Automated imaging protocol formulation for medical images using large language models

Abstract

A system and method for automated imaging protocol formulation are provided. The system receives patient data from one or more patient databases. Using a first machine learning model, imaging protocol formulation information is extracted from the patient data. Using a second machine learning model, an imaging protocol template is selected from multiple candidate templates based on the extracted information. Using a third machine learning model, the selected template is fine-tuned based on the patient data. The fine-tuned template is output.

Description

Technical Field

[0001] This invention generally relates to LLM (Large Language Model), and particularly to the formulation of automated imaging protocols for medical images using LLM. Background Technology

[0002] Imaging protocol formulation refers to the process of reviewing the order in which clinicians acquire medical images of patients and assigning specific imaging protocols to guide the acquisition of these images. Protocols typically define imaging modalities, anatomical regions of interest, and acquisition parameters used to acquire the medical images. Recently, rule-based text processing methods using patient data have been proposed for routine automated imaging protocol formulation. However, such routine rule-based methods do not generalize well across clinical settings. Summary of the Invention

[0003] According to one or more embodiments, a system and method for automated imaging protocol formulation are provided. The system receives patient data from one or more patient databases. Using a first machine learning model, imaging protocol formulation information is extracted from the patient data. Using a second machine learning model, an imaging protocol template is selected from a plurality of candidate imaging protocol templates based on the extracted imaging protocol formulation information. Using a third machine learning model, the selected imaging protocol template is fine-tuned based on the patient data. The fine-tuned imaging protocol template is output.

[0004] In one embodiment, the extracted text of the proposed imaging protocol information and descriptions of multiple candidate imaging protocol templates are encoded into vectors. A similarity measure is determined between each of the vectors of the extracted text of the proposed imaging protocol information and the vectors of the descriptions of the multiple candidate imaging protocol templates. Based on the similarity measure, the candidate imaging protocol template with the vector most similar to the vector of the extracted proposed imaging protocol information is selected.

[0005] In one embodiment, a third machine learning model is used to fine-tune the selected imaging protocol template based on patient data and protocol formulation criteria at the clinical location. The selected imaging protocol template can be fine-tuned using a decision forest.

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

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

[0008] In one embodiment, a first machine learning model is used to outline information in the patient data related to the formulation of imaging protocols.

[0009] In one embodiment, a first machine learning model is used to retrieve patient data from one or more patient databases.

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

[0011] These and other advantages of the invention will be apparent to those skilled in the art from the following detailed description and accompanying drawings. Attached Figure Description

[0012] Figure 1 A schematic diagram of an automated imaging protocol formulation system for acquiring medical images of a patient, according to one or more embodiments, is shown. Figure 2 A method for developing an automated imaging protocol for acquiring medical images of a patient, according to one or more embodiments, is illustrated; Figure 3 An exemplary artificial neural network that can be used to implement one or more embodiments is shown; Figure 4 This illustrates a convolutional neural network that can be used to implement one or more embodiments; Figure 5 A schematic structure is shown that can be used to implement a recursive machine learning model of one or more embodiments; and Figure 6 A high-level block diagram of a computer that can be used to implement one or more embodiments is shown. Detailed Implementation

[0013] This invention generally relates to methods and systems for automatically formulating imaging protocols for medical images using LLM (Large Language Model). Embodiments of the invention described herein are intended to provide a visual understanding of such methods and systems. Digital images typically consist of digital representations of one or more objects (or shapes). In this document, the digital representations of objects are generally described based on identification and manipulation of the objects. Such manipulation is a virtual manipulation implemented in the memory or other circuitry / hardware of a computer system. Therefore, it is to be understood that embodiments of the invention can be executed within a computer system using data stored within the computer system.

[0014] The embodiments described herein use LLM to provide automated imaging protocol formulation for medical images. The automated imaging protocol formulation according to the embodiments described herein avoids the vulnerabilities of conventional rule-based methods by directly extracting imaging protocol formulation information from unstructured patient data using various LLMs, selecting an imaging protocol template based on the extracted imaging protocol formulation information, and fine-tuning the selected imaging protocol template.

[0015] Figure 1 A schematic diagram 100 of an automated imaging protocol formulation system for acquiring medical images of a patient, according to one or more embodiments, is shown. As illustrated in schematic diagram 100, the automated imaging protocol formulation system includes a data collection and preprocessing module 104, a protocol template selection module 106, a protocol refinement module 110, and optionally, a protocol template definition module 108 and a protocol review interface 112. Modules 104-112 of the automated imaging protocol formulation system can be powered by one or more suitable computing devices (such as, for example...) Figure 6 The computer (602) is used for implementation.

[0016] Data collection and preprocessing module 104 collects data from one or more patient databases 102-A, 102-B, ..., 102- n Patient data is retrieved and cleaned from the patient database (collectively referred to as patient database 102), and imaging protocol formulation information is extracted from the patient data. Protocol template selection module 106 selects the best-matching imaging protocol template based on the extracted imaging protocol formulation information. Protocol refinement module 110 fine-tunes the selected imaging protocol template according to, for example, protocol formulation criteria based on patient data and clinical location. Optionally, candidate protocol templates can be defined using protocol template definition module 108, and the fine-tuned imaging protocol templates can be reviewed and modified using protocol review interface 112.

[0017] Figure 2 A method 200 for acquiring medical images of a patient, according to one or more embodiments, is illustrated. The steps of method 200 can be performed by one or more suitable computing devices (such as, for example...). Figure 6 The computer (602) will be used to execute this. Further reference will be provided. Figure 1 A schematic diagram of an automatic imaging protocol formulation system is used to describe this. Figure 2 Method 200.

[0018] exist Figure 2 At step 202, patient data is received from one or more patient databases. In one example, as shown in schematic diagram 100, patient data may be retrieved from one or more patient databases 102 by data collection and preprocessing module 104.

[0019] Patient data may include any suitable data about the patient. For example, patient data may include text-based data about the patient, such as demographic information, vital signs, medical history, family history, laboratory results, medications, measurements, and information extracted from medical images. Text-based patient data may include structured and / or unstructured data organized into specific fields, each with a defined purpose. In another example, patient data may include imaging data about the patient, such as medical images, scanned notes, etc. Medical images may be any suitable modality or combination of modalities, such as CT (computed tomography), MRI (magnetic resonance imaging), US (ultrasound), X-rays, etc. Medical images may be 2D (two-dimensional) images and / or 3D (three-dimensional) volumes. One or more patient databases may include, for example, 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 for storing patient data.

[0020] In one embodiment, a first LLM retrieves patient data from one or more patient databases based on a prompted API (Application Programming Interface) request. The first LLM receives one or more prompts as input, which include instructions for outputting a function call sequence (e.g., "Current Medication:"), and generates that function call sequence as output. The instructions convey at least 1) the task (e.g., collecting and summarizing information from a patient database related to image protocol formulation) and 2) the available API function call sequences and what information they will return. The LLM continues by outputting the function call sequence as appropriate (replacing the function call sequence with the data returned by the API). In response to the function call sequence, one or more API calls are made to one or more patient databases to retrieve patient data at the fields corresponding to the function call sequence in the patient databases. This allows the first LLM to dynamically determine what information is needed for a given scenario based on previously retrieved patient data. Once patient data is retrieved via the API, the first LLM is further prompted to summarize the patient data for image protocol formulation, after which the first LLM outputs the summary. In another embodiment, patient data is retrieved from one or more patient databases based on the available APIs of one or more patient databases. The available APIs of one or more patient databases are fully invoked to retrieve extensive patient data from one or more patient databases.

[0021] Patient data can be loaded into a storage device or memory, for example, into one or more patient databases (e.g., Figure 6The computer 602's memory 610 or storage device 612) and / or via a network interface through one or more remote patient databases (e.g., Figure 6 The network interface 606 of the computer 602 receives patient data directly.

[0022] exist Figure 2 At step 204, a first machine learning model is used to extract imaging protocol formulation information from the patient data. In one example, as shown in schematic diagram 100, the data collection and preprocessing module 104 extracts imaging protocol formulation information from the patient data.

[0023] Imaging protocol formulation information may include any suitable information relating to the formulation of the imaging protocol. For example, imaging protocol formulation information may include scan instructions, imaging history, patient history, allergies, etc. In one embodiment, imaging protocol formulation information is an overview of relevant information in the patient data relating to the formulation of the imaging protocol.

[0024] The first machine learning model can be any suitable machine learning-based model. In one embodiment, the first machine learning model is a first LLM. The first LLM receives one or more prompts as input, the prompts including patient data and instructions for extracting (e.g., outlining) imaging protocol formulation information from the patient data, and generates the extracted imaging protocol formulation information as output. These one or more prompts can be, for example, manually entered into the first LLM by a user, or obtained from a computing system (e.g., via one or more APIs). Figure 6 The computer (602) inputs the patient data into the first LLM. The patient data may include unstructured patient data input into the first LLM via one or more prompts with context labels (e.g., indicating the type, time, and origin of the patient data), and / or may include structured patient data input into the first LLM as prose with structure labels via one or more prompts (e.g., the patient's age is <age>).

[0025] The first LLM can be any suitable pre-trained deep learning-based LLM. For example, the LLM can be based on a transformer architecture that uses a self-attention mechanism to capture long-range dependencies in text. An example of a transformer-based architecture is GPT (Generative Pre-trained Transformer), which has a multi-layer transformer decoder architecture that can be pre-trained to optimize the next lexical prediction task and then fine-tuned for various downstream tasks using labeled data. GPT-based LLMs can be trained using reinforcement learning with human feedback to perform various natural language processing tasks. Other exemplary transformer-based architectures include BLOOM (BigScience Open Access Multilingual Language Model) and BERT (Bidirectional Encoder Representation Based on Transformer). The first LLM can be a multimodal LLM that receives both text-based patient data and patient imaging data. In one embodiment, the first LLM is a task-independent LLM. In another embodiment, the first LLM is fine-tuned for tasks such as extracting imaging protocol formulation information from patient data using, for example, training with a human in a loop.

[0026] exist Figure 2 At step 206, a second machine learning model is used to select an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocol formulation information. Each of the plurality of candidate imaging protocol templates includes a predefined set of protocols for acquiring medical images of a patient. Exemplary protocols may include an imaging modality, an anatomical region of interest, and acquisition parameters for acquiring medical images (e.g., acquisition protocol, field of view, slice thickness, appropriate coils in MRI, reconstruction nuclei in CT, etc.). In one example, as illustrated in schematic diagram 100, the imaging protocol template is selected by protocol template selection module 106.

[0027] The second machine learning model can 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 one or more prompts as input, the prompts including extracted imaging protocol formulation information, descriptions of multiple candidate imaging protocol templates, and optionally, other patient data (e.g., patient history), and generates candidate imaging protocol templates as output, the candidate imaging protocol templates being most similar to the extracted imaging protocol formulation information (and optionally, other patient data), as the selected imaging protocol template. One or more prompts can be, for example, manually entered into the second LLM by a user, or obtained from a computing system (e.g., via one or more APIs). Figure 6The computer (602) inputs the data into the second LLM. The second LLM can be any suitable pre-trained deep learning-based encoder-based LLM. For example, the second LLM can be based on a transformer architecture such as GPT, BLOOM, BERT, etc. The second LLM can be a multimodal LLM used to receive both text-based data and imaging data from the patient.

[0028] The second LLM determines the most similar candidate imaging protocol templates based on semantic similarity, measuring how close the text of the extracted imaging protocol proposal information (and optionally, other patient data) is in meaning to the descriptions of multiple candidate imaging protocol templates. For example, a description of a candidate imaging protocol template including "headache" and an extracted imaging protocol proposal information including instructions for a scan including "migraine" would be semantically similar. In one embodiment, the second LLM encodes the text of the extracted imaging protocol proposal information (and optionally, other patient data) and the descriptions of multiple candidate imaging protocol templates into corresponding vectors. A similarity measure is then determined between the vector of the text of the extracted imaging protocol proposal information (and optionally, other patient data) and the vectors of the descriptions of the multiple candidate imaging protocol templates. The similarity measure can be, for example, cosine similarity, or any suitable measure of similarity or distance. The candidate imaging protocol template with the vector most similar to the vector of the extracted imaging protocol proposal information (and optionally, other patient data) is then selected.

[0029] exist Figure 2 At step 208, a third machine learning model is used to fine-tune the selected imaging protocol template based on patient data. In one example, as shown in schematic diagram 100, the selected imaging protocol template is fine-tuned by protocol refinement module 110.

[0030] In one embodiment, the patient data used to fine-tune the selected imaging protocol template includes any patient data related to the selection or modification of the imaging protocol, such as allergies, contraindications, complications, etc. For example, some patients may not be able to receive contrast agents or have metal implants requiring special precautions. The selected imaging protocol template may be additionally or alternatively fine-tuned based on protocol development criteria of the clinical location (e.g., hospital). Protocol development criteria include a set of rules or preferences defining how the protocol should be tailored to the clinical location, such as naming, formatting, customization for different scenarios, etc. For example, some hospitals may prefer shorter scan times or lower contrast doses compared to other hospitals.

[0031] The third machine learning model can be any suitable machine learning-based model. In one embodiment, the third machine learning model is a third LLM. The third LLM receives one or more prompts as input, which include a selected imaging protocol template, patient data, and / or clinical criteria for the clinical location, and generates a fine-tuned imaging protocol template as output. The one or more prompts can be, for example, manually entered into the third LLM by a user, or obtained from a computing system (e.g., via one or more APIs). Figure 6 The computer (602) inputs the data into the third LLM. The third LLM can be any suitable pre-trained deep learning-based LLM. For example, the third LLM can be based on a transformer architecture such as GPT, BLOOM, BERT, etc. The third LLM can be a multimodal LLM used to receive text-based data and imaging data from the patient.

[0032] In one embodiment, fine-tuning can be encoded in a decision forest, with free-text queries at each node. For example, one node in the decision tree might query whether the patient has renal insufficiency or a contrast agent allergy. If so, an endogenous contrast agent is used; otherwise, an injected contrast agent is used. A third LLM will be prompted to answer the query and refine the imaging protocol template accordingly, following the decision tree. By structuring decision points as free-text queries, protocol drafting criteria can be set or updated without reprogramming.

[0033] exist Figure 2 At step 210, the fine-tuned imaging protocol template is output. For example, the fine-tuned imaging protocol template can be output in the following manner: on the display device of the computer system (e.g., Figure 6 The finely tuned imaging protocol template is displayed on the input / output device 608 of the computer 602, and stored in the memory or storage device of the computer system (e.g., ...). Figure 6 The fine-tuned imaging protocol template is stored on the memory 610 or storage device 612 of the computer 602, or by transmitting the fine-tuned imaging protocol template to a remote computer system (e.g., Figure 6 Computer 602).

[0034] In one embodiment, such as in a clinical workflow for achieving just-in-time protocol formulation or just-in-time imaging, medical images are acquired automatically or semi-automatically based on a finely tuned imaging protocol template. Some steps (e.g., contrast agent injection) may require user approval before being performed.

[0035] In one embodiment, optionally, the fine-tuned imaging protocol template can be output to a user interface (e.g., a display device) for review by a user (e.g., a radiologist or another clinician). In one example, as shown in schematic diagram 100, the fine-tuned imaging protocol template is output to a protocol review interface 112. The user interface can present the fine-tuned imaging protocol template, which is annotated with references from patient data that justify the choice of protocol or summarized information. Specific protocol selections made during fine-tuning can also be shown using references from patient data that justify the fine-tuning or summarized information. Users can thus quickly review and approve protocol selections and fine-tune or modify the fine-tuned imaging protocol template.

[0036] In one embodiment, optionally, multiple candidate imaging protocol templates can be automatically defined (e.g., during a previous preprocessing stage). In one example, as illustrated in schematic diagram 100, multiple candidate imaging protocol templates can be automatically defined by protocol template definition module 108. Data from various information systems and databases at clinical sites is analyzed to identify clusters of similar imaging protocols and dimensions of variability within each cluster. For example, dimensions of variability within similar imaging protocols describe protocol changes (e.g., inclusion or exclusion of specific sequences, use of contrast agents). Dimensions of variability within patient scenarios describe factors that might lead a user to select one of the aforementioned protocol variations (e.g., patient weight, tinnitus history, adequate renal function). Clusters of similar imaging protocols and dimensions of variability within each cluster can be used to automatically define candidate imaging protocol templates. Patient data from various patients is also analyzed to identify profiles of patient scenarios that trigger each cluster. The variability within the patient scenario that triggers each cluster is then associated with the dimensions of variability within that cluster. The profiles of the patient scenario that triggers each cluster can be used to automatically define candidate imaging protocol template descriptions. Patient scenarios associated with the dimension of variability can be used to draft protocol formulation standards, which can be used or modified to fine-tune imaging protocol templates (e.g., in...). Figure 2 (Step 208). For example, there might be a cluster of protocols requiring vascular imaging, and within this cluster, patients with renal insufficiency would receive an ASL (Artery Spin Labeling) sequence, while patients with adequate renal function would receive an angiographic scan. In the protocol selection step, a protocol with vascular imaging will be selected, and in the fine-tuning, either ASL or exogenous angiography will be selected.

[0037] 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, have different cues, etc.). The LLM is trained using a large training data corpus during an initial offline or training phase. Once trained, the trained LLM is applied during an online or inference phase, for example, by performing... Figure 2 Method 200.

[0038] In one embodiment, if one or more prompts input to the first LLM, second LLM, and / or third LLM exceed the maximum size of their respective context windows, data for one or more prompts can be stored in a vector database as needed. Such data can be retrieved in segments within the maximum size of the context windows as required. For example, relevant data can be retrieved from the vector database, and the corresponding LLM can perform analysis based on the retrieved relevant data, save the output, and discard the context window. This can be repeated as necessary.

[0039] Advantageously, the embodiments described herein provide automated imaging protocol formulation, resulting in time savings for technicians and radiologists. Furthermore, the embodiments described herein utilize LLM to extract information from free text natural language input, rather than structured input. This enables integration into existing clinical information systems without requiring a consistent EHR format and content, while also making the user interface more intuitive and reducing training costs. Additionally, the protocol template definition module allows for the intuitive definition of protocol formulation decision trees without requiring programming. This reduces the burden on application experts and clinicians debugging the system. Moreover, the embodiments described herein provide a protocol review interface that demonstrates the reasonableness of the protocol formulation decisions, making it understandable and more trustworthy for end users.

[0040] The embodiments described herein are described with respect to the claimed system and the claimed method. Features, advantages, or alternative embodiments described herein may be assigned to other claimed objects, and vice versa. In other words, the claims and embodiments for the system may be improved using features described or claimed in the context of the corresponding method. In this case, the functional features of the method are implemented through the physical units of the system.

[0041] Furthermore, certain embodiments described herein are described with respect to methods and systems utilizing trained machine learning models, and with respect to methods and systems for providing trained machine learning models. Features, advantages, or alternative embodiments described herein may be assigned to other claimed objects, and vice versa. In other words, the claims and embodiments for providing trained machine learning models may be modified using features described or claimed in the context of utilizing trained machine learning models, and vice versa. Specifically, the dataset used in methods and systems for utilizing trained machine learning models may have the same properties and characteristics as the corresponding dataset used in methods and systems for providing trained machine learning models, and the trained machine learning model provided by the corresponding method and system may be used in methods and systems for utilizing trained machine learning models.

[0042] Generally, trained machine learning models mimic human cognitive functions associated with other human thought processes. Specifically, through training on training data, machine learning models can adapt to new situations and detect and infer patterns. Another term for a “trained machine learning model” is a “trained function.”

[0043] Generally, the parameters of a machine learning model can be adapted through training. Specifically, 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. Specifically, the parameters of a machine learning model can be iteratively adapted through several training steps. Specifically, within training, a certain cost function can be minimized. Specifically, within the training of a neural network, the backpropagation algorithm can be used.

[0044] In particular, machine learning models, such as, for example, in Figure 2 The first machine learning model used in steps 202 and 204, the second machine learning model used in step 206, and the third machine learning model used in step 208 may include, for example, neural networks, support vector machines, decision trees, and / or Bayesian methods, and / or may be based on machine learning models such as k-means clustering, Q-learning, genetic algorithms, and / or association rules. Specifically, the neural network may be, for example, a deep neural network, a convolutional neural network, or a convolutional deep neural network. Furthermore, the neural network may be, for example, an adversarial network, a deep adversarial network, and / or a generative adversarial network.

[0045] Figure 3 An embodiment of an artificial neural network 300 that can be used to implement one or more machine learning models described herein is shown. Alternative terms for “artificial neural network” are “neural network,” “artificial neural network,” or “neural network.”

[0046] The artificial neural network 300 includes nodes 320, ..., 332 and edges 340, ..., 342, where each edge 340, ..., 342 is a directed connection from a first node 320, ..., 332 to a second node 320, ..., 332. Generally, the first nodes 320, ..., 332 and the second nodes 320, ..., 332 are different nodes 320, ..., 332; however, it is also possible that the first nodes 320, ..., 332 and the second nodes 320, ..., 332 are the same. For example, in... Figure 3 In the diagram, edge 340 is a directed connection from node 320 to node 323, and edge 342 is a directed connection from node 330 to node 332. Edges 340, ..., 342 from the first node 320, ..., 332 to the second node 320, ..., 332 are also represented as the "incoming edges" of the second node 320, ..., 332 and the "outgoing edges" of the first node 320, ..., 332.

[0047] In this embodiment, nodes 320, ..., 332 of the artificial neural network 300 can be arranged in layers 310, ..., 313, wherein these layers may include an inherent order introduced by edges 340, ..., 342 between nodes 320, ..., 332. Specifically, edges 340, ..., 342 may exist only between adjacent layers of nodes. In the illustrated embodiment, there exists an input layer 310 comprising only nodes 320, ..., 322 without incoming edges, an output layer 313 comprising only nodes 331, 332 without outgoing edges, and hidden layers 311, 312 between the input layer 310 and the output layer 313. Generally, the number of hidden layers 311, 312 can be arbitrarily chosen. The number of nodes 320, ..., 322 in the input layer 310 is typically related to the number of input values ​​of the neural network, and the number of nodes 331, 332 in the output layer 313 is typically related to the number of output values ​​of the neural network.

[0048] Specifically, (real) numbers can be assigned as values ​​to each node 320, ..., 332 of the neural network 300. Here, x (n) i This represents the value of the i-th node 320, ..., 332 in the n-th layer 310, ..., 313. The values ​​of nodes 320, ..., 322 in the input layer 310 are equivalent to the input values ​​of the neural network 300, and the values ​​of nodes 331, 332 in the output layer 313 are equivalent to the output values ​​of the neural network 300. Furthermore, each edge 340, ..., 342 may include a weight (which is a real number), specifically, a real number within the interval [-1, 1] or the interval [0, 1]. Here, w (m,n) i,jLet w represent the weight of the edge between the i-th node 320, ..., 332 of layer m (310, ..., 313) and the j-th node 320, ..., 332 of layer n (310, ..., 313). Furthermore, for the weight w... (n,n+1) i,j Define the abbreviation w (n) i,j .

[0049] Specifically, in order to calculate the output value of neural network 300, the input value is propagated through the neural network. Specifically, the values ​​of nodes 320, ..., 332 in the (n+1)th layer 310, ..., 313 can be calculated based on the values ​​of nodes 320, ..., 332 in the nth layer 310, ..., 313 using the following formula: .

[0050] In this paper, the function f is the transfer function (another term is "activation function"). Known transfer functions are step functions, sigmoid functions (e.g., logistic functions, generalized logistic functions, hyperbolic tangent functions, arctangent functions, error functions, smoothstep functions, or rectifier functions). Transfer functions are primarily used for normalization purposes.

[0051] Specifically, these values ​​are propagated layer by layer through a neural network, wherein the value of the input layer 310 is given by the input of the neural network 300, wherein the value of the first hidden layer 311 can be calculated based on the value of the input layer 310 of the neural network, wherein the value of the second hidden layer 312 can be calculated based on the value of the first hidden layer 311, and so on.

[0052] To set the value of the edge The neural network 300 must be trained using training data. Specifically, the training data includes training input data and training output data (denoted as t). i For the training step, neural network 300 is applied to the training input data to generate the computed output data. Specifically, the training data and the computed output data include a number of values ​​equal to the number of nodes in the output layer.

[0053] Specifically, the weights within the neural network (up to 300) are recursively adapted using a comparison between the calculated output data and the training data (backpropagation algorithm). Specifically, the weights are changed according to the following formula: Where γ is the learning rate, and in the case that the (n+1)th layer is not the output layer, based on Count The equation is calculated recursively as follows: And in the case that the (n+1)th layer is the output layer 313, the number The equation is calculated recursively as follows: in It is the first derivative of the activation function, and It is the comparison training value of the j-th node of the output layer 313.

[0054] A convolutional neural network (CNN) is a neural network that uses convolution operations instead of general matrix multiplication in at least one of its layers (so-called "convolutional layers"). Specifically, a convolutional layer performs a dot product of one or more convolutional kernels with the input data / image of the convolutional layer, where the entries for the one or more convolutional kernels are parameters or weights adapted through training. In particular, Frobenius inner products and ReLU activation functions can be used. A CNN may include additional layers such as pooling layers, fully connected layers, and normalization layers.

[0055] By using convolutional neural networks, input images can be processed very efficiently because convolutional operations based on different kernels can extract various image features. This allows relevant image features to be found during training by adapting the weights of the convolutional kernels. Furthermore, due to weight sharing within the convolutional kernels, fewer parameters need to be trained, preventing overfitting during training and allowing for faster training or more layers in the network, thus improving network performance.

[0056] Figure 4 An embodiment of a convolutional neural network 400 that can be used to implement one or more machine learning models described herein is shown. In the shown embodiment, the convolutional neural network 400 includes an input node layer 410, a convolutional layer 411, a pooling layer 413, a fully connected layer 414, an output node layer 416, and hidden node layers 412, 414. Alternatively, the convolutional neural network 400 may include a plurality of convolutional layers 411, a plurality of pooling layers 413, and a plurality of fully connected layers 415, as well as other types of layers. The order of the layers can be arbitrarily chosen, but typically the fully connected layer 415 is used as the last layer before the output layer 416.

[0057] Specifically, within the convolutional neural network 400, nodes 420, 422, and 424 of node layers 410, 412, and 414 can be viewed as arranged as a d-dimensional matrix or a d-dimensional image. In particular, in the two-dimensional case, the values ​​of nodes 420, 422, and 424 indexed by i and j in the nth node layer 410, 412, and 414 can be represented as x(n)[i,j]. However, the arrangement of nodes 420, 422, and 424 in a node layer 410, 412, and 414 itself has no effect on the computations performed within the convolutional neural network 400, as these are given only by the weights and structure of the edges.

[0058] Convolutional layer 411 is a connection layer between the preceding node layer 410 (with node value x(n-1)) and the following node layer 412 (with node value x(n)). Specifically, convolutional layer 411 is characterized by the structure and weights of the input edges that form the convolution operation based on a certain number of kernels. In particular, the structure and weights of the edges of convolutional layer 411 are selected such that the value x(n) of node 422 in the following node layer 412 is computed as a convolution based on the value x(n-1) of node 420 in the preceding node layer 410. In the two-dimensional case, convolution* is defined as: .

[0059] Here, the kernel K is a d-dimensional matrix (in this embodiment, a two-dimensional matrix), which is typically small compared to the number of nodes 420, 422 (e.g., a 3×3 or 5×5 matrix). Specifically, this implies that the weights of the edges in convolutional layer 411 are not independent, but are chosen such that they produce the convolution equation. In particular, for a 3×3 kernel, there are only 9 independent weights (each entry in the kernel matrix corresponds to one independent weight), regardless of the number of nodes 420, 422 in the preceding node layer 410 and the following node layer 412.

[0060] Generally, convolutional neural networks 400 use node layers 410, 412, and 414 with multiple channels, especially due to the use of multiple kernels in convolutional layer 411. In those cases, the node layer can be considered as a (d+1)-dimensional matrix (the first dimension indexing the channels). Then, the action of convolutional layer 411 is defined as a two-dimensional example of the following: in Corresponding to the a-th channel of the preceding node layer 410, Corresponding to the b-th channel of the subsequent node layer 412, and This corresponds to one of the kernels. If convolutional layer 411 acts on the preceding node layer 410 with A channels and outputs the following node layer 412 with B channels, then there exist A·B independent d-dimensional kernels. .

[0061] Generally, in a convolutional neural network 400, an activation function is used. In this embodiment, ReLU (an abbreviation for "Rectified Linear Unit") is used, where... This makes the action of convolutional layer 411 in the two-dimensional example: .

[0062] Other activation functions may also be used, such as ELU (an abbreviation for "Exponential Linear Unit"), LeakyReLU, Sigmoid, Tanh, or Softmax.

[0063] In the shown embodiment, the input layer 410 includes 36 nodes 420 arranged as a two-dimensional 6×6 matrix. The first hidden node layer 412 includes 72 nodes 422 arranged as two two-dimensional 6×6 matrices, each of which is the result of convolving the values ​​of the input layer with a 3×3 kernel within the convolutional layer 411. Equivalently, the nodes 422 of the first hidden node layer 412 can be interpreted as being arranged as a three-dimensional 2×6×6 matrix, where the first dimension corresponds to the channel dimension.

[0064] The advantage of using convolutional layer 411 is that it can take advantage of the spatial local correlation of the input data by enforcing a local connectivity pattern between nodes in neighboring layers, in particular by connecting each node only to a small region of nodes in the previous layer.

[0065] Pooling layer 413 is the connecting layer between the preceding node layer 412 (with node value x(n-1)) and the following node layer 414 (with node value x(n)). Specifically, pooling layer 413 can be characterized by the activation function and edge structure and weights that form the pooling operation based on the nonlinear pooling function f. For example, in the two-dimensional case, the value x(n) of node 424 in the following node layer 414 can be calculated based on the value x(n-1) of node 422 in the preceding node layer 412 as follows: In other words, by using pooling layer 413, the number of nodes 422 and 424 can be reduced by replacing a number of adjacent nodes 422 in the preceding node layer 412 with a single node 422 from the subsequent node layer 414, the replacement being calculated based on the values ​​of the number of adjacent nodes. Specifically, the pooling function f can be a maximum value function, an average value function, or an L2 norm function. Specifically, for pooling layer 413, the weights of the incoming edges are fixed and are not modified through training.

[0066] The advantage of using pooling layer 413 is that it reduces the number of nodes 422 and 424 and the number of parameters. This results in a reduction in the computational cost of the network and better control over overfitting.

[0067] In the illustrated embodiment, pooling layer 413 is a max-pooling layer that replaces four adjacent nodes with only one node, the value of which is the maximum of the four adjacent node values. Max-pooling is applied to each d-dimensional matrix of the previous layer; in this embodiment, max-pooling is applied to each of two two-dimensional matrices, thereby reducing the number of nodes from 72 to 18.

[0068] Generally, the last layer of the convolutional neural network 400 is a fully connected layer 415. The fully connected layer 415 is the connection layer between the preceding node layer 414 and the following node layer 416. The fully connected layer 413 can be characterized by the fact that most, in particular all, edges exist between nodes 414 of the preceding node layer 414 and nodes 416 of the following node layer, and the weight of each of these edges can be adjusted individually.

[0069] In this embodiment, the nodes 424 of the preceding node layer 414 of the fully connected layer 415 are displayed as a two-dimensional matrix, and additionally as unrelated nodes (indicated as a row of nodes, where the number of nodes is reduced for better presentation). This operation is also referred to as "flattening". In this embodiment, the number of nodes 426 in the following node layer 416 of the fully connected layer 415 is less than the number of nodes 424 in the preceding node layer 414. Alternatively, the number of nodes 426 can be equal to or greater than the number of nodes 424.

[0070] Furthermore, in this embodiment, a Softmax activation function is used within the fully connected layer 415. By applying the Softmax function, the sum of the values ​​of all nodes 426 in the output layer 416 is 1, and all values ​​of all nodes 426 in the output layer 416 are real numbers between 0 and 1. In particular, if the input data is classified using the convolutional neural network 400, the values ​​of the output layer 416 can be interpreted as the probability that the input data falls into one of the different categories.

[0071] Specifically, a convolutional neural network 400 can be trained based on the backpropagation algorithm. To prevent overfitting, regularization methods can be used, such as dropout of nodes 420, ..., 424, random pooling, use of artificial data, and weight decay based on L1 or L2 norm or maximum norm constraints.

[0072] According to one aspect, a machine learning model may include one or more Residual Networks (ResNets). Specifically, a ResNet is an artificial neural network that includes at least one jump or skip connection for hopping at at least one layer of the artificial neural network. Specifically, a ResNet may be a convolutional neural network that includes one or more skip connections that skip one or more convolutional layers respectively. According to some examples, a ResNet may be represented as an m-layer ResNet, where m is the number of layers in the corresponding architecture, and according to some examples, it may take values ​​of 34, 50, 101, or 152. According to some examples, such an m-layer ResNet may correspondingly include (m-2) / 2 skip connections.

[0073] Skip connections can be viewed as bypassing a process where the output of a previous layer is directly fed into a layer following that layer via one or more bypassed layers. Instead of directly fitting the desired mapping, the bypassed layer must then fit a residual mapping that "balances" the output of the directly fed layer.

[0074] Fitting residual mappings is computationally easier to optimize than directional mappings. More importantly, this mitigates the vanishing / exploding gradient problem during optimization when training machine learning models: if a bypassed layer encounters this problem, its contribution can be skipped by regularizing the output of the direct feed. Therefore, the advantage of using ResNet is that much deeper networks can be trained.

[0075] Specifically, a recursive machine learning model is a machine learning model whose output depends not only on the input values ​​and the parameters of the machine learning model adapted through the training process, but also on the hidden state vector, which is based on previous inputs used on the recursive machine learning model. In particular, the recursive machine learning model may include additional stored states or additional structures that incorporate time delays or include feedback loops.

[0076] Specifically, the basic structure of a recurrent machine learning model can be a neural network, which can be represented as a recurrent neural network. Such a recurrent neural network can be described as an artificial neural network where the connections between nodes form a directional graph along a time series. Specifically, a recurrent neural network can be interpreted as a directional acyclic graph. Specifically, the recurrent neural network can be a finite-spiking recurrent neural network or an infinite-spiking recurrent neural network (where a finite-spiking network can be unfolded and replaced with a strictly feedforward neural network, and an infinite-spiking network cannot be unfolded and replaced with a strictly feedforward neural network).

[0077] Specifically, recurrent neural networks can be trained based on the BPTT algorithm (an acronym for "backward propagation by time"), the RTRL algorithm (an acronym for "real-time recursive learning"), and / or genetic algorithms.

[0078] By using a recursive machine learning model, input data including sequences of variable length can be used. Specifically, this implies that the method cannot be used only with a fixed number of input datasets (and requires different training for all other numbers of input datasets used as input), but can be used with any number of input datasets. This means that, independent of the number of input datasets contained in different sequences, the entire training dataset can be used within training, and the training data is not reduced to training data corresponding to a fixed number of consecutive input datasets.

[0079] Figure 5 The schematic structure of a recursive machine learning model F is illustrated using both recursive representation 502 and unfolded representation 504. This recursive machine learning model F can be used to implement one or more machine learning models described in this paper. The recursive machine learning model takes several input datasets x, x1, ..., x2. N 506 is used as input, and corresponding output datasets y, y1, ..., y2 are created. N A set of 508. Furthermore, the output depends on the so-called hidden vectors h1, h2, ..., h3. N 510, the hidden vectors implicitly include information about the input dataset previously used as input to the recursive machine learning model F 512. This is achieved by using these hidden vectors h, h1, ..., h2. N 510, the order of the input dataset can be utilized.

[0080] In a single processing step, the recursive machine learning model F512 takes the hidden vector created in the previous step. and the input dataset x n As input. Within this step, the recursive machine learning model F generates the updated hidden vector h. n and the output dataset yn As output. In other words, a processing step calculates... Alternatively, by splitting the recursive machine learning model F512 into a part F(y) that computes the output data and a part F(h) that computes the hidden vectors, one processing step computes... and For the first processing step, h0 can be randomly selected or filled with all entries that are zero. The parameters of the recurrent machine learning model F 512, which was previously trained on the training dataset, remain unchanged between the different processing steps.

[0081] In particular, the output data and hidden vectors of the processing step depend on all the previous input datasets used in the previous steps. and .

[0082] The systems, apparatus, and methods described herein can be implemented using digital circuitry or using one or more computers employing 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 disks, internal hard disks and removable disks, magneto-optical disks, optical disks, etc.

[0083] The systems, apparatus, and methods described herein can be implemented using computers operating in a client-server relationship. Typically, in such a system, the client computer is located remotely from the server computer and interacts via a network. The client-server relationship can be defined and controlled by computer programs running on the respective client and server computers.

[0084] The systems, apparatuses, and methods described herein can be implemented within a network-based cloud computing system. In such a system, a server or another processor connected to the network communicates with one or more client computers via the network. Client computers can communicate with the server via, for example, a web browser application residing and operating on the client computer. Client computers can store data on the server and access that data via the network. Client computers can transmit requests for data or for online services to the server via the network. The server can perform the requested service and provide data to one or more client computers(s). The server can also transmit data suitable for causing the client computer to perform specified functions (e.g., perform calculations, display specified data on a screen, etc.). For example, the server can transmit requests suitable for causing the client computer to perform one or more steps or functions of the methods and workflows described herein, including... Figure 1 Or one or more of the steps or functions in step 2. Certain steps or functions of the methods and workflows described herein (including...) Figure 1 One or more of the steps or functions in step 2 may be performed by a server or by another processor in a network-based cloud computing system. Certain steps or functions of the methods and workflows described herein (including...) Figure 1 (or one or more of steps 2) can be performed by a client computer in a web-based cloud computing system. The steps or functions of the methods and workflows described herein (including...) Figure 1 (or one or more of steps 2) can be performed by a server and / or by a client computer in a web-based cloud computing system in any combination.

[0085] The systems, apparatuses, and methods described herein can be implemented using a computer program product tangibly embodied in an information carrier (e.g., embodied in a non-transitory machine-readable storage device) for execution by a programmable processor; and the methods and workflow steps described herein (including Figure 1 (or one or more of the steps or functions in step 2) can be implemented using one or more computer programs 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 produce a certain result. A computer program can be written in any form of programming language (including compiled or interpreted languages) and can be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0086] Figure 6 This document depicts a high-level block diagram of an example computer 602 that can be used to implement the systems, apparatus, and methods described herein. Computer 602 includes a processor 604 operatively coupled to a data storage device 612 and a memory 610. Processor 604 controls this operation by executing computer program instructions that define the overall operation of computer 602. The computer program instructions may be stored in the data storage device 612 or other computer-readable medium and loaded into memory 610 when execution of the computer program instructions is desired. Therefore, Figure 1 The methods and workflow steps or functions of option 2 can be defined by computer program instructions stored in memory 610 and / or data storage device 612, and can be controlled by processor 604 that executes the computer program instructions. For example, the computer program instructions can be implemented to be programmed by those skilled in the art to perform... Figure 1 Alternatively, it may be computer-executable code containing the methods, workflow steps, or functions of step 2. Therefore, by executing computer program instructions, processor 604 performs... Figure 1Alternatively, methods and workflow steps or functions may be included. 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 (e.g., monitor, keyboard, mouse, speaker, buttons, etc.) that enable a user to interact with computer 602.

[0087] Processor 604 may include both general-purpose microprocessors and special-purpose microprocessors, and may be the sole processor of computer 602 or one of multiple processors. For example, processor 604 may include one or more central processing units (CPUs). Processor 604, data storage device 612 and / or memory 610 may include one or more application-specific integrated circuits (ASICs) and / or one or more field-programmable gate arrays (FPGAs), supplementing or incorporating the foregoing.

[0088] 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 rate synchronous 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 disk storage devices (e.g., internal hard disks and removable disks), magneto-optical disk storage devices, optical disk storage devices, flash memory devices, semiconductor memory devices (e.g., 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) discs), or other non-volatile solid-state memory devices.

[0089] Input / output device 608 may include peripheral devices such as printers, scanners, displays, etc. For example, input / output device 608 may include display devices (such as cathode ray tube (CRT) or liquid crystal display (LCD) monitors) for displaying information to a user, a keyboard, and pointing devices such as mice or trackballs through which the user can provide input to computer 602.

[0090] Image acquisition device 614 can be connected to computer 602 to input image data (e.g., medical images) into computer 602. It is possible to implement image acquisition device 614 and computer 602 as a single device. It is also possible for image acquisition device 614 and computer 602 to communicate wirelessly via a network. In a possible embodiment, computer 602 can be remotely located relative to image acquisition device 614.

[0091] Any or all of the systems, apparatuses, and methods discussed herein may be implemented using one or more computers (such as computer 602).

[0092] Those skilled in the art will recognize that the actual implementation of a computer or computer system may have other structures and may include other components, and for illustrative purposes, Figure 6 It is a high-level representation of some components in such a computer.

[0093] Individuals with male or female gender identity are included in the term, independent of the use of grammatical terms.

[0094] The foregoing detailed descriptions should be understood in every respect as illustrative and exemplary, not restrictive, and the scope of the invention disclosed herein is not determined by these detailed descriptions, but rather by the claims as interpreted under the full breadth permitted by patent law. It is to be understood that the embodiments shown and described herein are merely illustrative descriptions of the principles of the invention, and various modifications can be made by those skilled in the art without departing from the scope and spirit of the invention. Various other combinations of features can be implemented by those skilled in the art without departing from the scope and spirit of the invention.

[0095] The following is a list of non-limiting illustrative embodiments disclosed herein: Illustrative Example 1. A computer-implemented method comprising: receiving patient data from one or more patient databases; extracting imaging protocol formulation 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 protocol formulation 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.

[0096] Illustrative Example 2. A computer-implemented method according to Illustrative Example 1, wherein selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocol formulation information using a second machine learning model includes: encoding the text of the extracted imaging protocol formulation information and the descriptions of the plurality of candidate imaging protocol templates into vectors; determining a similarity measure between each of the vector of the text of the extracted imaging protocol formulation information and the vector of the description 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 extracted imaging protocol formulation information based on the similarity measure. Illustrative Example 3. A computer-implemented method according to any one of Illustrative Examples 1-2, wherein fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model includes: using the third machine learning model to fine-tune the selected imaging protocol template based on the patient data and protocol formulation criteria of the clinical location.

[0097] Illustrative Example 4. A computer-implemented method according to any one of Illustrative Examples 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.

[0098] Illustrative Example 5. The computer-implemented method according to any one of Illustrative Examples 1-4 further includes: identifying clusters of imaging protocols in systems at one or more clinical locations; 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.

[0099] Illustrative Example 6. A computer-implemented method according to any one of Illustrative Examples 1-5, wherein outputting the fine-tuned imaging protocol template comprises: presenting a fine-tuned selected imaging protocol template, the fine-tuned selected imaging protocol template being annotated with information from the patient data that demonstrates the rationality of the protocol selection and the rationality of the fine-tuning.

[0100] Illustrative Example 7. A computer-implemented method according to any one of Illustrative Examples 1-6, wherein extracting imaging protocol formulation information from the patient data using a first machine learning model includes: using the first machine learning model to summarize information in the patient data relating to imaging protocol formulation.

[0101] Illustrative Example 8. A computer-implemented method according to any one of Illustrative Examples 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.

[0102] Illustrative Example 9. A computer-implemented method according to any one of Illustrative Examples 1-8, wherein the first machine learning model includes a first LLM (Large Language Model), the second machine learning model includes a second LLM, and the third machine learning model includes a third LLM.

[0103] Illustrative Example 10. An apparatus comprising: means for receiving patient data from one or more patient databases; means for extracting imaging protocol formulation information from the patient data using a first machine learning model; means for selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocol formulation 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.

[0104] Illustrative Example 11. According to the apparatus of Illustrative Example 10, the component for selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocol formulation information using a second machine learning model includes: a component for encoding the text of the extracted imaging protocol formulation information and the descriptions of the plurality of candidate imaging protocol templates into vectors; a component for determining a similarity measure between each of the vector of the text of the extracted imaging protocol formulation information and the vector of the description of the plurality of candidate imaging protocol templates; and a component for selecting, based on the similarity measure, a candidate imaging protocol template having a vector most similar to the vector of the extracted imaging protocol formulation information.

[0105] Illustrative Example 12. The apparatus according to any one of Illustrative Examples 10-11, wherein the component for fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model comprises: a component for fine-tuning the selected imaging protocol template based on the patient data and protocol formulation criteria of the clinical location using the third machine learning model.

[0106] Illustrative Example 13. The apparatus according to Illustrative Examples 10-12, wherein the component for fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model includes: a component for fine-tuning the selected imaging protocol template according to a decision forest.

[0107] Illustrative Example 14. The apparatus according to Illustrative Examples 10-13 further includes: a component for identifying clusters of imaging protocols in a system at one or more clinical locations; a component for identifying patient scenarios that trigger the clusters; and a component for automatically generating one or more of the plurality of candidate imaging protocol templates based on the identified patient scenarios.

[0108] Illustrative Example 15. A non-transitory computer-readable storage medium including instructions that, when executed by a computer, cause the computer to perform operations including: receiving patient data from one or more patient databases; extracting imaging protocol formulation 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 protocol formulation 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.

[0109] Illustrative Example 16. According to the non-transitory computer-readable storage medium of Illustrative Example 15, wherein using a second machine learning model, selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocol formulation information includes: encoding the text of the extracted imaging protocol formulation information and the descriptions of the plurality of candidate imaging protocol templates into vectors; determining a similarity measure between each of the vector of the text of the extracted imaging protocol formulation information and the vector of 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 extracted imaging protocol formulation information based on the similarity measure.

[0110] Illustrative Example 17. According to the non-transitory computer-readable storage medium of Illustrative Examples 15-16, outputting the fine-tuned imaging protocol template includes: presenting a fine-tuned selected imaging protocol template, the fine-tuned selected imaging protocol template being annotated with information from the patient data demonstrating the rationality of the protocol selection and the rationality of the fine-tuning.

[0111] Illustrative Example 18. According to the non-transitory computer-readable storage medium of Illustrative Examples 15-17, extracting imaging protocol formulation information from the patient data using a first machine learning model includes: using the first machine learning model to summarize information in the patient data related to imaging protocol formulation.

[0112] Illustrative Example 19. According to the non-transitory computer-readable storage medium of Illustrative Examples 15-18, receiving patient data from one or more patient databases includes: retrieving the patient data from the one or more patient databases using the first machine learning model.

[0113] Illustrative Example 20. A non-transitory computer-readable storage medium according to Illustrative Examples 15-19, wherein the first machine learning model includes a first LLM (Large Language Model), the second machine learning model includes a second LLM, and the third machine learning model includes a third LLM.

Claims

1. A computer-implemented method, comprising: Receive (202) patient data from one or more patient databases (102); Using a first machine learning model, (204) imaging protocol formulation information is extracted from the patient data; Using a second machine learning model, an imaging protocol template (206) is selected from multiple candidate imaging protocol templates based on the extracted imaging protocol formulation information; The selected imaging protocol template (208) is fine-tuned using a third machine learning model based on the patient data; and Output (210) a finely tuned imaging protocol template.

2. The computer-implemented method according to claim 1, wherein using a second machine learning model to select an imaging protocol template from multiple candidate imaging protocol templates based on the extracted imaging protocol formulation information includes: The extracted text of the proposed imaging protocol information and the descriptions of the multiple candidate imaging protocol templates are encoded into vectors; Determine a similarity measure between the vector of the extracted imaging protocol formulation information text and the vector of the description of each of the plurality of candidate imaging protocol templates; as well as Based on the similarity metric, a candidate imaging protocol template is selected from the vectors that are most similar to the vectors with the extracted imaging protocol formulation information.

3. The computer-implemented method of claim 1, wherein using a third machine learning model to fine-tune the selected imaging protocol template based on the patient data comprises: Using the third machine learning model, the selected imaging protocol template is fine-tuned based on the patient data and protocol formulation standards of the clinical location.

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

5. The computer-implemented method according to claim 1, further comprising: Identify clusters of imaging protocols in systems at one or more clinical sites; Identify the patient scenario that triggered the cluster; as well as Based on the identified patient scenario, one or more of the multiple candidate imaging protocol templates are automatically generated.

6. The computer-implemented method according to claim 1, wherein outputting the fine-tuned imaging protocol template comprises: The selected imaging protocol template is presented after fine-tuning, and the fine-tuned selected imaging protocol template is annotated with information from the patient data that proves the rationality of the protocol selection and the rationality of the fine-tuning.

7. The computer-implemented method of claim 1, wherein extracting imaging protocol formulation information from the patient data using a first machine learning model comprises: Using the first machine learning model, information related to the formulation of imaging protocols in the patient data is summarized.

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

9. The computer-implemented method of claim 1, wherein the first machine learning model includes a first LLM (Large Language Model), the second machine learning model includes a second LLM, and the third machine learning model includes a third LLM.

10. An apparatus comprising: A component for receiving (202) patient data from one or more patient databases (102); A component for extracting (204) imaging protocol formulation information from the patient data using a first machine learning model; The component used to select (206) imaging protocol templates from multiple candidate imaging protocol templates based on the extracted imaging protocol formulation information using a second machine learning model; A component for fine-tuning (208) the selected imaging protocol template based on the patient data using a third machine learning model; as well as Components used to output (210) a finely tuned imaging protocol template.

11. The apparatus of claim 10, wherein the component for selecting an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocol formulation information using a second machine learning model comprises: A component used to encode the text of the extracted imaging protocol formulation information and the descriptions of the multiple candidate imaging protocol templates into vectors; A component for determining the similarity measure between the vector of the text of the extracted imaging protocol formulation information and the vector of the description of each of the plurality of candidate imaging protocol templates; as well as A component for selecting, based on the similarity metric, a candidate imaging protocol template that has the vector most similar to the vector with the extracted imaging protocol formulation information.

12. The apparatus of claim 10, wherein the component for fine-tuning the selected imaging protocol template based on the patient data using a third machine learning model comprises: A component for fine-tuning the selected imaging protocol template using the third machine learning model, based on the patient data and clinical location protocol formulation criteria.

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

14. The apparatus of claim 10, further comprising: A component used to identify clusters of imaging protocols in systems at one or more clinical sites; Components used to identify the patient scenario that triggered the cluster; as well as Used to automatically generate one or more components from the plurality of candidate imaging protocol templates based on the identified patient scenario.

15. A non-transitory computer-readable storage medium including instructions that, when executed by a computer, cause the computer to perform operations, the operations including: Receive (202) patient data from one or more patient databases (102); Using a first machine learning model, (204) imaging protocol formulation information is extracted from the patient data; Using a second machine learning model, an imaging protocol template (206) is selected from multiple candidate imaging protocol templates based on the extracted imaging protocol formulation information; The selected imaging protocol template (208) is fine-tuned using a third machine learning model based on the patient data; and Output (210) a finely tuned imaging protocol template.

16. The non-transitory computer-readable storage medium of claim 15, wherein using a second machine learning model to select an imaging protocol template from a plurality of candidate imaging protocol templates based on the extracted imaging protocol formulation information includes: The extracted text of the proposed imaging protocol information and the descriptions of the multiple candidate imaging protocol templates are encoded into vectors; Determine a similarity measure between the vector of the extracted imaging protocol formulation information text and the vector of the description of each of the plurality of candidate imaging protocol templates; as well as Based on the similarity metric, a candidate imaging protocol template is selected from the vectors that are most similar to the vectors with the extracted imaging protocol formulation information.

17. The non-transitory computer-readable storage medium of claim 15, wherein outputting the fine-tuned imaging protocol template comprises: The selected imaging protocol template is presented after fine-tuning, and the fine-tuned selected imaging protocol template is annotated with information from the patient data that proves the rationality of the protocol selection and the rationality of the fine-tuning.

18. The non-transitory computer-readable storage medium of claim 15, wherein extracting imaging protocol formulation information from the patient data using a first machine learning model comprises: The first machine learning model is used to summarize information related to the formulation of imaging protocols in the patient data.

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

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