnormal IA for the provided input image ID is above a certain customizable threshold ξd, the initial input image ID with the labeled abnormal IA may then be inserted into a subsequent abnormal classification unit 44, which classifies the abnormal IA into a category i with probability values ​​P(i) for different categories, which category i provides meaningful input to the scan recommendation unit 45. Suitable classification categories i may include, for example, one or more of the following exemplary classification groups: disease categories (tumor, hemorrhage, MS, etc.), abnormal anatomical regions (white matter, dark matter, brainstem, CSF, etc.), abnormal metrics (white matter volume, vessel wall thickness, carotid artery cross section, etc.). Then, the category i with a high predicted classification probability P(i) is forwarded to the scan recommendation unit 45 as one of the inputs. The selection of which categories are forwarded to the scan recommendation unit 45 may be performed, for example, by selecting a custom threshold probability value ξc, such that all categories i are forwarded whose probability P(i) satisfies P(i)>ξc. In another variant, the forwarding criterion for category i may alternatively be set to

[0078] Meet max i P(i) and P(i)>ξc.

[0079] Additionally, the personal scan preparation system 40 also includes a third input interface 47 for receiving electronic health record information IHRI. These electronic health record information IHRI of the patient may include, for example, known diseases, symptoms, clinical indications for scanning, contraindications such as contrast intolerance, and may be extracted, for example, from a RIS or another electronic system containing patient data. In addition, the personal scan preparation system 40 includes a manual input interface 48 as a fourth input interface for receiving some user interface inputs UII, for example, containing suspected diseases or manually selected protocols to be included in the scan workflow. The additional received data IHRI, UII are sent to a keyword extraction unit 46, which extracts some relevant keywords EK-IHRI, EK-UII from the received data IHRI, UII. In addition, keywords are assigned to topics, which are mapped to possible scan workflow directions PDI based on the electronic health record information IHRI and possible scan workflow directions PDU based on the user interface input UII as evaluation values. These evaluation values ​​are then also sent to the scan recommendation system 45. The scan recommendation system 45 performs automatic scan recommendations based on a combination of various inputs CIA, CMA, PDI, PDU to identify the next recommended protocol NRP. The identified next recommended protocol NRP is output by the output interface 49, which is also part of the personal scan preparation system 40. In various embodiments, the personal scan preparation system 40 can be implemented as any suitable number of hardware circuits, processors and / or software (e.g., executed algorithms). As an example, the scan recommendation system 40 can be implemented as any suitable type of classical or machine learning system, such as a fully connected neural network (FCNN). In an embodiment, the personal scan preparation system 40 can work in combination with other classical or machine learning processes, processors, hardware components and / or interfaces, and / or additional executed algorithms to receive data via any suitable number of inputs. For example, the personal scan preparation system 40 includes four input interfaces 41, 42, 43, 44, and works based on four inputs ID, IM, IHRI, UII. Although this is by way of example and not limitation. Embodiments include personal scan preparation systems 40 applicable to any suitable number of inputs that may be greater than or less than four inputs, and may include Figure 4 Examples of different alternative inputs are shown. In various embodiments, each of the input interfaces of the scan recommendation system 45 can be implemented using any suitable number of hardware circuits, processors, and / or software (e.g., algorithms executed). In various embodiments, each of the input interfaces of the scan recommendation system 45 can be implemented as a system (e.g., a classical or machine learning system) that is optimally or preferably configured according to the specific type of data received at each corresponding input interface.

[0080] The scan recommendation system 45 may be implemented based on, for example, an expert system, a machine learning or deep learning method, or a combination thereof. The scan recommendation algorithm of the scan recommendation system 45 may be implemented, for example, by one or several classical or machine learning algorithms that are able to accept evaluation data (predicted upstream by the anomaly classification unit and the keyword extraction unit), and then predict the most appropriate scan protocol to be added to the scan queue (as can be seen in Figure 6 As an example, a method based on natural language processing with suitable word embeddings may be suitable for this task. In an embodiment, the scan recommendation system 45 can be designed / trained, for example, based on clinical expert knowledge and medical association guidelines.

[0081] exist Figure 5 In the figure, it is schematically shown that Figure 4 The medical imaging system 50 of the personal scan preparation system 40 is described in detail in the context of . The medical imaging system 50 also includes an MR scanning system 51, which provides T1-map data T1-M and T2-map data T2-M to the personal scan preparation system 40. In addition, the medical imaging system 50 includes a contrast image generation unit 52, which generates some synthetic contrast images based on the T1-map data T1-M and T2-map data T2-M for use by radiologists who are accustomed to a specific type of image with a specific contrast. The scan recommendation system of the personal scan preparation system 40 proposes a next recommended protocol NRP for the MR imaging process.

[0082] exist Figure 6In the block diagram depicted, a scan workflow execution system 60 is shown, which has a personal scan preparation system 40 according to an embodiment of the present invention and an additional iterative protocol determination and measurement system 60a that can run until the relevant scan abort criteria ACR combination is met. The iterative protocol determination and measurement system 60a includes a scan queue management unit 61, which receives the next recommended protocol NRP from the scan recommendation system 45 of the personal scan preparation system 40, and further receives some manually or electronically selected protocol EPs to be scanned, and creates a scan queue including the received protocol NRP, EP. In addition to the data-driven new recommended protocol NRP, the embodiment includes manually or electronically adding the protocol to the scan queue. This option may be required to make the scan workflow comply with, for example, local hospital standards, health insurance provider requirements, medical association guidelines, user preferences, etc. Therefore, the scan protocol NRP, EP from two channels, namely the personal scan preparation system 40 and the manual / electronic input, is fed to the scan queue management unit 61. The scan queue management unit 61 then sorts and / or manages the scan protocol jobs NRP, EP according to, for example, scheduling time, relevance, redundancy, or input channels. The scan queue management unit 61 can (depending on operator and user preferences) accept both protocol input channels, or in two extreme cases, reject one or the other input channel to implement manual or automatic mode at a high level. Depending on operator and user preferences, the scan queue management unit 61 can be configured to add only specific protocols, protocols with specific parameters, or protocols that meet specific requirements to the scan queue. The protocols in the scan queue can be executed during the next MR scanner measurement, resulting in a reconstructed MR image. Where applicable, quality assurance steps and optional rescans (if indicated) can be inserted at this time or before image reconstruction.

[0083] Therefore, the iterative protocol determination and measurement system 60a also includes an abnormality detection unit 62, which is configured to detect abnormal MR-IA in the MR image data MR-ID received from the MR scanning system 51. The abnormal data MR-IA is sent to an abnormality classification unit 63, which is also part of the iterative protocol determination and measurement system 60a. Therefore, the image MR-ID that is easy to reconstruct is inserted into the abnormality detection unit 62 (for example, similar to the above abnormality detection unit 43) suitable for the corresponding MR scanning protocol. If abnormal MR-IA is detected, the abnormal MR-IA can be classified by the subsequent abnormality classification unit 63 described above, which is similar to the above abnormality classification unit 44 in function. The abnormality classification unit 63 arranges the abnormal data MR-IA in the abnormal category MR-CIA. Also referred to as the classification data MR-CIA is sent to the iteration unit 64, which determines whether the scan is completed based on the classification data MR-CIA and the predetermined standard ACR or whether an additional MR scan must be planned based on the classification abnormal MR-CIA. The probability P(i) of abnormal category i in the classification data MR-CIA is less than the threshold θ c If the probability P(i) of at least one of the abnormal categories i is higher than the threshold θ c , it is determined whether the termination criteria ACR is met, which may be that the time limit is reached, the maximum number of scan protocols is reached, or there are no scan protocols remaining in the queue. If one of these termination criteria ACRs is met, the scan is completed. If none of the termination criteria ACRs is met, a new recommended protocol NRP is created based on the detected abnormal MR-CA. That is, the classification data MR-CIA can be used as evaluation data for the scan recommendation system 45 of the scan preparation system 40, and further MR scans are performed using the newly recommended scan protocol.

[0084] Therefore, after the anomaly classification, the following steps may be recommended: Check if any scan abort flags requiring completion of the scan workflow have occurred, for example due to reaching a time limit, reaching a maximum number of scan protocols, no scan protocols remaining in the queue, etc. If no scan workflow abort criteria ACR is affirmative (i.e. identified), the scan recommendation system 45 may be queried again with the classification data MR-CIA of the new scan image data MR-ID for one or several additional scan recommendation cycles until one or a relevant combination of the following criteria ACRs are met and result in the completion of the scan workflow:

[0085] 1. No additional abnormal MR-IA was detected.

[0086] 2. The classification probability P(i) of all abnormal categories i is lower than a certain threshold

[0087] P(i)≤θ c .

[0088] 3. There are no items left in the scan queue.

[0089] 4. Scan Workflow Abort Criteria Any or relevant combination of ACRs is affirmative.

[0090] If the iteration unit 64 decides that the scan must be completed, a command DO is sent to the MR scanning unit 51 for sending the image data MR-ID to the medical personnel.

[0091] In the case that the abnormality detection unit 62 does not detect the abnormality MR-IA in the MR image data MR-ID from the MR scanning system 51, the information N-IA is sent to the scan queue management unit 61, which determines whether there are additional items remaining in the queue. The scan queue status can then determine whether another pending scan will be performed next, or whether the scan workflow is completed. In the first case, the scan queue management unit 61 continues with the next protocol in the queue.

[0092] Although a recurring configuration of an anomaly detection unit with an anomaly classification unit connected in series is described herein, this is not the only practical configuration and is provided by way of example and not limitation. Alternatively, embodiments described herein may feed images, image metrics, or other suitable data into both the anomaly detection unit and the anomaly classification unit simultaneously for later combination of the results of the two modules.

[0093] In addition, embodiments also include setting thresholds (e.g., ξd, ξc, ) is modified to different optimal values, depending on, for example, the location of the corresponding unit within the scanning workflow execution system 60, the scanning protocol NSP adopted for image generation and other factors, specific applications, etc.

[0094] The methods and systems described herein allow a trained or untrained technician to perform a highly personalized MRI scanning workflow for each patient and generate the optimal set of MRI images required to diagnose that patient's condition, taking into account potential abnormalities detected on the recorded image data, electronic health record information, and manual or electronic setup requirements from the operator.

[0095] The techniques discussed herein can therefore provide the basis for a fully autonomous and optimized MR imaging workflow that still allows it to be customized manually or electronically by the operator as needed. MR scanner operators can choose to run the MR scanning workflow in fully autonomous, fully manual, or hybrid modes.

[0096] In autonomous mode, depending on the operator’s preference, by setting an appropriate threshold ξ c , By selecting the associated scan abort criteria ACR, the MR scan workflow can be configured to, for example, minimize the total scan time per patient and / or minimize the probability of missing the associated MR protocol required for image reading and diagnosis.

[0097] In another variation, the operator may first record a standard MR protocol set (with appropriate settings Figure 3 Workflow steps 3.II, 3.III) in , and employing the scan recommendation system 45 as needed to determine and record any additional MR scans that may be used for additional indications for the current patient.

[0098] exist Figure 7 In , an example for anomaly classification of image data is shown for a more detailed understanding. Figure 7 In the upper part of the figure, the training process of image data anomaly classification is depicted. First, the labeled image data L-ID as training data is input to the combined anomaly detection and classification units 43, 44. The combined anomaly detection and classification units 43, 44 include a convolutional neural network, which creates classification data CIA as an answer to the labeled image data L-ID. Then, the classification data CIA is compared with the labeled classification data L-CIA, and the convolutional neural network is adjusted so that the classification data CIA is assimilated into the labeled classification data L-CIA. Figure 7 In , some categories of image abnormality IA are described, such as high signal area hea, low signal area hoa, abnormal tissue at, midline shift ms, vascular stenosis vs, plaque pl, demyelination demy, and mass effect me. The abnormality classification data can be represented by a 1×i vector. Alternatively, tissue-specific abnormalities or abnormal IA can also be used for abnormality classification. In this case, abnormal IA is classified by tissue type and abnormality type. For example, the tissue type for the brain can be white matter, gray matter, hippocampus, cerebellum, or corpus callosum. The abnormality type can be hypertrophy, atrophy, or abnormal morphology. In Figure 7 The classification process is described in the lower part of FIG. 1 . First, the image data ID is input to the anomaly detection and classification units 43, 44. The trained convolutional neural network of the anomaly detection and classification units 43, 44 now outputs classification data CIA, which includes a 1×i vector including a probability value P(i) for the above anomaly IA.

[0099] exist Figure 8 In , an example for metric anomaly classification is shown for a more detailed understanding. Figure 8In the upper part of , the training process of metric anomaly classification is depicted. First, the labeled metric data L-IM is input to the anomaly detection and classification units 43, 44 as training data. The anomaly detection and classification units 43, 44 include a convolutional neural network, which creates classification data CMA as an answer to the labeled metric data L-IM. Then, the classification data CMA is compared with the labeled classification data L-CMA, and the convolutional neural network is adjusted so that the classification data CMA is assimilated into the labeled classification data L-CMA. Figure 8 In , the measurement data IM includes values ​​about WMv volume, GMv volume, arterial cross section acs, venous cross section vcs, ventricular volume vv, and additional information about the patient's age pa and the patient's gender sm. The measurement data IM can be represented by a 1×k vector, and the abnormal classification data CMA can be represented by a 1×j vector, where j=k-2, because the classification data CMA does not include information about the patient's age and gender. Figure 8 The classification process is described in the lower part of FIG. 1 . First, the measurement data IM is input to the anomaly detection and classification units 43, 44. The trained convolutional neural network of the anomaly detection and classification units 43, 44 now outputs classification data CMA, which includes a 1×j vector including the probability value P(j) for the above anomaly MA.

[0100] exist Fig. 9 In, an example for keyword analysis based on electronic health record input data IHRI or user interface input data UII is shown in detail. The electronic health record input data IHRI may include information about suspected diagnosis SD, symptoms SP, reference indication RIN, specified body part BR and MR protocol of PRR request. Keyword extraction starts with data preprocessing step 9.I, such as lemmatization and stemming. Then, in step 9.II, bag of words generation is performed. Stemming is a basic NLP (NLP = Natural Language Programming) data preprocessing step, by which the morphological variations (inflections and inflections) of words are traced back to the root (tripped -> tripped). Lemmatization is an important NLP data preparation step, which reduces words from different inflected forms to their basic form (better -> good). Bag of words generation aims to generate an unclassified catalog of preprocessed, clean words (see above) from text documents, where the frequency of words is parsed. Thereafter, in step 9.III, topic detection based on a topic model is performed, where different topics T1, T2, T3, ..., TN are detected. Each identified topic as a keyword group is then mapped to a possible scan workflow direction PDI as the recommended next scan protocol group T1w, T2w, T2 FL, DWI, SWI. This mapping can be based on expert knowledge, customer preferences, or additionally specified machine learning or deep learning based algorithms. Fig. 9In , the group PDI of the recommended next scan protocol is represented by a 1×n vector. The group PDI of the recommended next scan protocol includes, for example, T1 weighted protocol T1w, T2 weighted protocol T2w, T2 FLAIR protocol T2 FL, diffusion weighted imaging protocol DWI and susceptibility weighted imaging protocol SWI. If human mapping from subject to protocol is not desired or possible, another possibility is to skip subject detection 9.III and instead train a model, for example, a model such as a convolutional neural network that predicts the protocol directly from step 9.II. This is more direct, but provides poor interpretability to the user.

[0101] exist Fig.10 , the function of the scan recommendation 45 is shown in detail. On the left, an abnormality classification unit 44 for classification of image abnormalities IA and metric abnormalities MA is depicted. In addition, a keyword extraction unit 46 for extracting keywords from the electronic health record information data IHRI and the user input data UII is described. Figure 7 , Figure 8 and Fig. 9 As explained in the context of , different classification data CIA, CMA are output by the classification unit 44, and possible scan workflow direction data PDI, PDU are output by the keyword extraction unit 46. These data are input into the scan recommendation system 45 as evaluation data. Then, the abnormal classification data CIA, CMA are processed by the mapping unit 101, which operates based on a protocol mapping model such as a random forest model or a fully connected neural network. In this way, some probability values ​​P(m) of the next recommended protocol or possible scan workflow direction PDA are created, which have a format similar to the scan workflow direction data PDI, PDU. All these recommended next protocol data PDA, PDI, PDU are then input into the final protocol recommendation unit 102, which determines the final recommended protocol NRP by using a weighted average or a random forest model or a fully connected neural network. Therefore, a high probability value P (p=3) for the T2-FLAIR protocol T2 FL is determined. Therefore, the T2-FLAIR protocol T2 FL is recommended for the next MR scan.

[0102] The above description is merely a preferred embodiment of the present disclosure, but is not intended to limit the present disclosure, and any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A scanning recommendation method, The following steps are involved: - receiving assessment data comprising classification data (CIA, CMA) classifying anomalies (IA, MA) of image-derived metrics (IM) detected in input data (ID, IM, IHRI, UII) regarding a medical condition of a patient, - Based on the type of the abnormality determined by classification of the abnormality by classification data included in the evaluation data, automatically determining a next recommended MR protocol (NRP) to be performed adapted to the type of the abnormality from the MR scan of the patient.

2. The scanning recommendation method according to claim 1, in, Automatic determination of the next MR protocol (NRP) is based on the use of at least one of the following: A machine learning model comprising at least one of: - Convolutional Neural Networks, - fully connected neural networks, - Shallow neural network models, - Deep neural network models, - Decision trees, - Random Forest, - Support Vector Machines, - Expert systems, -Classic model algorithm.

3. The scanning recommendation method according to claim 1 or 2, in, The evaluation data comprises a first possible scanning workflow direction (PDI, PDU) by combining a subject (T) based on the extracted keywords (EK-IHRI, EK-UII) about the medical condition of the patient in the input data (IHRI, UII) with a 1 , T 2 , T 3 , ..., T N ) is mapped to the probability value (P(n), P(o)) of the next recommended protocol (NRP).

4. The scanning recommendation method according to claim 1 or 2, in, Automatic determination of the next recommended MR protocol (NRP) includes the following steps: - determining a second possible scanning workflow direction (PDA) by automatically mapping said classification data (CIA, CMA) to a probability (P(m)) of the next recommended protocol (NRP), - automatically and finally recommending said Next Recommended Protocol (NRP) based on said second possible scan workflow direction (PDA).

5. The scanning recommendation method according to claim 3, in, Automatic determination of the next recommended MR protocol (NRP) includes the following steps: - determining a second possible scanning workflow direction (PDA) by automatically mapping said classification data (CIA, CMA) to a probability (P(m)) of the next recommended protocol (NRP), - automatically and finally recommending said next recommended protocol (NRP) based on said second possible scanning workflow direction (PDA) and based on said first possible scanning workflow direction (PDI, PDU).

6. A method for preparing a personal scan, The following steps are involved: - receiving input data (ID, IM, IHRI, UII) regarding a medical condition of a patient, wherein said input data (ID, IM, IHRI, UII) comprises image-derived metrics (IM) based on image data (ID), - Determining evaluation data based on the input data (ID, IM, IHRI, UII), comprising the following steps: - detecting anomalies (IA, MA) of image derived metrics (IM) in said input data (ID, IM, IHRI, UII), - determining classification data (CIA, CMA) as evaluation data based on the detected anomalies (IA, MA), - Based on the evaluation data, executing the scanning recommendation method according to any one of claims 1 to 5.

7. The personal scanning preparation method according to claim 6, in, - said input data (ID, IM, IHRI, UII) comprises image data (ID), and - The step of detecting anomalies (IA, MA) comprises detecting anomalies (IA) in said image data (ID).

8. The personal scanning preparation method according to claim 7, in, - the input data (ID, IM, IHRI, UII) also includes electronic health record information (IHRI) and / or user interface input data (UII), and the step of determining the assessment data comprises the following steps: - extracting keywords (EK-IHRI, EK-UII) from the electronic health record information (IHRI) and / or the user interface input (UII), - Determine the topic (T) based on the extracted keywords (EK-IHRI, EK-UII) 1 , T 2 , T 3 , ..., T N ), - By applying the subject (T 1 , T 2 , T 3 , ..., T N ) is mapped to the probability value (P(n), P(o)) of the probability of the next recommended protocol (NRP) to determine the first possible scanning workflow direction (PDI, PDU).

9. A medical imaging method, The following steps are involved: - determining a Next Recommended Protocol (NRP) based on the method according to any one of claims 1 to 8, - sending said next recommended protocol (NRP) to the MR scanning system (51), - An MR scan is performed by the MR scanning system (51) based on the received protocol (NRP), wherein MR image data (MR-ID) from the patient are recorded.

10. A scanning workflow execution method, The following steps are involved: -(a) performing a medical imaging method according to claim 9, - (b) detecting anomalies (MR-IA) in the reconstructed MR image data (MR-ID), - (c) determining classification data (MR-CIA) based on the detected abnormality (MR-IA), - The above three steps (a)-(c) are iteratively performed based on the determined classification data (MR-CIA) until the abort criterion (ACR) is obtained.

11. A scanning recommendation system (45), include: - an input interface (41, 42) for receiving evaluation data including classification data (CIA, CMA) classifying anomalies (IA, MA) of image-derived metrics (IM) detected based on input data (ID, IM, IHRI, UII) regarding a medical condition of a patient, - A recommendation unit (101, 102) for automatically determining a next recommended MR protocol (NRP) to be performed adapted to the type of abnormality based on the type of the abnormality determined by classification of the abnormality by classification data included in the received evaluation data, according to the MR scan of the patient.

12. The scanning recommendation system according to claim 11, in, The recommendation unit (101, 102) includes at least one of the following: - A machine learning model comprising at least one of the following: - Convolutional Neural Networks (CNN), - Fully Connected Neural Network (FCNN), - Shallow neural network models, - Deep neural network models, - Decision trees, - Random Forest, - Support Vector Machines, - Expert systems, -Classic algorithm.

13. The scanning recommendation system according to claim 11 or 12, comprising an input interface for receiving evaluation data, the evaluation data comprising a first possible scanning workflow direction (PDI, PDU), the first possible scanning workflow direction (PDI, PDU) being selected by dividing a topic (T) based on the extracted keywords (EK-IHRI, EK-UII) about the medical condition of the patient 1 , T 2 , T 3 , ..., T N ) is mapped to the probability value (P(n), P(o)) of the next recommended protocol (NRP).

14. The scanning recommendation system according to claim 11 or 12, in, The recommendation unit (101, 102) comprises: - a mapping unit (101) for determining a second possible scanning workflow direction (PDA) by automatically mapping the classification data (CIA, CMA) to a probability value (P(m)) of the probability of a next recommended protocol (NRP), - a final protocol recommendation unit (102), configured to finally recommend the next recommended protocol (NRP) based on the second possible scanning workflow direction (PDA).

15. The scanning recommendation system according to claim 13, in, The recommendation unit (101, 102) comprises: - a mapping unit (101) for determining a second possible scanning workflow direction (PDA) by automatically mapping the classification data (CIA, CMA) to a probability value (P(m)) of the probability of a next recommended protocol (NRP), - a final protocol recommendation unit (102), configured to finally recommend the next recommended protocol (NRP) based on the second possible scanning workflow direction (PDA) and based on the first possible scanning workflow direction (PDI, PDU).

16. A personal scanning preparation system (40), include: - an input interface (41, 42, 47, 48) for receiving input data (ID, IM, IHRI, UII) regarding a medical condition of a patient, wherein the input data (ID, IM, IHRI, UII) comprises image-derived metrics (IM) based on the image data (ID), -Assessment units including: - a detection unit (43) for detecting anomalies (IA, MA) of image derived metrics (IM) in said input data (ID, IM, IHRI, UII), a classification unit (44) for determining classification data (CIA, CMA) as evaluation data based on the detected anomaly (IA, MA), - A scanning recommendation system (45) according to any one of claims 11 to 15.

17. The personal scan preparation system according to claim 16, in, - the input interface (41) is configured to receive input data (ID) comprising medical image data (ID), and - the detection unit (43) is configured to detect anomalies (IA) in the image data (ID).

18. The personal scan preparation system according to claim 16, in, - the input interface (47, 48) is configured to receive input data (IHRI, UII) comprising electronic health record information (IHRI) and / or user interface input data (UII), - The evaluation unit comprises: - an extraction unit, configured to extract keywords (EK-IHRI, EK-UII) from the electronic health record information (IHRI) and / or the user interface input (UII), - A topic discovery unit for determining topics (T 1 , T 2 , T 3 , ..., T N ), - a mapping unit for mapping the subject (T 1 , T 2 , T 3 , ..., T N ) is mapped to the probability value (P(m), P(o)) of the probability of the next recommended protocol (NRP) to determine the first possible scanning workflow direction as evaluation data.

19. A medical imaging system (50), include: - MR scanning system (51), - A system (40) according to any one of claims 11 to 18, which is coupled to the MR scanning system (51) to determine a next recommended protocol (NRP) and sends the next recommended protocol (NRP) to the MR scanning system (51) for performing an MR scan.

20. A scanning workflow execution system (60), include: - The medical imaging system (50) according to claim 19, - an anomaly detection unit (62) for detecting anomalies (MR-IA) in the reconstructed MR image data (MR-ID), - a classification unit (63) for determining classification data (MR-CIA) based on the detected abnormality (MR-IA), An iteration unit (64) is used to determine whether to iteratively perform the three steps of medical imaging, abnormality detection and abnormality classification based on the determined classification data (MR-CIA) until a termination criterion (ACR) is obtained.

21. A computer program product having a computer program, which can be directly loaded into a storage device of a control device of a medical imaging system (50), and when the computer program is executed in the control device, the computer program product has a program part for executing all steps of the method according to any one of claims 1 to 10.

22. A computer-readable medium having stored thereon a program part which can be read in and executed by a computer unit of a medical imaging system (50), so that when the program part is executed in the computer unit, all steps of the method according to any one of claims 1 to 10 are performed.

Images