Automatic imaging protocoling of medical images using large language models

Abstract

Systems and methods for automatic imaging protocoling are provided. Patient data from one or more patient databases is received. Imaging protocoling information is extracted from the patient data using a first machine learning model. An imaging protocol template is selected from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model. The selected imaging protocol template is fine-tuned based on the patient data using a third machine learning model. The fine-tuned imaging protocol template is output.

Description

AUTOMATIC IMAGING PROTOCOLING OF MEDICAL IMAGES USING LARGE LANGUAGE MODELSTECHNICAL FIELD

[0001] The present invention relates generally to LLMs (large language models), and in particular to automatic imaging protocoling of medical images using LLMs.BACKGROUND

[0002] Imaging protocoling refers to the process of reviewing clinician orders for the acquisition of medical images of a patient and assigning specific imaging protocols directing the acquisition of the medical images. The protocols typically define the imaging modality, the anatomical region of interest, and the acquisition parameters for acquiring the medical images. Recently, conventional automatic imaging protocoling approaches have been proposed using rules-based text processing of patient data. However, such conventional rules-based approaches do not generalize well across clinical sites.BRIEF SUMMARY OF THE INVENTION

[0003] In accordance with one or more embodiments, systems and methods for automatic imaging protocoling are provided. Patient data from one or more patient databases is received. Imaging protocoling information is extracted from the patient data using a first machine learning model. An imaging protocol template is selected from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model. The selected imaging protocol template is fine-tuned based on the patient data using a third machine learning model. The fine-tuned imaging protocol template is output.

[0004] In one embodiment, text of the extracted imaging protocoling information and descriptions of the plurality of candidate imaging protocol templates are encoded into vectors. A similarity measure between the vector of the text of the extracted imaging protocoling information and each of the vectors of the descriptions of the plurality of candidate imaging protocol templates is determined. A candidate imaging protocoltemplate having a vector most similar to the vector of the of the extracted imaging protocoling information is selected based on the similarity measures.

[0005] In one embodiment, the selected imaging protocol template is fine-tuned based on the patient data and protocoling standards for a clinical site using the third machine learning model. The selected imaging protocol template may be fine-tuned according to a decision forest.

[0006] In one embodiment, clusters of imaging protocols in systems of one or more clinical sites are identified. Patient scenarios that trigger the clusters are identified. One or more of the plurality of candidate imaging protocol templates are automatically generated based on the identified patient scenarios.

[0007] In one embodiment, the fine-tuned selected imaging protocol template annotated with information from the patient data justifying selection of protocols and justifying the fine-tuning is presented.

[0008] In one embodiment, information relating to imaging protocoling in the patient data is summarized using the first machine learning model.

[0009] In one embodiment, the patient data is retrieved from the one or more patient databases using the first machine learning model.

[0010] In one embodiment, the first machine learning model comprises a first LLM, the second machine learning model comprises a second LLM, and the third machine learning model comprises a third LLM.

[0011] These and other advantages of the invention will be apparent to those of ordinary skill in the art by reference to the following detailed description and the accompanying drawings.BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 shows a schematic diagram for an automatic imaging protocoling system for acquisition of medical images of a patient, in accordance with one or more embodiments;

[0013] Figure 2 shows a method for automatic imaging protocoling for acquisition of medical images of a patient, in accordance with one or more embodiments;

[0014] Figure 3 shows an exemplary artificial neural network that may be used to implement one or more embodiments;

[0015] Figure 4 shows a convolutional neural network that may be used to implement one or more embodiments;

[0016] Figure 5 shows a schematic structure of a recurrent machine learning model that may be used to implement one or more embodiments; and

[0017] Figure 6 shows a high-level block diagram of a computer that may be used to implement one or more embodiments.DETAILED DESCRIPTION

[0018] The present invention generally relates to methods and systems for automatic imaging protocoling of medical images using LLMs (large language models). Embodiments of the present invention are described herein to give a visual understanding of such methods and systems. A digital image is often composed of digital representations of one or more objects (or shapes). The digital representation of an object is often described herein in terms of identifying and manipulating the objects. Such manipulations are virtual manipulations accomplished in the memory or other circuitry / hardware of a computer system. Accordingly, is to be understood that embodiments of the present invention may be performed within a computer system using data stored within the computer system.

[0019] Embodiments described herein provide for automatic imaging protocoling of medical images using LLMs. The automatic imaging protocoling in accordance with embodiments described herein avoids the brittleness of conventional rules-based approaches by leveraging various LLMs to extract imaging protocoling information directly from unstructured patient data, select an imaging protocol template based on the extracted imaging protocoling information, and fine-tune the selected imaging protocol template.

[0020] Figure 1 shows a schematic diagram 100 for an automatic imaging protocoling system for acquisition of medical images of a patient, in accordance with one or more embodiments. As shown in schematic diagram 100, the automatic imaging protocolingsystem comprises data collection and preprocessing module 104, protocol template selection module 106, protocol refinement module 110, and, optionally, protocol template definition module 108 and protocol review interface 112. The modules 104-112 of the automatic imaging protocoling system may be implemented by one or more suitable computing devices, such as, e.g., computer 602 of Figure 6.

[0021] Data collection and preprocessing module 104 retrieves and cleans patient data from one or more patient databases 102-A, 102-B, . . ., 102-n (collectively referred to as patient databases 102) and extracts imaging protocoling information from the patient data. Protocol template selection module 106 selects a best matching imaging protocol template based on the extracted imaging protocoling information. Protocol refinement module 110 fine-tunes the selected imaging protocol template according to, e.g., the patient data and protocoling standards for a clinical site. Optionally, protocol template definition module 108 may be applied for defining candidate protocol templates and protocol review interface 112 may be applied for reviewing and modifying the fine-tuned imaging protocol template.

[0022] Figure 2 shows a method 200 for automatic imaging protocoling for acquisition of medical images of a patient, in accordance with one or more embodiments. The steps of method 200 may be performed by one or more suitable computing devices, such as, e.g., computer 602 of Figure 6. Method 200 of Figure 2 will be described with continued reference to schematic diagram of the automatic imaging protocoling system of Figure 1.

[0023] At step 202 of Figure 2, patient data is received from one or more patient databases. In one example, as shown in schematic diagram 100, the patient data may be retrieved by data collection and preprocessing module 104 from one or more patient databases 102.

[0024] The patient data may comprise any suitable data of a patient. For example, the patient data may comprise text-based data of the patient, such as, e.g., demographic information, vital signs, medical history, family history, laboratory results, medications, measurements and information extracted from medical images, etc. The text-based data of the patient may comprise structured data organized into specific fields each having a defined purpose and / or unstructured data. In another example, the patient data maycomprise imaging data of the patient, such as, e.g., medical images, scanned notes, etc. The medical images may be of any suitable modality or combination of modalities, such as, e.g., CT (computed tomography), MRI (magnetic resonance imaging), US (ultrasound), x-ray, etc. The medical images may be 2D (two dimensional) images and / or 3D (three dimensional) volumes. The one or more patient databases may comprise, for example, an EHR (electronic health record), EMR (electronic medical record), PHR (personal health record), HIS (health information system), RIS (radiology information system), PACS (picture archiving and communication system), LIMS (laboratory information management system), or any other suitable database or system storing patient data.

[0025] In one embodiment, the patient data is retrieved from the one or more patient databases using a first LLM based on prompted API (application programming interface) requests. The first LLM receives as input one or more prompts comprising instructions to output a function call sequence (e.g., “Current Medication:”) and generates as output the function call sequence. The instructions convey at least 1 ) the task (e.g., to collect and summarize the information from the patient databases that is relevant to image protocoling) and 2) the usable API function call sequences and what information will be returned by them. The LLM proceeds by outputting, as appropriate, the function call sequences (which get replaced with the API returned data). In response to the function call sequence, one or more API calls are made to the one or more patient databases to retrieve the patient data at the fields of the patient databases corresponding to the function call sequence. This enables the first LLM to determine which information is needed for a given scenario dynamically based on patient data previously retrieved. Once the patient data has been retrieved via the API, the first LLM would be further prompted to summarize the patient data for image protocoling, after which the first LLM would output the summary. In another embodiment, the patient data is retrieved from the one or more patient databases based on available APIs of the one or more patient databases. Available APIs of the one or more patient databases are comprehensively called to retrieve extensive patient data from the one or more patient databases.

[0026] The patient data may be received directly by, for example, loading patient data a storage or memory of the one or more patient databases (e.g., memory 610 or storage 612 of computer 602 of Figure 6) and / or receiving patient data via a network interface of one or more remote patient databases (e.g., network interface 606 of computer 602 of Figure 6).

[0027] At step 204 of Figure 2, imaging protocoling information is extracted from the patient data using a first machine learning model. In one example, as shown in schematic diagram 100, the imaging protocoling information is extracted from the patient data by data collection and preprocessing module 104.

[0028] The imaging protocoling information may comprise any suitable information relating to imaging protocoling. For example, the imaging protocoling information may comprise indications for a scan, imaging history, patient history, allergies, etc. In one embodiment, the imaging protocoling information is a summary of pertinent information relating to imaging protocoling in the patient data.

[0029] The first machine learning model may be any suitable machine learning based model. In one embodiment, the first machine learning model is the first LLM. The first LLM receives as input one or more prompts comprising the patient data and instructions for extracting (e.g., summarizing) the imaging protocoling information from the patient data and generates as output the extracted imaging protocoling information. The one or more prompts may be input to the first LLM, e.g., manually by a user or from a computing system (e.g., computer 602 of Figure 6) via one or more APIs. The patient data may comprise unstructured patient data input to the first LLM via the one or more prompts with context tabs (e.g., denoting the type, time, and origin of the patient data) and / or may comprise structured patient data input to the first LLM via the one or more prompts as prose with structure tags (e.g., the patient’s age is <age>).

[0030] The first LLM may be any suitable pre-trained deep learning based LLM. For example, the LLM may be based on the transformer architecture, which uses a selfattention mechanism to capture long-range dependencies in text. One example of a transformer-based architecture is GPT (generative pre-training transformer), which has a multilayer transformer decoder architecture that may be pretrained to optimize the nexttoken prediction task and then fine-tuned with labelled data for various downstream tasks. GPT-based LLMs may be trained using reinforcement learning with human feedback for performing various natural language processing tasks. Other exemplary transformerbased architectures include BLOOM (BigScience Large Open-science Open-access Multilingual Language Model) and BERT (Bidirectional Encoder Representations from Transformers). The first LLM may be a multi-modal LLM to receive the text-based data of the patient and the imaging data of the patient. In one embodiment, the first LLM is a non-task-specific LLM. In another embodiment, the first LLM is fine-tuned for the task of extracting imaging protocoling information from patient data using, e.g., human-in-the- loop training.

[0031] At step 206 of Figure 2, an imaging protocol template is selected from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model. Each of the plurality of candidate imaging protocol templates comprises a set of predefined protocols for acquiring medical images of the patient. Exemplary protocols may include the imaging modality, the anatomical region of interest, and the acquisition parameters for acquiring the medical images (e.g., acquisition protocol, field of view, slice thickness, appropriate coil in MRIs, reconstruction kernel in CT, etc.). In one example, as shown in schematic diagram 100, the imaging protocol template is selected by protocol template selection module 106.

[0032] The second machine learning model may be any suitable machine learning based model (e.g., an encoder network). In one embodiment, the second machine learning model is a second LLM. The second LLM receives as input one or more prompts comprising the extracted imaging protocoling information, descriptions of the plurality of candidate imaging protocol templates, and optionally other patient data (e.g., patient history records) and generates as output a candidate imaging protocol template that is most similar to the extracted imaging protocoling information (and optionally the other patient data) as the selected imaging protocol template. The one or more prompts may be input to the second LLM, e.g., manually by a user or from a computing system (e.g., computer 602 of Figure 6) via one or more APIs. The second LLM may be any suitable pre-trained deep learning based encoder-based LLM. For example, the second LLM maybe based on the transformer architecture, such as, e.g., GPT, BLOOM, BERT, etc. The second LLM may be a multi-modal LLM to receive the text-based data of the patient and the imaging data of the patient.

[0033] The second LLM determines the most similar candidate imaging protocol template based on a semantic similarity, which measures how closely the text of the extracted imaging protocoling information (and optionally the other patient data) and the descriptions of the plurality of candidate imaging protocol templates are in terms of meaning. For example, a description of a candidate imaging protocol template including “headache” and extracted imaging protocoling information comprising an indication for scan including “migraines” would be semantically similar. In one embodiment, the second LLM encodes the text of the extracted imaging protocoling information (and optionally the other patient data) and the descriptions of the plurality of candidate imaging protocol templates into respective vectors. A similarity measure is then determined between the vector of the text of the extracted imaging protocoling information (and optionally the other patient data) and each of the vectors of the descriptions of the plurality of candidate imaging protocol templates. The similarity measure may be, for example, cosine similarity or any suitable measure of similarity or distance. The candidate imaging protocol template having a vector that is most similar to the vector of the extracted imaging protocoling information (and optionally the other patient data) is then selected.

[0034] At step 208 of Figure 2, the selected imaging protocol template is fine-tuned based on the patient data using a third machine learning model. In one example, as shown in schematic diagram 100, the selected imaging protocol template is fine-tuned by protocol refinement module 110.

[0035] In one embodiment, the patient data for fine-tuning the selected imaging protocol template comprises any patient data relevant to the choice or modification of imaging protocols, such as, e.g., allergies, contraindications, comorbidities, etc. For example, some patients may not be able to receive contrast agents or have metal implants that require particular precautions. The selected imaging protocol template may additionally or alternatively be fine-tuned based on protocoling standards for a clinical site (e.g., a hospital). The protocoling standards comprise a set of rules or preferencesdefining how protocols should be, e.g., named, formatted, customized for different scenarios, etc. for the clinical site. For example, some hospitals may prefer shorter scan times or lower contrast doses than others.

[0036] The third machine learning model may be any suitable machine learning based model. In one embodiment, the third machine learning model is a third LLM. The third LLM receives as input one or more prompts comprising the selected imaging protocol template, the patient data, and / or the clinical standards for a clinical site and generates as output the fine-tuned imaging protocol template. The one or more prompts may be input to the third LLM, e.g., manually by a user or from a computing system (e.g., computer 602 of Figure 6) via one or more APIs. The third LLM may be any suitable pretrained deep learning based LLM. For example, the third LLM may be based on the transformer architecture, such as, e.g., GPT, BLOOM, BERT, etc. The third LLM may be a multi-modal LLM to receive the text-based data of the patient and the imaging data of the patient.

[0037] In one embodiment, the fine-tuning may be encoded in a decision forest with free text queries at each node. For example, one tree in the decision forest may query if the patient has renal insufficiency or contrast allergies. If yes, an endogenous contrast is used, otherwise an injected contrast is used. The third LLM would be prompted to answer the queries and follow the decision tree accordingly to refine the imaging protocol template. By structuring the decision points as free text queries, the protocoling standard can be set or updated without the need for reprogramming.

[0038] At step 210 of Figure 2, the fine-tuned imaging protocol template is output. For example, the fine-tuned imaging protocol template can be output by displaying the finetuned imaging protocol template on a display device of a computer system (e.g., input / output device 608 of computer 602 of Figure 6), storing the fine-tuned imaging protocol template on a memory or storage of a computer system (e.g., memory 610 or storage 612 of computer 602 of Figure 6), or by transmitting the fine-tuned imaging protocol template to a remote computer system (e.g., computer 602 of Figure 6).

[0039] In one embodiment, for example in clinical workflows implementing just-in-time protocoling or just-in-time imaging, medical images are automatically or semi-automatically acquired based on the fine-tuned imaging protocol template. Some steps (e.g., contrast injection) may require user approval prior to being performed.

[0040] In one embodiment, optionally, the fine-tuned imaging protocol template may be output to a user interface (e.g., a display device) for review by a user (e.g., radiologist or another clinician). In one example, as shown in schematic diagram 100, the fine-tuned imaging protocol template is output to protocol review interface 112. The user interface may present the fine-tuned imaging protocol template annotated with quotes or summarized information from the patient data that justifies the selection of the protocols. The specific protocol choices made during the fine-tuning may also be shown with quotes or summarized information from the patient data that justifies the fine-tuning. A user could thereby quickly review and approve the protocol selection and fine-tuning or modify the fine-tuned imaging protocol template.

[0041] In one embodiment, optionally, the plurality of candidate imaging protocol templates may be automatically defined (e.g., during a prior preprocessing stage). In one example, as shown in schematic diagram 100, the plurality of candidate imaging protocol templates may be automatically defined by protocol template definition module 108. Data of various information systems and databases for a clinical site are analyzed to identify clusters of similar imaging protocols and the dimensions of variability within each cluster. For example, dimensions of variability within the similar imaging protocols describe protocol changes (e.g., the inclusion or exclusion of a particular sequence, the use of contrast). Dimensions of variability within the patient scenarios describe factors that might lead a user to choose one of the aforementioned protocol variants (e.g., patient weight, history of tinnitus, renal sufficiency). The clusters of similar imaging protocols and the dimensions of variability within each cluster may be used to automatically define candidate imaging protocol templates. Patient data of various patients is also analyzed to identify profiles of patient scenarios that trigger each cluster. Variability within the patient scenarios that trigger each cluster are then correlated with the dimensions of variability within that cluster. The profiles of patient scenarios that trigger each cluster may be used to automatically define candidate imaging protocol template descriptions. The patient scenarios correlated with the dimensions of variability may be used fordrafting protocoling standards that may be used or modified for use for fine-tuning imaging protocol templates (e.g., at step 208 of Figure 2). For example, there might be a protocol cluster that requires vascular imaging and within that cluster a correlation that patients with renal insufficiency would receive an ASL (arterial spin labeling) sequence while those with renal sufficiency would receive a contrasted scan. In the protocol selection step, the protocol with vascular imaging would be selected, and in the fine-tuning, either ASL or extrinsic contrast would be selected.

[0042] In one embodiment, the first LLM, the second LLM, and / or the third LLM are the same LLM. In another embodiment, the first LLM, the second LLM, and / or the third LLM are different LLMs (e.g., LLMs that are fine-tuned differently, prompted differently, etc.). The LLMs are trained during a prior offline or training stage using a large corpus of training data. Once trained, the trained LLMs are applied during an online or inference stage, e.g., to perform method 200 of Figure 2.

[0043] In one embodiment, where the one or more prompts input to the first LLM, the second LLM, and / or the third LLM exceeds the maximum size of their respective context window, a vector database may be utilized to store data of the one or more prompts as needed. Such data may be retrieved piecewise as needed within the maximum size of the context window. For example, the relevant data may be retrieved from the vector database, an analysis may be run on the retrieved relevant data by the respective LLM, the output is saved and the context window is discarded. This can be repeated as necessary.

[0044] Advantageously, embodiments described herein provide for automatic imaging protocoling, resulting in time savings for technicians and radiologists. Further, embodiments described herein utilize LLMs to extract information from free text natural language inputs, rather than structured inputs. This enables integration into existing clinical information systems without requiring consistent EHR formats and content, while also making the user interface more intuitive, reducing training costs. Additionally, the protocol template definition module allows for the definition of protocoling decision trees in an intuitive way without the need for programming. This reduces the burden on application specialists commissioning the system and clinicians. Further, embodimentsdescribed herein provide for a protocol review interface that justifies the protocoling decisions, making it understandable to the end user and more trustworthy.

[0045] Embodiments described herein are described with respect to the claimed systems as well as with respect to the claimed methods. Features, advantages or alternative embodiments herein can be assigned to the other claimed objects and vice versa. In other words, claims and embodiments for the systems can be improved with features described or claimed in the context of the respective methods. In this case, the functional features of the method are implemented by physical units of the system.

[0046] Furthermore, certain embodiments described herein are described with respect to methods and systems utilizing trained machine learning models, as well as with respect to methods and systems for providing trained machine learning models. Features, advantages or alternative embodiments herein can be assigned to the other claimed objects and vice versa. In other words, claims and embodiments for providing trained machine learning models can be improved with features described or claimed in the context of utilizing trained machine learning models, and vice versa. In particular, datasets used in the methods and systems for utilizing trained machine learning models can have the same properties and features as the corresponding datasets used in the methods and systems for providing trained machine learning models, and the trained machine learning models provided by the respective methods and systems can be used in the methods and systems for utilizing the trained machine learning models.

[0047] In general, a trained machine learning model mimics cognitive functions that humans associate with other human minds. In particular, by training based on training data the machine learning model is able to adapt to new circumstances and to detect and extrapolate patterns. Another term for “trained machine learning model” is “trained function.”

[0048] In general, parameters of a machine learning model can be adapted by means of training. In particular, supervised training, semi-supervised training, unsupervised training, reinforcement learning and / or active learning can be used. Furthermore, representation learning (an alternative term is “feature learning”) can be used. In particular, the parameters of the machine learning models can be adapted iteratively byseveral steps of training. In particular, within the training a certain cost function can be minimized. In particular, within the training of a neural network the backpropagation algorithm can be used.

[0049] In particular, a machine learning model, such as, e.g., the first machine learning model utilized at steps 202 and 204, the second machine learning model utilized at step 206, and the third machine learning model utilized at step 208 of Figure 2, can comprise, for example, a neural network, a support vector machine, a decision tree and / or a Bayesian network, and / or the machine learning model can be based on, for example, k- means clustering, Q-learning, genetic algorithms and / or association rules. In particular, a neural network can be, e.g., a deep neural network, a convolutional neural network or a convolutional deep neural network. Furthermore, a neural network can be, e.g., an adversarial network, a deep adversarial network and / or a generative adversarial network.

[0050] Figure 3 shows an embodiment of an artificial neural network 300 that may be used to implement one or more machine learning models described herein. Alternative terms for “artificial neural network” are “neural network”, “artificial neural net” or “neural net”.

[0051] The artificial neural network 300 comprises nodes 320, ... , 332 and edges 340, ... , 342, wherein each edge 340, ... , 342 is a directed connection from a first node 320, ... , 332 to a second node 320, ... , 332. In general, the first node 320, ... , 332 and the second node 320, ... , 332 are different nodes 320, ... , 332, it is also possible that the first node 320, ... , 332 and the second node 320, ... , 332 are identical. For example, in Figure 3 the edge 340 is a directed connection from the node 320 to the node 323, and the edge 342 is a directed connection from the node 330 to the node 332. An edge 340, ... , 342 from a first node 320, ... , 332 to a second node 320, ... , 332 is also denoted as “ingoing edge” for the second node 320, ... , 332 and as “outgoing edge” for the first node 320, ... , 332.

[0052] In this embodiment, the nodes 320, ... , 332 of the artificial neural network 300 can be arranged in layers 310, ... , 313, wherein the layers can comprise an intrinsic order introduced by the edges 340, ... , 342 between the nodes 320, ... , 332. In particular, edges 340, ... , 342 can exist only between neighboring layers of nodes. In the displayedembodiment, there is an input layer 310 comprising only nodes 320, ... , 322 without an incoming edge, an output layer 313 comprising only nodes 331 , 332 without outgoing edges, and hidden layers 311 , 312 in-between the input layer 310 and the output layer 313. In general, the number of hidden layers 311 , 312 can be chosen arbitrarily. The number of nodes 320, ... , 322 within the input layer 310 usually relates to the number of input values of the neural network, and the number of nodes 331 , 332 within the output layer 313 usually relates to the number of output values of the neural network.

[0053] In particular, a (real) number can be assigned as a value to every node 320, ... , 332 of the neural network 300. Here, x(n)i denotes the value of the i-th node 320, ... , 332 of the n-th layer 310, ... , 313. The values of the nodes 320, ... , 322 of the input layer 310 are equivalent to the input values of the neural network 300, the values of the nodes 331 , 332 of the output layer 313 are equivalent to the output value of the neural network 300. Furthermore, each edge 340, ... , 342 can comprise a weight being a real number, in particular, the weight is a real number within the interval [-1 , 1 ] or within the interval [0, 1 ], Here, w(m n)i,j denotes the weight of the edge between the i-th node 320, ... , 332 of the m-th layer 310, ... , 313 and the j-th node 320, ... , 332 of the n-th layer 310, ... , 313. Furthermore, the abbreviation w(n)ij is defined for the weight w(n n+1)i,j.

[0054] In particular, to calculate the output values of the neural network 300, the input values are propagated through the neural network. In particular, the values of the nodes 320, ... , 332 of the (n+1 )-th layer 310, ... , 313 can be calculated based on the values of the nodes 320, ... , 332 of the n-th layer 310, ... , 313 by

[0055] Herein, the function f is a transfer function (another term is “activation function”). Known transfer functions are step functions, sigmoid function (e.g., the logistic function, the generalized logistic function, the hyperbolic tangent, the Arctangent function, the error function, the smoothstep function) or rectifier functions. The transfer function is mainly used for normalization purposes.

[0056] In particular, the values are propagated layer-wise through the neural network, wherein values of the input layer 310 are given by the input of the neural network 300, wherein values of the first hid-den layer 311 can be calculated based on the values of the input layer 310 of the neural network, wherein values of the second hidden layer 312 can be calculated based in the values of the first hidden layer 311 , etc.

[0057] In order to set the valuesfor the edges, the neural network 300 has to be trained using training data. In particular, training data comprises training input data and training output data (denoted as tj). For a training step, the neural network 300 is applied to the training input data to generate calculated output data. In particular, the training data and the calculated output data comprise a number of values, said number being equal with the number of nodes of the output layer.

[0058] In particular, a comparison between the calculated output data and the training data is used to recursively adapt the weights within the neural network 300 (backpropagation algorithm). In particular, the weights are changed according towherein y is a learning rate, and the numbers 5(n)j can be recursively calculated asbased on 5(n+1)j, if the (n+1 )-th layer is not the output layer, andif the (n+1 )-th layer is the output layer 313, wherein f’ is the first derivative of the activation function, and t(n+1)j is the comparison training value for the j-th node of the output layer 313.

[0059] A convolutional neural network is a neural network that uses a convolution operation instead general matrix multiplication in at least one of its layers (so-called “convolutional layer”). In particular, a convolutional layer performs a dot product of one ormore convolution kernels with the convolutional layer's input data I image, wherein the entries of the one or more convolution kernel are the parameters or weights that are adapted by training. In particular, one can use the Frobenius inner product and the ReLLI activation function. A convolutional neural network can comprise additional layers, e.g., pooling layers, fully connected layers, and normalization layers.

[0060] By using convolutional neural networks input images can be processed in a very efficient way, because a convolution operation based on different kernels can extract various image features, so that by adapting the weights of the convolution kernel the relevant image features can be found during training. Furthermore, based on the weightsharing in the convolutional kernels less parameters need to be trained, which prevents overfitting in the training phase and allows to have faster training or more layers in the network, improving the performance of the network.

[0061] Figure 4 shows an embodiment of a convolutional neural network 400 that may be used to implement one or more machine learning models described herein. In the displayed embodiment, the convolutional neural network comprises 400 an input node layer 410, a convolutional layer 411 , a pooling layer 413, a fully connected layer 414 and an output node layer 416, as well as hidden node layers 412, 414. Alternatively, the convolutional neural network 400 can comprise several convolutional layers 411 , several pooling layers 413 and several fully connected layers 415, as well as other types of layers. The order of the layers can be chosen arbitrarily, usually fully connected layers 415 are used as the last layers before the output layer 416.

[0062] In particular, within a convolutional neural network 400 nodes 420, 422, 424 of a node layer 410, 412, 414 can be considered to be arranged as a d-dimensional matrix or as a d-dimensional image. In particular, in the two-dimensional case the value of the node 420, 422, 424 indexed with i and j in the n-th node layer 410, 412, 414 can be denoted as x(n)[i, j]. However, the arrangement of the nodes 420, 422, 424 of one node layer 410, 412, 414 does not have an effect on the calculations executed within the convolutional neural network 400 as such, since these are given solely by the structure and the weights of the edges.

[0063] A convolutional layer 411 is a connection layer between an anterior node layer410 (with node values x(n-1 )) and a posterior node layer 412 (with node values x(n)). In particular, a convolutional layer 411 is characterized by the structure and the weights of the incoming edges forming a convolution operation based on a certain number of kernels. In particular, the structure and the weights of the edges of the convolutional layer411 are chosen such that the values x(n) of the nodes 422 of the posterior node layer 412 are calculated as a convolution x(n) = K * x(n-1 ) based on the values x(n-1 ) of the nodes 420 anterior node layer 410, where the convolution * is defined in the two-dimensional case as

[0064] Here the kernel K is a d-dimensional matrix (in this embodiment, a two- dimensional matrix), which is usually small compared to the number of nodes 420, 422 (e.g., a 3x3 matrix, or a 5x5 matrix). In particular, this implies that the weights of the edges in the convolution layer 411 are not independent, but chosen such that they produce said convolution equation. In particular, for a kernel being a 3x3 matrix, there are only 9 independent weights (each entry of the kernel matrix corresponding to one independent weight), irrespectively of the number of nodes 420, 422 in the anterior node layer 410 and the posterior node layer 412.

[0065] In general, convolutional neural networks 400 use node layers 410, 412, 414 with a plurality of channels, in particular, due to the use of a plurality of kernels in convolutional layers 411. In those cases, the node layers can be considered as (d+1 )- dimensional matrices (the first dimension indexing the channels). The action of a convolutional layer 411 is then a two-dimensional example defined aswhere x(n l)acorresponds to the a-th channel of the anterior node layer 410, x(n)t corresponds to the b-th channel of the posterior node layer 412 and Ka bcorresponds to one of the kernels. If a convolutional layer 411 acts on an anterior node layer 410 with Achannels and outputs a posterior node layer 412 with B channels, there are A B independent d-dimensional kernels Ka b.

[0066] In general, in convolutional neural networks 400 activation functions are used. In this embodiment re ReLU (acronym for “Rectified Linear Units”) is used, with R(z) = max(0, z), so that the action of the convolutional layer 411 in the two-dimensional example is

[0067] It is also possible to use other activation functions, e.g., ELU (acronym for “Exponential Linear Unit”), LeakyReLU, Sigmoid, Tanh or Softmax.

[0068] In the displayed embodiment, the input layer 410 comprises 36 nodes 420, arranged as a two-dimensional 6x6 matrix. The first hidden node layer 412 comprises 72 nodes 422, arranged as two two-dimensional 6x6 matrices, each of the two matrices being the result of a convolution of the values of the input layer with a 3x3 kernel within the convolutional layer 411 . Equivalently, the nodes 422 of the first hidden node layer 412 can be interpreted as arranged as a three-dimensional 2x6x6 matrix, wherein the first dimension correspond to the channel dimension.

[0069] The advantage of using convolutional layers 411 is that spatially local correlation of the input data can exploited by enforcing a local connectivity pattern between nodes of adjacent layers, in particular by each node being connected to only a small region of the nodes of the preceding layer.

[0070] A pooling layer 413 is a connection layer between an anterior node layer 412 (with node values x(n-1 )) and a posterior node layer 414 (with node values x(n)). In particular, a pooling layer 413 can be characterized by the structure and the weights of the edges and the activation function forming a pooling operation based on a non-linear pooling function f. For example, in the two-dimensional case the values x(n) of the nodes 424 of the posterior node layer 414 can be calculated based on the values x(n-1 ) of the nodes 422 of the anterior node layer 412 as

[0071] In other words, by using a pooling layer 413 the number of nodes 422, 424 can be reduced, by re-placing a number d1 d2 of neighboring nodes 422 in the anterior node layer 412 with a single node 422 in the posterior node layer 414 being calculated as a function of the values of said number of neighboring nodes. In particular, the pooling function f can be the max-function, the average or the L2-Norm. In particular, for a pooling layer 413 the weights of the incoming edges are fixed and are not modified by training.

[0072] The advantage of using a pooling layer 413 is that the number of nodes 422, 424 and the number of parameters is reduced. This leads to the amount of computation in the network being reduced and to a control of overfitting.

[0073] In the displayed embodiment, the pooling layer 413 is a max-pooling layer, replacing four neighboring nodes with only one node, the value being the maximum of the values of the four neighboring nodes. The max-pooling is applied to each d-dimensional matrix of the previous layer; in this embodiment, the max-pooling is applied to each of the two two-dimensional matrices, reducing the number of nodes from 72 to 18.

[0074] In general, the last layers of a convolutional neural network 400 are fully connected layers 415. A fully connected layer 415 is a connection layer between an anterior node layer 414 and a posterior node layer 416. A fully connected layer 413 can be characterized by the fact that a majority, in particular, all edges between nodes 414 of the anterior node layer 414 and the nodes 416 of the posterior node layer are present, and wherein the weight of each of these edges can be adjusted individually.

[0075] In this embodiment, the nodes 424 of the anterior node layer 414 of the fully connected layer 415 are displayed both as two-dimensional matrices, and additionally as non-related nodes (indicated as a line of nodes, wherein the number of nodes was reduced for a better presentability). This operation is also denoted as “flattening”. In this embodiment, the number of nodes 426 in the posterior node layer 416 of the fully connected layer 415 smaller than the number of nodes 424 in the anterior node layer 414. Alternatively, the number of nodes 426 can be equal or larger.

[0076] Furthermore, in this embodiment the Softmax activation function is used within the fully connected layer 415. By applying the Softmax function, the sum the values of allnodes 426 of the output layer 416 is 1 , and all values of all nodes 426 of the output layer 416 are real numbers between 0 and 1. In particular, if using the convolutional neural network 400 for categorizing input data, the values of the output layer 416 can be interpreted as the probability of the input data falling into one of the different categories.

[0077] In particular, convolutional neural networks 400 can be trained based on the backpropagation algorithm. For preventing overfitting, methods of regularization can be used, e.g., dropout of nodes 420, ... , 424, stochastic pooling, use of artificial data, weight decay based on the L1 or the L2 norm, or max norm constraints.

[0078] According to an aspect, the machine learning model may comprise one or more residual networks (ResNet). In particular, a ResNet is an artificial neural network comprising at least one jump or skip connection used to jump over at least one layer of the artificial neural network. In particular, a ResNet may be a convolutional neural network comprising one or more skip connections respectively skipping one or more convolutional layers. According to some examples, the ResNets may be represented as m-layer ResNets, where m is the number of layers in the corresponding architecture and, according to some examples, may take values of 34, 50, 101 , or 152. According to some examples, such an m-layer ResNet may respectively comprise (m-2) / 2 skip connections.

[0079] A skip connection may be seen as a bypass which directly feeds the output of one preceding layer over one or more bypassed layers to a layer succeeding the one or more bypassed layers. Instead of having to directly fit a desired mapping, the bypassed layers would then have to fit a residual mapping “balancing” the directly fed output.

[0080] Fitting the residual mapping is computationally easier to optimize than the directed mapping. What is more, this alleviates the problem of vanishing / exploding gradients during optimization upon training the machine learning models: if a bypassed layer runs into such problems, its contribution may be skipped by regularization of the directly fed output. Using ResNets thus brings about the advantage that much deeper networks may be trained.

[0081] In particular, a recurrent machine learning model is a machine learning model whose output does not only depend on the input value and the parameters of the machine learning model adapted by the training process, but also on a hidden state vector, whereinthe hidden state vector is based on previous inputs used on for the recurrent machine learning model. In particular, the recurrent machine learning model can comprise additional storage states or additional structures that incorporate time delays or comprise feedback loops.

[0082] In particular, the underlying structure of a recurrent machine learning model can be a neural network, which can be denoted as recurrent neural network. Such a recurrent neural network can be described as an artificial neural network where connections between nodes form a directed graph along a temporal sequence. In particular, a recurrent neural network can be interpreted as directed acyclic graph. In particular, the recurrent neural network can be a finite impulse recurrent neural network or an infinite impulse recurrent neural network (wherein a finite impulse network can be unrolled and replaced with a strictly feedforward neural network, and an infinite impulse network cannot be unrolled and replaced with a strictly feedforward neural network).

[0083] In particular, training a recurrent neural network can be based on the BPTT algorithm (acronym for“backpropagation through time”), on the RTRL algorithm (acronym for “real-time recurrent learning”) and / or on genetic algorithms.

[0084] By using a recurrent machine learning model input data comprising sequences of variable length can be used. In particular, this implies that the method cannot be used only for a fixed number of input datasets (and needs to be trained differently for every other number of input datasets used as input), but can be used for an arbitrary number of input datasets. This implies that the whole set of training data, independent of the number of input datasets contained in different sequences, can be used within the training, and that training data is not reduced to training data corresponding to a certain number of successive input datasets.

[0085] Fig. 5 shows the schematic structure of a recurrent machine learning model F, both in a recurrent representation 502 and in an unfolded representation 504, that may be used to implement one or more machine learning models described herein. The recurrent machine learning model takes as input several input datasets x, xi, ... , XN 506 and creates a corresponding set of output datasets y, yi, ... , yN 508. Furthermore, the output depends on a so-called hidden vector h, hi, ... , hN 510, which implicitly comprisesinformation about input datasets previously used as input for the recurrent machine learning model F 512. By using these hidden vectors h, hi, ... , hN 510, a sequentiality of the input datasets can be leveraged.

[0086] In a single step of the processing, the recurrent machine learning model F 512 takes as input the hidden vector hn-1 created within the previous step and an input dataset Xn. Within this step, the recurrent machine learning model F generates as output an updated hidden vector hn and an output dataset yn. In other words, one step of processing calculates (yn, hn) = F(xn, hn-1 ), or by splitting the recurrent machine learning model F 512 into a part F(y) calculating the output data and F(h) calculating the hidden vector, one step of processing calculates yn= F(y)(xn, hn-i) and hn = F(h)(xn, hn-i). For the first processing step, ho can be chosen randomly or filled with all entries being zero. The parameters of the recurrent machine learning model F 512 that were trained based on training datasets before do not change between the different processing steps.

[0087] In particular, the output data and the hidden vector of a processing step depend on all the previous input datasets used in the previous steps. yn= F(y)(xn, F(h)(xn-i, hn-2))hn-2)).

[0088] Systems, apparatuses, and methods described herein may be implemented using digital circuitry, or using one or more computers using well-known computer processors, memory units, storage devices, computer software, and other components. Typically, a computer includes a processor for executing instructions and one or more memories for storing instructions and data. A computer may also include, or be coupled to, one or more mass storage devices, such as one or more magnetic disks, internal hard disks and removable disks, magneto-optical disks, optical disks, etc.

[0089] Systems, apparatuses, and methods described herein may be implemented using computers operating in a client-server relationship. Typically, in such a system, the client computers are located remotely from the server computer and interact via a network. The client-server relationship may be defined and controlled by computer programs running on the respective client and server computers.

[0090] Systems, apparatuses, and methods described herein may be implemented within a network-based cloud computing system. In such a network-based cloudcomputing system, a server or another processor that is connected to a network communicates with one or more client computers via a network. A client computer may communicate with the server via a network browser application residing and operating on the client computer, for example. A client computer may store data on the server and access the data via the network. A client computer may transmit requests for data, or requests for online services, to the server via the network. The server may perform requested services and provide data to the client computer(s). The server may also transmit data adapted to cause a client computer to perform a specified function, e.g., to perform a calculation, to display specified data on a screen, etc. For example, the server may transmit a request adapted to cause a client computer to perform one or more of the steps or functions of the methods and workflows described herein, including one or more of the steps or functions of Figures 1 or 2. Certain steps or functions of the methods and workflows described herein, including one or more of the steps or functions of Figures 1 or 2, may be performed by a server or by another processor in a network-based cloudcomputing system. Certain steps or functions of the methods and workflows described herein, including one or more of the steps of Figures 1 or 2, may be performed by a client computer in a network-based cloud computing system. The steps or functions of the methods and workflows described herein, including one or more of the steps of Figures 1 or 2, may be performed by a server and / or by a client computer in a network-based cloud computing system, in any combination.

[0091] Systems, apparatuses, and methods described herein may be implemented using a computer program product tangibly embodied in an information carrier, e.g., in a non-transitory machine-readable storage device, for execution by a programmable processor; and the method and workflow steps described herein, including one or more of the steps or functions of Figures 1 or 2, may be implemented using one or more computer programs that are executable by such a processor. A computer program is a set of computer program instructions that can be used, directly or indirectly, in a computer to perform a certain activity or bring about a certain result. A computer program can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand-alone program or asa module, component, subroutine, or other unit suitable for use in a computing environment.

[0092] A high-level block diagram of an example computer 602 that may be used to implement systems, apparatuses, and methods described herein is depicted in Figure 6. Computer 602 includes a processor 604 operatively coupled to a data storage device 612 and a memory 610. Processor 604 controls the overall operation of computer 602 by executing computer program instructions that define such operations. The computer program instructions may be stored in data storage device 612, or other computer readable medium, and loaded into memory 610 when execution of the computer program instructions is desired. Thus, the method and workflow steps or functions of Figures 1 or 2 can be defined by the computer program instructions stored in memory 610 and / or data storage device 612 and controlled by processor 604 executing the computer program instructions. For example, the computer program instructions can be implemented as computer executable code programmed by one skilled in the art to perform the method and workflow steps or functions of Figures 1 or 2. Accordingly, by executing the computer program instructions, the processor 604 executes the method and workflow steps or functions of Figures 1 or 2. Computer 602 may also include one or more network interfaces 606 for communicating with other devices via a network. Computer 602 may also include one or more input / output devices 608 that enable user interaction with computer 602 (e.g., display, keyboard, mouse, speakers, buttons, etc.).

[0093] Processor 604 may include both general and special purpose microprocessors, and may be the sole processor or one of multiple processors of computer 602. Processor 604 may include one or more central processing units (CPUs), for example. Processor 604, data storage device 612, and / or memory 610 may include, be supplemented by, or incorporated in, one or more application-specific integrated circuits (ASICs) and / or one or more field programmable gate arrays (FPGAs).

[0094] Data storage device 612 and memory 610 each include a tangible non- transitory computer readable storage medium. Data storage device 612, and memory 610, may each include high-speed random access memory, such as dynamic random access memory (DRAM), static random access memory (SRAM), double data ratesynchronous dynamic random access memory (DDR RAM), or other random access solid state memory devices, and may include non-volatile memory, such as one or more magnetic disk storage devices such as internal hard disks and removable disks, magnetooptical disk storage devices, optical disk storage devices, flash memory devices, semiconductor memory devices, such as erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM), digital versatile disc read-only memory (DVD-ROM) disks, or other non-volatile solid state storage devices.

[0095] Input / output devices 608 may include peripherals, such as a printer, scanner, display screen, etc. For example, input / output devices 608 may include a display device such as a cathode ray tube (CRT) or liquid crystal display (LCD) monitor for displaying information to the user, a keyboard, and a pointing device such as a mouse or a trackball by which the user can provide input to computer 602.

[0096] An image acquisition device 614 can be connected to the computer 602 to input image data (e.g., medical images) to the computer 602. It is possible to implement the image acquisition device 614 and the computer 602 as one device. It is also possible that the image acquisition device 614 and the computer 602 communicate wirelessly through a network. In a possible embodiment, the computer 602 can be located remotely with respect to the image acquisition device 614.

[0097] Any or all of the systems, apparatuses, and methods discussed herein may be implemented using one or more computers such as computer 602.

[0098] One skilled in the art will recognize that an implementation of an actual computer or computer system may have other structures and may contain other components as well, and that Figure 6 is a high-level representation of some of the components of such a computer for illustrative purposes.

[0099] Independent of the grammatical term usage, individuals with male, female or other gender identities are included within the term.

[0100] The foregoing Detailed Description is to be understood as being in every respect illustrative and exemplary, but not restrictive, and the scope of the invention disclosed herein is not to be determined from the Detailed Description, but rather fromthe claims as interpreted according to the full breadth permitted by the patent laws. It is to be understood that the embodiments shown and described herein are only illustrative of the principles of the present invention and that various modifications may be implemented by those skilled in the art without departing from the scope and spirit of the invention. Those skilled in the art could implement various other feature combinations without departing from the scope and spirit of the invention.

[0101] The following is a list of non-limiting illustrative embodiments disclosed herein:

[0102] Illustrative embodiment 1. A computer-implemented method comprising: receiving patient data from one or more patient databases; extracting imaging protocoling information from the patient data using a first machine learning model; selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model; fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model; and outputting the fine-tuned imaging protocol template.

[0103] Illustrative embodiment 2. The computer-implemented method of illustrative embodiment 1 , wherein selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model comprises: encoding text of the extracted imaging protocoling information and descriptions of the plurality of candidate imaging protocol templates into vectors; determining a similarity measure between the vector of the text of the extracted imaging protocoling information and each of the vectors of the descriptions of the plurality of candidate imaging protocol templates; and selecting a candidate imaging protocol template having a vector most similar to the vector of the of the extracted imaging protocoling information based on the similarity measures. Illustrative embodiment 3. The computer-implemented method of any one of illustrative embodiments 1 -2, wherein fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model comprises: fine-tuning the selected imaging protocol template based on the patient data and protocoling standards for a clinical site using the third machine learning model.

[0104] Illustrative embodiment 4. The computer-implemented method of any one of illustrative embodiments 1 -3, wherein fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model comprises: fine-tuning the selected imaging protocol template according to a decision forest.

[0105] Illustrative embodiment 5. The computer-implemented method of any one of illustrative embodiments 1 -4, further comprising: identifying clusters of imaging protocols in systems of one or more clinical sites; identifying patient scenarios that trigger the clusters; and automatically generating one or more of the plurality of candidate imaging protocol templates based on the identified patient scenarios.

[0106] Illustrative embodiment 6. The computer-implemented method of any one of illustrative embodiments 1 -5, wherein outputting the fine-tuned imaging protocol template comprises: presenting the fine-tuned selected imaging protocol template annotated with information from the patient data justifying selection of protocols and justifying the fine- tuning.

[0107] Illustrative embodiment 7. The computer-implemented method of any one of illustrative embodiments 1 -6, wherein extracting imaging protocoling information from the patient data using a first machine learning model comprises: summarizing information relating to imaging protocoling in the patient data using the first machine learning model.

[0108] Illustrative embodiment 8. The computer-implemented method of any one of illustrative embodiments 1 -7, wherein receiving patient data from one or more patient databases comprises: retrieving the patient data from the one or more patient databases using the first machine learning model.

[0109] Illustrative embodiment 9. The computer-implemented method of any one of illustrative embodiments 1 -8, wherein the first machine learning model comprises a first LLM (large language model), the second machine learning model comprises a second LLM, and the third machine learning model comprises a third LLM.

[0110] Illustrative embodiment 10. An apparatus comprising: means for receiving patient data from one or more patient databases; means for extracting imaging protocoling information from the patient data using a first machine learning model; means for selecting an imaging protocol template from a plurality of candidate imaging protocoltemplates based on the extracted imaging protocoling information using a second machine learning model; means for fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model; and means for outputting the fine-tuned imaging protocol template.

[0111] Illustrative embodiment 11. The apparatus of illustrative embodiment 10, wherein the means for selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model comprises: means for encoding text of the extracted imaging protocoling information and descriptions of the plurality of candidate imaging protocol templates into vectors; means for determining a similarity measure between the vector of the text of the extracted imaging protocoling information and each of the vectors of the descriptions of the plurality of candidate imaging protocol templates; and means for selecting a candidate imaging protocol template having a vector most similar to the vector of the of the extracted imaging protocoling information based on the similarity measures.

[0112] Illustrative embodiment 12. The apparatus of any one of illustrative embodiments 10-11 , wherein the means for fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model comprises: means for fine-tuning the selected imaging protocol template based on the patient data and protocoling standards for a clinical site using the third machine learning model.

[0113] Illustrative embodiment 13. The apparatus of illustrative embodiments 10-12, wherein the means for fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model comprises: means for fine-tuning the selected imaging protocol template according to a decision forest.

[0114] Illustrative embodiment 14. The apparatus of illustrative embodiments 10-13, further comprising: means for identifying clusters of imaging protocols in systems of one or more clinical sites; means for identifying patient scenarios that trigger the clusters; and means for automatically generating one or more of the plurality of candidate imaging protocol templates based on the identified patient scenarios.

[0115] Illustrative embodiment 15. A non-transitory computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out operations comprising: receiving patient data from one or more patient databases; extracting imaging protocoling information from the patient data using a first machine learning model; selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model; fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model; and outputting the fine-tuned imaging protocol template.

[0116] Illustrative embodiment 16. The non-transitory computer-readable storage medium of illustrative embodiment 15, wherein selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model comprises: encoding text of the extracted imaging protocoling information and descriptions of the plurality of candidate imaging protocol templates into vectors; determining a similarity measure between the vector of the text of the extracted imaging protocoling information and each of the vectors of the descriptions of the plurality of candidate imaging protocol templates; and selecting a candidate imaging protocol template having a vector most similar to the vector of the of the extracted imaging protocoling information based on the similarity measures.

[0117] Illustrative embodiment 17. The non-transitory computer-readable storage medium of illustrative embodiments 15-16, wherein outputting the fine-tuned imaging protocol template comprises: presenting the fine-tuned selected imaging protocol template annotated with information from the patient data justifying selection of protocols and justifying the fine-tuning.

[0118] Illustrative embodiment 18. The non-transitory computer-readable storage medium of illustrative embodiments 15-17, wherein extracting imaging protocoling information from the patient data using a first machine learning model comprises: summarizing information relating to imaging protocoling in the patient data using the first machine learning model.

[0119] Illustrative embodiment 19. The non-transitory computer-readable storage medium of illustrative embodiments 15-18, wherein receiving patient data from one or more patient databases comprises: retrieving the patient data from the one or more patient databases using the first machine learning model.

[0120] Illustrative embodiment 20. The non-transitory computer-readable storage medium of illustrative embodiments 15-19, wherein the first machine learning model comprises a first LLM (large language model), the second machine learning model comprises a second LLM, and the third machine learning model comprises a third LLM.

Claims

CLAIMS:1 . A computer-implemented method comprising: receiving (202) patient data from one or more patient databases (102); extracting (204) imaging protocoling information from the patient data using a first machine learning model; selecting (206) an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model; fine-tuning (208) the selected imaging protocol template based on the patient data using a third machine learning model; and outputting (210) the fine-tuned imaging protocol template.

2. The computer-implemented method of claim 1 , wherein selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model comprises: encoding text of the extracted imaging protocoling information and descriptions of the plurality of candidate imaging protocol templates into vectors; determining a similarity measure between the vector of the text of the extracted imaging protocoling information and each of the vectors of the descriptions of the plurality of candidate imaging protocol templates; and selecting a candidate imaging protocol template having a vector most similar to the vector of the of the extracted imaging protocoling information based on the similarity measures.

3. The computer-implemented method of claim 1 , wherein fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model comprises:fine-tuning the selected imaging protocol template based on the patient data and protocoling standards for a clinical site using the third machine learning model.

4. The computer-implemented method of claim 1 , wherein fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model comprises: fine-tuning the selected imaging protocol template according to a decision forest.

5. The computer-implemented method of claim 1 , further comprising: identifying clusters of imaging protocols in systems of one or more clinical sites; identifying patient scenarios that trigger the clusters; and automatically generating one or more of the plurality of candidate imaging protocol templates based on the identified patient scenarios.

6. The computer-implemented method of claim 1 , wherein outputting the finetuned imaging protocol template comprises: presenting the fine-tuned selected imaging protocol template annotated with information from the patient data justifying selection of protocols and justifying the fine- tuning.

7. The computer-implemented method of claim 1 , wherein extracting imaging protocoling information from the patient data using a first machine learning model comprises: summarizing information relating to imaging protocoling in the patient data using the first machine learning model.

8. The computer-implemented method of claim 1 , wherein receiving patient data from one or more patient databases comprises: retrieving the patient data from the one or more patient databases using the first machine learning model.

9. The computer-implemented method of claim 1 , wherein the first machine learning model comprises a first LLM (large language model), the second machine learning model comprises a second LLM, and the third machine learning model comprises a third LLM.

10. An apparatus comprising: means for receiving (202) patient data from one or more patient databases (102); means for extracting (204) imaging protocoling information from the patient data using a first machine learning model; means for selecting (206) an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model; means for fine-tuning (208) the selected imaging protocol template based on the patient data using a third machine learning model; and means for outputting (210) the fine-tuned imaging protocol template.

11. The apparatus of claim 10, wherein the means for selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model comprises: means for encoding text of the extracted imaging protocoling information and descriptions of the plurality of candidate imaging protocol templates into vectors; means for determining a similarity measure between the vector of the text of the extracted imaging protocoling information and each of the vectors of the descriptions of the plurality of candidate imaging protocol templates; and means for selecting a candidate imaging protocol template having a vector most similar to the vector of the of the extracted imaging protocoling information based on the similarity measures.

12. The apparatus of claim 10, wherein the means for fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model comprises: means for fine-tuning the selected imaging protocol template based on the patient data and protocoling standards for a clinical site using the third machine learning model.

13. The apparatus of claim 10, wherein the means for fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model comprises: means for fine-tuning the selected imaging protocol template according to a decision forest.

14. The apparatus of claim 10, further comprising: means for identifying clusters of imaging protocols in systems of one or more clinical sites; means for identifying patient scenarios that trigger the clusters; and means for automatically generating one or more of the plurality of candidate imaging protocol templates based on the identified patient scenarios.

15. A non-transitory computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out operations comprising: receiving (202) patient data from one or more patient databases (102); extracting (204) imaging protocoling information from the patient data using a first machine learning model; selecting (206) an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model; fine-tuning (208) the selected imaging protocol template based on the patient data using a third machine learning model; andoutputting (210) the fine-tuned imaging protocol template.

16. The non-transitory computer-readable storage medium of claim 15, wherein selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocoling information using a second machine learning model comprises: encoding text of the extracted imaging protocoling information and descriptions of the plurality of candidate imaging protocol templates into vectors; determining a similarity measure between the vector of the text of the extracted imaging protocoling information and each of the vectors of the descriptions of the plurality of candidate imaging protocol templates; and selecting a candidate imaging protocol template having a vector most similar to the vector of the of the extracted imaging protocoling information based on the similarity measures.

17. The non-transitory computer-readable storage medium of claim 15, wherein outputting the fine-tuned imaging protocol template comprises: presenting the fine-tuned selected imaging protocol template annotated with information from the patient data justifying selection of protocols and justifying the fine- tuning.

18. The non-transitory computer-readable storage medium of claim 15, wherein extracting imaging protocoling information from the patient data using a first machine learning model comprises: summarizing information relating to imaging protocoling in the patient data using the first machine learning model.

19. The non-transitory computer-readable storage medium of claim 15, wherein receiving patient data from one or more patient databases comprises:retrieving the patient data from the one or more patient databases using the first machine learning model.

20. The non-transitory computer-readable storage medium of claim 15, wherein the first machine learning model comprises a first LLM (large language model), the second machine learning model comprises a second LLM, and the third machine learning model comprises a third LLM.

Images