Automated documentation of medical imaging examinations

Abstract

Techniques for automated documentation of medical imaging exams are disclosed. Acquired medical images are stored in an image buffer. A quality metric is determined for at least a portion of the medical images, and the quality metric is compared to a first threshold value. A content metric is determined for the at least a portion of the medical images, the content metric including an abnormality score characterizing at least one abnormality in the at least a portion of the medical images, and the content metric is compared to a second threshold value. At least a subset of the medical images is identified as representative images based on the first threshold value and the second threshold value, and the representative images are stored outside the buffer. The quality metric and / or the content metric can be generated using a neural network.

Description

[0001] 2024PF00252

[0002] AUTOMATED DOCUMENTATION OF MEDICAL IMAGING EXAMINATIONS

[0003] GOVERNMENT INTEREST

[0004]

[0001] This invention was made with United States government support awarded by the United States Department of Health and Human Services under the grant number HHS / ASPR / BARDA 75A50120C00097. The United States has certain rights in this invention.

[0005] TECHNICAL FIELD

[0006]

[0002] The present disclosure relates to evaluating medical images and documenting medical imaging examinations. More specifically, this application relates to automated documentation of medical imaging exams, such as point-of-care (POC) ultrasound exams.

[0007] BACKGROUND

[0008]

[0003] Various medical imaging modalities can be used for clinical analysis and medical intervention, as well as visual representation of the function of organs and tissues, such as magnetic resonance imaging, ultrasound, or computed tomography. For example, point-of- care (POC) ultrasound exams may be performed in an emergency room or intensive care unit to quickly evaluate a condition or status of a subject (e.g., a patient). While consultative ultrasound may be performed, for example, by a sonographer, POC exams are typically performed by physicians (e.g., emergency physicians, intensivists, anaesthesiologists, internists, or the like). POC ultrasound exams may be very short in duration, sometimes taking less than a minute.

[0009]

[0004] Documentation of medical imaging exams can refer to identification and retention of at least a portion of acquired medical images and generation / retention of relevant data. Complete and accurate documentation can be important for providing quality care to subjects and allowing for communication among medical teams and / or to subjects. Examples of documentation include reports generated to document an exam. A report may serve as an instrument of communication between providers, subjects, and / or other users, and provides 2024PF00252 information about conditions of a subject at a particular time, which may inform follow-up procedures and exams. Documentation may include static images or videos / cineloops, which may be included because they are representative of the exam, such as by including particular views, target anatomies or objects, abnormalities, or the like. Various data may be included in documentation, such as subject identifying information, facility identifying information, date, time, clinician information, anatomy or other object identification, annotations, labels, bounding boxes, image orientation, description of exams and / or procedures, clinician comments, subject reactions or complications, measurements, findings, diagnoses, recommendations, and so forth.

[0010] SUMMARY

[0011]

[0005] Apparatuses, systems, and methods for automated documentation of medical imaging exams are disclosed. For example, the disclosed technology can automatically identify representative images acquired during an ultrasound exam, and the disclosed technology can automatically generate labels or other annotations for the representative images.

[0012] In accordance with at least one example disclosed herein, methods are disclosed for automated documentation of medical imaging exams. A plurality of medical images is stored in an image buffer. For example, the buffer can be a first-in first-out (FIFO) buffer representing a predetermined time period for acquisition of at least some of the plurality of medical images. A quality metric is determined for at least a portion of the plurality of medical images, and the quality metric is compared to a first threshold value. A content metric is determined for the at least a portion of the plurality of medical images, the content metric including an abnormality score characterizing at least one abnormality in the at least a portion of the plurality of medical images. The content metric is compared to a second threshold value. At least a subset of the plurality of medical images is identified as representative images based on the first threshold value and the second threshold value. For example, the representative images can be identified when they are determined to be of sufficient quality and when they contain desired image content, such as at least one 2024PF00252 abnormality. In some implementations, at least one annotation is automatically generated for the at least a subset of the plurality of medical images, such as an abnormality label or another content label. In various embodiments, the at least a subset of the plurality of medical images is automatically formatted. In these and other embodiments, the at least a subset of the plurality of medical images is stored as a cineloop. The representative images are stored, for example, in a non-transitory computer-readable medium outside the buffer.

[0013]

[0006] In various implementations, the quality metric, the content metric, or both are generated using a neural network. In these and other implementations, the neural network is trained using a training dataset comprising medical images, associated annotations, and at least one of quality metrics or content metrics. In various implementations, the methods include training one or more neural networks.

[0014]

[0007] In various implementations, the methods further include acquiring the plurality of medical images via an ultrasound imaging system, and the at least a subset of the plurality of medical images is identified as the representative images in real time during the medical imaging exam. In these and other implementations, a medical exam type is determined for the medical imaging exam based at least in part on an exam preset, and the quality metric, the content metric, or both are based at least in part on the medical exam type.

[0015]

[0008] In various implementations, identifying the at least a subset of the plurality of medical images as the representative images includes determining an elapsed time after storing previous representative images, and the at least a subset of the plurality of medical images is identified as the representative images when the elapsed time meets or exceeds a threshold time.

[0016]

[0009] In various implementations, identifying the at least a subset of the plurality of medical images as the representative images includes determining a correlation coefficient between the at least a subset of the plurality of medical images and previous representative images, and the at least a subset of the plurality of medical images is identified as the representative images when the correlation is below a threshold value.

[0017]

[0010] In various implementations, identifying the at least a subset of the plurality of medical images as the representative images includes determining a first view of the at least 2024PF00252 a subset of the plurality of medical images and comparing the first view to a second view of previous representative images, and the at least a subset of the plurality of medical images is identified as the representative images when the first view is different than the second view.

[0018] [Oil] In various implementations, identifying the at least a subset of the plurality of medical images as the representative images includes identifying a first subset of the plurality of medical images having respective abnormality labels and identifying a second subset of the plurality of medical images not having respective abnormality labels.

[0019]

[0012] In accordance with at least one example disclosed herein, the disclosed technology includes a system configured to perform one or more disclosed methods. In some implementations, the system comprises a medical imaging system, which may be an ultrasound imaging system including an ultrasound imaging probe configured to acquire the medical images. In some implementations, the system comprises a computing system including at least one processor and at least one memory carrying instructions configured to cause the computing system to perform one or more methods disclosed herein.

[0020]

[0013] In accordance with at least one example disclosed herein, a non-transitory computer- readable medium is disclosed carrying instructions that, when executed, cause a processor to execute operations to perform at least one method disclosed herein.

[0021] BRIEF DESCRIPTION OF THE DRAWINGS

[0022]

[0014] FIG. 1 is a block diagram of an ultrasound imaging system arranged in accordance with principles of the present disclosure.

[0023]

[0015] FIG. 2 is a flow diagram illustrating a workflow for automated documentation of medical imaging exams, according to principles of the present disclosure.

[0024]

[0016] FIG. 3 is a block diagram illustrating a neural network, according to principles of the present disclosure.

[0025]

[0017] FIG. 4 is a flow diagram illustrating a process for automated documentation of medical imaging exams, according to principles of the present disclosure.

[0026] 1 2024PF00252

[0027] DESCRIPTION

[0028]

[0018] The following description of certain examples is illustrative in nature and is in no way intended to limit the disclosed technology or its applications or uses. In the following detailed description of examples of the present apparatuses, systems, and methods, reference is made to the accompanying drawings, which form a part hereof, and in which are shown by way of illustration specific examples in which the described apparatuses, systems and methods may be practiced. These examples are described in sufficient detail to enable those skilled in the art to practice the presently disclosed apparatuses, systems, and methods, and it is to be understood that other examples may be utilized and that structural and logical changes can be made without departing from the spirit and scope of the present disclosure. Moreover, for the purpose of clarity, detailed descriptions of certain features will not be discussed when they would be apparent to those with skill in the art, so as not to obscure the description of the present system. The following detailed description is therefore not to be taken in a limiting sense, and the scope of the present technology is defined only by the appended claims.

[0029]

[0019] Users of medical imaging systems, such as ultrasound imaging systems, face technical challenges related to documenting medical imaging exams. For example, existing systems may require manual or substantially manual documentation of exams, such that a user may be required to select representative images, generate annotations or other data, and / or prepare a report for an exam. These systems are inefficient and time-consuming. Additionally, existing technologies may not provide sufficient storage to save all or substantially all medical images from an exam for later review and documentation. Due to these and other shortcomings of existing systems, users may simply choose to omit documentation of an exam, which contributes to the problem of “ghost scans” - that is, exams that are not documented, annotated, recorded, saved, billed, or the like. As a result, existing systems may be unable to efficiently provide the benefits of exam documentation, such as availability of exam results for reference and comparison to other exams (e.g., future imaging exams of the same subject) and the ability to bill for exams. These problems are likely to occur, for example, in relation to POC exams, which may be performed by a 2024PF00252 physician during a very short time period (e.g., a minute or less). Due to the technical shortcomings of existing systems, physicians frequently fail to document POC exams, and POC exams may not be stored or reported for future use or billing purposes. Accordingly, there exists a need for technologies that address the foregoing problems and provide other benefits.

[0030]

[0020] Disclosed herein are systems and related methods for automated documentation of medical imaging exams. The disclosed technology automatically identifies and stores representative images (e.g., as cineloops) of an exam that can be used to create an exam report, rather than requiring users to manually identify the representative images. In some implementations, the disclosed technology generates annotations or other data for the representative images, such as labels. The representative images can be identified to provide representative information about a condition of a subject (e.g., a patient). Representative images can be identified based on various factors, including but not limited to image quality, presence of a target anatomy or other object, viewing angle, presence of one or more abnormalities, and so forth. The representative images can be used to create a report after the exam has ended, without having to interrupt the scan procedure for image selection and annotation. Moreover, the disclosed technology can identify the representative images in real time (e.g., in seconds or less and / or while an exam is occurring).

[0031]

[0021] To identify representative images, the disclosed technology may include, use, and / or train one or more artificial intelligence (Al) models, which can be neural networks (e.g., deep learning networks and / or convolutional neural networks) and / or other machine learning models. For example, a model can be trained using training datasets comprising representative images for a particular exam type (e.g., lung ultrasound, cardiac ultrasound, abdominal ultrasound). The training datasets may include additional data, such as labels or other annotations, quality metrics (e.g., scores), indications of normality or abnormality, and so forth. Once a model has been trained, the model receives data comprising acquired medical images, and the model outputs representative images selected from the acquired medical images. Additionally or alternatively, one or more models can be trained and used to generate outputs used to select representative images, such as quality metrics or content 2024PF00252 metrics characterizing the medical images. The model may further output associated data, such as labels or other annotations. As used herein, a “model” can refer to a construct that is trained using training data to make predictions or provide probabilities for new data items, whether or not the new data items were included in the training data.

[0032]

[0022] For example, training data for supervised learning can include items with various parameters and an assigned classification. A new data item can have parameters that a model can use to assign a classification to the new data item. As another example, a model can be a probability distribution resulting from the analysis of training data, such as a likelihood of an n-gram occurring in a given language based on an analysis of a large corpus from that language. Examples of models and / or associated techniques include, without limitation: neural networks, support vector machines, decision trees, Parzen windows, Bayes, clustering, reinforcement learning, probability distributions, decision trees, decision tree forests, and others. Models can be configured for various situations, data types, sources, and output formats.

[0033]

[0023] A model trained or provided by the disclosed system can include a neural network with multiple input nodes that receive training datasets. The input nodes can correspond to functions that receive the input and produce results. These results can be provided to one or more levels of intermediate nodes that each produce further results based on a combination of lower-level node results. A weighting factor can be applied to the output of each node before the result is passed to the next layer node. At a final layer, (“the output layer,”) one or more nodes can produce a value classifying the input that, once the model is trained, can be used to evaluate new data (e.g., to evaluate medical imaging data to identify representative images, generate quality metrics, generate content metrics). In some implementations, such neural networks, known as deep neural networks, can have multiple layers of intermediate nodes with different configurations, can be a combination of models that receive different parts of the input and / or input from other parts of the deep neural network, or are convolutions — partially using output from previous iterations of applying the model as further input to produce results for the current input. 2024PF00252

[0034]

[0024] A model can be trained with supervised learning (e.g., self-supervised). Testing data can then be provided to the model to assess accuracy. Testing data can be, for example, a portion of the entire dataset (e.g., 10%) held back to use for evaluation of the model. Output from the model can be compared to the desired or expected output for the training data and, based on the comparison, the model can be modified, such as by changing weights between nodes of the neural network and / or parameters of the functions used at each node in the neural network (e.g., applying a loss function). Based on the results of the model evaluation, and after applying the described modifications, the model can then be retrained to evaluate new data.

[0035]

[0025] Advantages of the disclosed technology may include, without limitation, improved and / or more efficient documentation of medical imaging exams. Additionally, the disclosed technology can conserve resources (e.g., technological resources), such as by reducing the amount of storage needed to retain medical images for subsequent review and documentation. As a result of the technical improvements of the disclosed technology, clinicians may be able to generate documentation in a shorter amount of time, which allows clinicians to spend more time providing patient care. Additionally, the disclosed technology may provide more accurate documentation by automatically selecting images that are most representative of a subject’s condition and / or the results of an exam. While example implementations are described with reference to POC ultrasound imaging exams, the disclosed technology can also be applied to other kinds of ultrasound imaging exams, such as consultative ultrasound imaging exams. Additionally, while example embodiments are described with reference to ultrasound imaging exams, the disclosed technology can also be applied to other kinds of medical imaging exams, such as computed tomography or magnetic resonance imaging exams.

[0036]

[0026] FIG. 1 is a block diagram of an ultrasound imaging system 100 arranged in accordance with principles of the present disclosure. In the ultrasound imaging system 100 of FIG. 1, an ultrasound probe 112 includes a transducer array 114 for transmitting ultrasonic waves and receiving echo information. The transducer array 114 can be implemented as a linear array, convex array, a phased array, and / or a combination thereof. The transducer array 2024PF00252

[0037] 114, for example, can include a two-dimensional array (as shown) of transducer elements capable of scanning in both elevation and azimuth dimensions for 2D and / or 3D imaging. The transducer array 114 can be coupled to a microbeamformer 116 in the probe 112, which controls transmission and reception of signals by the transducer elements in the array. In this example, the microbeamformer 116 is coupled by the probe cable to a transmit / receive (T / R) switch 118, which switches between transmission and reception and protects the main beamformer 122 from high-energy transmit signals. In some embodiments, the T / R switch 118 and other elements in the system can be included in the ultrasound probe 112 rather than in a separate ultrasound system base. In some embodiments, the ultrasound probe 112 may be coupled to the ultrasound imaging system via a wireless connection (e.g., WiFi, Bluetooth).

[0038]

[0027] The transmission of ultrasonic beams from the transducer array 114 under control of the microbeamformer 116 is directed by the transmit controller 120 coupled to the T / R switch 118 and the beamformer 122, which receives input from the user’s operation of the user interface (e.g., control panel, touch screen, console) 125. The user interface 125 may include soft and / or hard controls. One of the functions controlled by the transmit controller 120 is the direction in which beams are steered. Beams may be steered straight ahead from (orthogonal to) the transducer array 114, or at different angles for a wider field of view. The partially beamformed signals produced by the microbeamformer 116 are coupled via channels 115 to a main beamformer 122 where partially beamformed signals from individual patches of transducer elements are combined into a fully beamformed signal. In some embodiments, microbeamformer 116 is omitted and the transducer array 114 is coupled via channels 115 to the beamformer 122. In some embodiments, the system 100 can be configured (e.g., include a sufficient number of channels 115 and have a transmit / receive controller programmed to drive the transducer array 114) to acquire ultrasound data responsive to a plane wave or diverging beams of ultrasound transmitted toward the subject. In some embodiments, the number of channels 115 from the ultrasound probe may be less than the number of transducer elements of the transducer array 114 and the system can be 2024PF00252 operable to acquire ultrasound data packaged into a smaller number of channels than the number of transducer elements.

[0039]

[0028] The beamformed signals are coupled to a signal processor 126. The signal processor 126 can process the received echo signals in various ways, such as bandpass filtering, decimation, I and Q component separation, and / or harmonic signal separation. The signal processor 126 can also perform additional signal enhancement such as speckle reduction, signal compounding, and noise elimination. The processed signals are coupled to a B-mode processor 128, which can employ amplitude detection for the imaging of structures in the body. The signals produced by the B-mode processor 128 are coupled to a scan converter 130 and a multiplanar reformatter 132. The scan converter 130 arranges the echo signals in the spatial relationship from which they were received in a desired image format. For instance, the scan converter 130 can arrange the echo signal into a two-dimensional (2D) sector-shaped format, or a pyramidal three-dimensional (3D) image. The multiplanar reformatter 132 can convert echoes, which are received from points in a common plane in a volumetric region of the body into an ultrasonic image of that plane, as described in U.S. Pat. No. 6,443,896 (Detmer).

[0040]

[0029] A volume Tenderer 134 converts the echo signals of a 3D data set into a projected 3D image as viewed from a given reference point, e.g., as described in U.S. Pat. No. 6,530,885 (Entrekin et al.) The 2D or 3D images can be coupled from the scan converter 130, multiplanar reformatter 132, and volume Tenderer 134 to at least one processor 137 for further image processing operations. For example, the at least one processor 137 can include an image processor 136 configured to perform further enhancement and / or buffering and temporary storage of imaging data for display on an image display 138. The display 138 can include a display device implemented using a variety of display technologies, such as LCD, LED, OLED, or plasma display technology. The at least one processor 137 can include a graphics processor 140, which can generate graphic overlays for display with the ultrasound images. These graphic overlays can contain, e.g., standard identifying information such as patient name, date and time of the image, imaging parameters, and the like. For these purposes the graphics processor 140 receives input from the user interface 125, such as a 2024PF00252 typed patient name. The user interface 125 can also be coupled to the multip lanar reformatter 132 for selection and control of a display of multiple multiplanar reformatted (MPR) images.

[0041]

[0030] The user interface 125 can include one or more mechanical controls, such as buttons, dials, a trackball, a physical keyboard, and others, which may also be referred to herein as hard controls. Alternatively or additionally, the user interface 125 can include one or more soft controls, such as buttons, menus, soft keyboard, and other user interface control elements implemented for example using touch-sensitive technology (e.g., resistive, capacitive, or optical touch screens). One or more of the user controls can be co-located on a control panel 124. For example one or more of the mechanical controls can be provided on a console and / or one or more soft controls can be co-located on a touch screen, which can be attached to or integral with the console. The display 138 and the user interface 125 can be included in an I / O component, via which outputs are provided by the system 100 and / or inputs are received by the system 100.

[0042]

[0031] In some implementations, the user interface 125 can provide one or more outputs of the disclosed system, including representative images from a medical imaging exam and / or associated data, such as annotations, labels, quality metrics, content metrics, and so forth. In some implementations the one or more outputs of the system can be provided via one or more graphical user interfaces (e.g., via the display 138). The user interface 125 also includes controls to select one or more exam presets corresponding to respective exam types, such as exam presets for lung ultrasound, cardiac ultrasound, abdominal ultrasound, or the like. The exam presets, when selected, configure the ultrasound imaging system 100 to acquire images for a respective exam type, such as by configuring settings (e.g., imaging parameters) for the ultrasound probe 112, the beamformer 122, the transmit controller 120 and / or other components.

[0043]

[0032] The at least one processor 137 (e.g., the image processor 136, the graphics processor 140, or a different processor) can also perform functions associated with automated documentation of medical imaging exams, as described herein. For example, the at least one processor 137 can evaluate medical imaging data to identify representative images. In various embodiments, the at least one processor 137 implements one or more neural

[0044] II 2024PF00252 networks to evaluate medical imaging data, as disclosed herein. In some implementations, the identification of the representative images is based at least in part on an exam preset. For example, when an exam preset is selected (e.g., via the user interface 125) for lung ultrasound, the disclosed system identifies representative images for a lung ultrasound exam type.

[0045]

[0033] Although described as separate processors, it will be understood that the functionality of any of the processors described herein can be implemented in a single processor (e.g., a CPU or GPU implementing the functionality of processor 137) or fewer number of processors than described in this example. In some embodiments, the at least one processor 137 can be hardware-based (e.g., include multiple layers of interconnected nodes implemented in hardware). In some embodiments, the at least one processor 137 can be implemented at other processing stages, e.g., prior to the processing performed by the image processor 136, volume Tenderer 134, multiplanar reformatter 132, and / or scan converter 130. In some embodiments, the at least one processor 137 can be implemented to process ultrasound data in the channel domain, beamspace domain (e.g., before or after beamformer 122), the IQ domain (e.g., before, after, or in conjunction with signal processor 126), and / or the k-space domain. As described, in some embodiments, functionality of two or more of the processing components (e.g., beamformer 122, signal processor 126, B-mode processor 128, scan converter 130, multiplanar reformatter 132, volume Tenderer 134, at least one processor 137, image processor 136, graphics processor 140, etc.) can be combined into a single processing unit and / or divided between multiple processing units. The processing units can be implemented in software, hardware, or a combination thereof. For example, the at least one processor 137 can include one or more graphical processing units (GPU). In another example, beamformer 122 can include an application specific integrated circuit (ASIC).

[0046]

[0034] The at least one processor 137 can be coupled to one or more computer-readable media (e.g., memory 142) included in the system 100, which can be non-transitory. The one or more computer-readable media can carry instructions and / or a computer program that, when executed, cause the at least one processor 137 to perform operations described herein. 2024PF00252

[0047] A computer program can be stored / distributed on any suitable medium, such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but can also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. Furthermore, embodiments can take the form of a computer program product accessible from a computer-readable medium providing program code for use by or in connection with a computer or any device or system that executes instructions.

[0048]

[0035] The one or more computer-readable media, such as memory 142, can store data received or generated by the system 100. For example, the memory 142 may store images generated based on the received ultrasound signals, user inputs (e.g., parameter measurements, patient names), and the like. According to embodiments of the present disclosure, the memory 142 may implement or include one or more buffers. For example, the memory 142 may include one or more first-in-first-out buffers for storing images acquired by the system 100.

[0049]

[0036] For the purposes of this disclosure, a computer-readable medium can generally be any tangible apparatus that can contain, store, communicate, propagate, and / or transport the program for use by or in connection with the instruction execution device. The computer- readable medium can be, for example, without limitation, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, and / or a propagation medium. Nonlimiting examples of a computer readable medium include a semiconductor or solid-state memory, magnetic tape, a removable computer diskette, a random-access memory (RAM), a read-only memory (ROM), a rigid magnetic disk, and / or an optical disk. Optical disks can include compact disk read only memory (CD-ROM), compact disk-read / write (CD-R / W), and / or DVD.

[0050]

[0037] FIG. 2 is a block diagram illustrating a workflow 200 for automated documentation of medical imaging exams. The workflow 200 can be performed, for example, using the system 100 of FIG. 1. Additionally or alternatively, at least a portion of the workflow 200 can be performed using a different computing system to automatically document a medical imaging exam performed using the system 100 of FIG. 1. The disclosed technology implements the workflow 200 to identify and save representative images for documentation 2024PF00252 of a medical imaging exam. In some implementations, the workflow 200 is performed in real time (e.g., while an exam is being performed), while in other implementations the workflow 200 can be performed using stored medical images from a medical imaging exam.

[0051]

[0038] The workflow 200 begins at block 205, where a medical imaging exam is commenced using a medical imaging device (e.g., ultrasound imaging system 100 of FIG. 1). The medical imaging exam can be, for example, a POC ultrasound exam performed by a physician using an ultrasound probe (e.g., 112 of FIG. 1). Upon commencing the medical imaging exam, the medical imaging device begins acquiring imaging data, which includes frames of medical images.

[0052]

[0039] The workflow 200 proceeds to block 210, where a first-in-first-out (FIFO) buffer is filled (or updated) with imaging data acquired using the medical imaging device. For example, the FIFO buffer may store a plurality of image frames representing approximately 3 seconds of imaging data (e.g., the most recently acquired images). While the FIFO buffer may store image frames representing various durations, in some examples the image frames represent 1 to 4 seconds of an imaging exam. In some embodiments, the image frames may be a cineloop. For example, the FIFO buffer can be continuously or iteratively updated as medical images are acquired, analyzed, and saved or discarded. In some embodiments, the FIFO buffer may be implemented in whole or in part by a memory device or a portion of a memory device, such as for example, memory 142.

[0053]

[0040] The workflow 200 proceeds to block 215, where the image data in the FIFO buffer is analyzed. For example, a deep neural network can be applied to the imaging data to compute a quality metric (e.g., for each image frame in the FIFO buffer), such as a numerical score characterizing image quality. The quality metric may indicate a likelihood that a respective image frame is a high-quality diagnostic image based on various criteria. For example, images higher resolution and / or higher signal to noise ratio (SNR) may be considered higher quality than images with lower resolution and / or lower SNR. Images that include the presence of a target anatomy and / or other object of interest may be considered higher quality than images without the target anatomy and / or other object. In some examples, operations at block 215 include applying a single-class classifier network trained using a 2024PF00252 large number (e.g., hundreds, thousands, tens of thousands) of training images for which trained users provided ground-truth annotations of “high quality,” “not high quality,” and / or other characterizations of image quality. During inferencing, the trained network will provide a confidence score (e.g., a numerical score) indicating how likely it is that the given image is a “high quality image.”

[0054]

[0041] The workflow 200 proceeds to block 220, where an overall quality metric (e.g., a cumulative metric) is computed for the set of images in the FIFO buffer. For example, the overall quality metric may comprise or be based on a sum of at least a portion of the quality metrics generated at block 215 for image frames in the FIFO buffer. In some implementations, other values associated with the images in the FIFO buffer can additionally or alternatively be computed, such as average, mean, maximum, or minimum quality values or the like.

[0055]

[0042] The workflow 200 proceeds to decision block 225, where the overall quality metric computed at block 220 is compared to a threshold value. In other words, it is determined at decision block 225 whether the images in the FIFO buffer satisfy a quality threshold for retention as potential representative images for the medical imaging exam. The images may be saved outside the FIFO buffer, for example, in another non-transitory computer-readable medium (e.g., another portion of memory 142). In some embodiments, the images may be discarded from the FIFO buffer after being stored outside the FIFO buffer. If it is determined at decision block 225 that the overall quality metric does not exceed the threshold value, then at least a portion of the medical imaging data (e.g., one or more of the images) in the FIFO buffer is discarded, and the workflow 200 returns to block 210 to fill (or update) the FIFO buffer with new medical imaging data (e.g., more recently acquired images).

[0056]

[0043] If it is determined at decision block 225 that the overall quality metric satisfied the threshold value, then the workflow 200 proceeds to decision block 230, where it is determined whether sufficient time has elapsed since the last save of a one or more images from the FIFO buffer (e.g., a cineloop), which represents a potentially representative image for the medical imaging exam. For example, one or more previous images may have already been saved according to the workflow 200. If a previous image or set of images is too close 2024PF00252 in time (e.g., 1 second, 5 seconds, 10 seconds, 30 seconds), as compared to images presently in the FIFO buffer, then it is determined that sufficient time has not elapsed since the last save of an image or images from the FIFO buffer. If images presently in the FIFO buffer are too close in time to a previous image or images, which has already been saved as a potentially representative image or images, then the present images are likely to be duplicative of the previous image(s) saved from the FIFO buffer, and there is little advantage in retaining the present image(s). Accordingly, if it is determined at decision block 230 that sufficient time has not elapsed since the last save of an image, then the workflow 200 returns to block 210, where at least a portion of the images in the FIFO buffer are discarded and the FIFO buffer is filled or updated to continue the workflow 200.

[0057]

[0044] In some implementations, operations at block 230 additionally or alternatively include evaluating whether a view has changed since a previously-saved or evaluated image, such as by computing a correlation coefficient between images currently stored in the FIFO buffer and images previously stored in the FIFO buffer (e.g., saved as a cineloop). A low correlation value (e.g., 50%, 40%, 30%, 10%, 0%) may indicate a substantial change in view (e.g., when the user switches from a parasternal long axis view of a heart to a parasternal short axis view of the heart), such that the images currently stored in the FIFO buffer are materially different from the images previously stored in the FIFO buffer. Accordingly, images currently stored in the FIFO buffer are more likely to be saved as a representative image because they are not duplicative of previously saved image(s).

[0058]

[0045] If it is determined at decision block 230 that sufficient time has elapsed since the last save of an image or images (or that the images currently stored in the FIFO buffer are substantially different), then the workflow 200 proceeds to block 235, where images in the FIFO buffer are saved to a temporary archive as a cineloop. The cineloop represents one or more potentially representative images for the medical imaging exam, which is temporarily saved for further analysis and, potentially, for long-term retention if additional criteria are met.

[0059]

[0046] The workflow 200 proceeds to block 240, where an abnormality metric is computed for each image frame in the FIFO buffer, such as a numerical score characterizing content 2024PF00252 of a medical image. For example, the abnormality metric can be computed using a neural network to indicate one or more characteristics of content of a respective image frame, such as whether the image frame is likely to include abnormal content. Examples of abnormalities, the presence of which may cause a high abnormality metric (e.g., a score of 60%, 70%, 80%, 90%), include consolidation in a lung ultrasound scan, free fluid in an abdominal ultrasound scan, or cardiomyopathy in a cardiac ultrasound scan. A high abnormality metric may increase the likelihood that an image will be saved as a potentially representative image because the image may depict abnormalities that can be used to document and characterize the condition of a subject, for example.

[0060]

[0047] While an example of an abnormality score is described with reference to the workflow 200, additional or alternative content metrics can be computed to characterize an image and / or a condition depicted in an image. For example, a metric computed at block 240 may relate to image content that is not abnormal, such as a metric that indicates whether a particular view of a target anatomy or object is captured. Table 1 below illustrates nonlimiting examples of image content that may affect an abnormality metric and / or other content metric computed at block 240. Presence and / or degree of presence of these characteristics may increase a value of one or more metrics computed at block 240, which increases the likelihood that an image will be retained as a potentially relevant image. In some implementations, a metric is computed for each of a plurality of categories of characteristics (e.g., abnormality categories). 2024PF00252

[0061]

[0048] An abnormality metric can be computed, for example, using a detection network, such as a network implementing YOLO techniques, which may be trained using a large number (e.g., hundreds, thousands, tens of thousands) of medical images in which the location of abnormalities has been annotated with localization boxes by trained users. Any suitable detection model may be used, including but not limited to YOLO models. An example of a suitable YOLO model is a YOLOv3 model as described in YOLOv3: An Incremental Improvement by Joseph Redmon et al., arXiv: 1804.02767 [cs.CV] (2018), which is incorporated herein by reference for any purpose. When running inferences with the trained network on newly acquired images, the network provides a number of potential detection boxes for the abnormalities with corresponding confidence values indicating how likely the corresponding box is a “true” detection of the abnormality. The highest confidence values in each frame can be used as abnormality scores for a given abnormality or other characteristic.

[0062]

[0049] The workflow 200 proceeds to block 245, where an overall abnormality metric for the images in the FIFO buffer is computed, such as by summing at least a portion of the 2024PF00252 abnormality metrics computed at block 240. While an example of an overall abnormality metric is described, one or more other overall content metrics can be additionally or alternatively be computed to characterize content of the images in the FIFO buffer, such as based on one or more characteristics described in Table 1 above. In some implementations, different overall metrics can be computed for respective categories of characteristics (e.g., abnormality categories).

[0063]

[0050] The workflow 200 proceeds to decision block 250, where the overall metric(s) computed at block 245 is compared to a threshold, such as an abnormality threshold. In other words, it is determined at decision block 250 whether the images in the FIFO buffer satisfy a content threshold (e.g., an abnormality threshold) for retention as potential representative images for the medical imaging exam. If it is determined at decision block 250 that the overall quality metric does not exceed the threshold value, then at least a portion of the medical imaging data (e.g., one or more images) in the FIFO buffer is discarded, and the workflow 200 returns to block 210 to fill (or update) the FIFO buffer.

[0064]

[0051] If it is determined at decision block 250 that the overall metric satisfies the threshold value, then the workflow 200 proceeds to 255, where one or more category labels and / or other annotations, data, or metadata are added to the saved image(s) (e.g., saved at block 235). Having been determined to exceed the threshold at decision block 250, the temporarily saved image or images (e.g., a cineloop) are retained as a potentially representative image(s), and the image(s) are associated with other relevant data and / or metadata, including but not limited to a relevant abnormality category or other category characterizing the image(s), such as a label for an image content category discussed above with reference to Table 1.

[0065]

[0052] The image or images and the associated data and / or metadata is retained, and the workflow 200 may return to block 210 to continue evaluating acquired images in the FIFO buffer (e.g., to identify additional images for retention). The workflow 200 may be repeated and / or performed continuously, for example, until the imaging exam is completed, such as by being terminated by a user, at which time the workflow proceeds to block 260, where the imaging exam ends and no further images are acquired or added to the FIFO buffer. 2024PF00252

[0066]

[0053] The workflow 200 may proceed to block 265, where at least some saved images may be deleted. For example, a desired or maximum number of images or sets of images (e.g., cineloops) and associated data / metadata may be saved as representative images for documenting the medical imaging exam, and remaining images may be discarded for various reasons. In some examples, the number of saved images or sets of images may be based on different image categories (e.g., to ensure that a desired number of images in each of a plurality of categories is retained). In some examples, the number of saved images or sets of images may be based on an amount of available storage. In some examples, a number of saved images (e.g., total saved images or sets of images and / or images or sets of images per category) is based on a predetermined or user-specified limit.

[0067]

[0054] In some implementations, saved images without an abnormality label are sorted according to respective quality metrics, and only images of sufficient quality are retained. For example, 10, 50, or 100 images or sets of images without an abnormality label may be retained according to highest quality metrics, and remaining images or sets of images without an abnormality label are discarded.

[0068]

[0055] In some implementations, saved images or sets of images having abnormality labels or other content labels are grouped according to content characteristics (e.g., abnormality type), and each group is sorted according to quality metrics. In each group, only images or sets of images of sufficient quality are retained. For example, 10, 50, or 100 images or sets of images in each group are retained according to highest quality metrics, and remaining images or sets of images are discarded.

[0069]

[0056] After any excess images or sets of images are deleted, the workflow 200 proceeds to block 270, where the workflow 200 terminates.

[0070]

[0057] After the workflow 200 has completed, the saved images may be further analyzed and / or processed (e.g., using manual or automated processes) in relation to documentation of the medical imaging exam, such as for generating a report. The images may be saved for a preset time period before being discarded (e.g., 24 hours, 48 hours, one week, one month, six months). In some implementations, the workflow 200 generates at least a portion of a report and / or automatically formats saved images according to a report format or other form 2024PF00252 of documentation. In some implementations, the workflow 200 further includes processing the saved images, such as by automatically, cropping, editing, labeling (e.g., time, location, patient identification), annotating (e.g., with text, bounding boxes, icons, images), and / or formatting the images. The processing may be performed, for example, using one or more models, as described herein, which may be trained to automatically process the saved images.

[0071]

[0058] The workflow 200 or operations thereof may be repeated while maintaining a similar functionality, and one or more operations of the workflow 200 may be performed in parallel. Additionally, one or more operations of the workflow 200 may be omitted in some implementations. While the workflow 200 is described with reference to ultrasound imaging (e.g., POC ultrasound imaging), the workflow 200 can also be applied to other imaging modalities or examinations. Additionally, while the example workflow 200 identifies sets of images such as cineloops as representative images in some applications, other embodiments may save still images (e.g., frames) and / or other imaging data.

[0072]

[0059] FIG. 3 is block diagram illustrating a neural network 300 that may be trained and used to analyze imaging data of a medical imaging exam in accordance with principles of the present disclosure. In some examples, the neural network 300 may be implemented by one or more processors of a medical imaging system (e.g., ultrasound imaging system 100 of FIG. 1) to implement an Al model. For example, instructions for implementing the neural network 300 may be stored in a non-transitory computer readable memory such as memory 142, and the instructions may be executed by the one or more processors 137. The neural network 300 can be used, for example, to implement one or more operations of the workflow 200 of FIG. 2, such as to generate image frame-specific and / or overall quality metrics or content metrics (e.g., abnormality metrics), to generate labels or annotations, and / or to automatically process medical imaging data. In some implementations, multiple neural networks 300 may be implemented (e.g., a quality neural network and a content neural network), while in other implementations a single neural network 300 can be trained to receive medical imaging data and output representative images and / or associated data / metadata. In some examples, neural network 300 may be a convolutional neural network (CNN) with single and / or multidimensional layers and / or a deep learning network. 2024PF00252

[0073]

[0060] The neural network 300 may include one or more input nodes 302. In some examples, the input nodes 302 may be organized in a layer of the neural network 300. The input nodes 302 may be coupled to one or more layers 308 of hidden units 306 by weights 304. In some examples, the hidden units 306 may perform operations on one or more inputs from the input nodes 302 based, at least in part, on the associated weights 304. In some examples, the hidden units 306 may be coupled to one or more layers 314 of hidden units 312 by weights 310. The hidden units 312 may perform operations on one or more outputs from the hidden units 306 based, at least in part, on the weights 310. The outputs of the hidden units 312 may be provided to an output node 316 to provide an output (e. g. , inference) of the neural network 300. Although one output node 316 is shown in FIG. 3, in some examples, the neural network 300 may have multiple output nodes 316. In some examples, the output may be accompanied by a confidence level. The confidence level may be a value from, and including, 0 to 1, where a confidence level 0 indicates the neural network 300 has no confidence that the output is correct and a confidence level of 1 indicates the neural network 300 is 100% confident that the output is correct.

[0074]

[0061] In some examples, inputs to the neural network 300 provided at the one or more input nodes 302 may include medical imaging data, such as acquired image frames, which may be in a FIFO buffer and / or included in a cineloop. In some examples, outputs provided at output node 316 may include, for example, image frame quality metrics, overall quality metrics for images stored in a FIFO buffer, normality or abnormality metrics, labels or other annotations, and / or indications of whether images (e.g., frames or cineloops) are representative images of a medical imaging exam. In some examples, multiple neural networks 300 may be used to implement different functions of the disclosed system, such as different portions of the workflow 200 (e.g., a first neural network 300 to generate frame quality metrics and a second neural network 300 to generate abnormality metrics). In these and other implementations, a single neural network 300 can be used to implement substantially all of the workflow 200, such that the neural network 300 continuously evaluates images in the FIFO buffer and outputs a set of representative images and / or associated data. 2024PF00252

[0075]

[0062] Although a convolutional neural network has been described herein, this machine learning model has been provided only as an example, and the principles of the present disclosure are not limited to this particular model.

[0076]

[0063] FIG. 4 is a block diagram illustrating a process 400 for automated documentation of a medical imaging exam, according to principles of the present disclosure. The process 400 can be performed using at least a portion of a medical imaging system (e.g., ultrasound imaging system 100 of FIG. 1) or a different computing system. The process 400 can implement at least a portion of the workflow 200 of FIG. 2, and at least a portion of the process 400 can be performed using one or more neural networks or other Al models (e.g., neural network 300 of FIG. 3).

[0077]

[0064] The process 400 begins at block 410, where medical images (e.g., a set of image frames) are stored in a buffer, which may be a first-in first-out (FIFO) representing a time period of a medical imaging exam (e.g., 1 second, 3 seconds, 5 seconds, 10 seconds). The medical images may be stored in the buffer as they are acquired, or they may be retrieved from a memory (e.g., after a medical imaging exam has been completed). The medical images are acquired via a medical imaging device, such as an ultrasound imaging system (e.g., 100 of FIG. 1), and the medical images are acquired as part of a medical imaging exam, such as an ultrasound exam. The medical images are stored in the buffer for evaluation to determine whether they are representative images for documentation of the medical imaging exam. In some embodiments, the buffer may be implemented by a non-transitory computer- readable medium, such as memory 142. In some implementations, additional data is accessed with the medical images, such as settings information or other imaging parameters, subject (e.g., patient) information, user (e.g., physician or sonographer) information, date and time information, exam type information, and so forth.

[0078]

[0065] In some implementations, the medical images are associated with an exam type, which may be determined based on a selected exam preset for the medical imaging device. For example, a user may select an exam preset via a user interface (e.g., 125 of FIG. 1) of the medical imaging system associated with an exam type, such as lung ultrasound, 2024PF00252 abdominal ultrasound, or cardiac ultrasound. In some implementations, the process 400 includes determining the exam type based on the exam preset.

[0079]

[0066] In some implementations, one or more portions of the process 400 may be performed continuously and / or iteratively. In these and other implementations, the buffer may be continuously and / or periodically updated as medical images are evaluated and stored or discarded according to the process 400 (e.g., according to the workflow 200 of FIG. 2). For example, at least a portion of the medical images can be removed from the buffer and replaced with new or updated medical images when medical images are stored as a cineloop, or discarded based on a quality metric, a content metric, or another criterion.

[0080]

[0067] The process 400 proceeds to block 420, where a quality metric is determined for the medical images in the buffer. For example, a quality metric can be computed for each of a plurality of image frames based on image clarity, brightness, contrast, visibility of target anatomy or objects, image orientation, or other image characteristics. In some implementations, determining the quality metric includes determining quality metrics for image frames of the medical images and determining an overall quality metric for at least a portion of the medical images in the buffer, which may include all or substantially all of the image frames in the buffer. The overall quality metric can include, for example, a sum, average, or median quality metric. The quality metric can be determined, for example, using a neural network or other Al model, such as the neural network 300 of FIG. 3.

[0081]

[0068] The process 400 proceeds to block 430, where the quality metric determined at block 420 is compared to a first threshold value. For example, a frame-specific and / or overall quality metric can be compared to a threshold value to determine whether one or more medical images are of sufficient quality for documentation of the medical imaging exam. If the quality metric does not satisfy the threshold value, then at least a portion of the medical images can be discarded and the buffer can be updated with new medical images for further evaluation (e.g., computation of quality metrics for the new medical images).

[0082]

[0069] The process 400 proceeds to block 440, where a content metric is determined for the medical images in the buffer. In some implementations, the content metric is determined only for medical images for which the quality metric determined at block 420 exceeds the 2024PF00252 threshold value evaluated at block 430. The content metric can include, for example, one or more abnormality metrics characterizing presence or absence of one or more abnormalities in the medical images. Additionally or alternatively, the content metric can include one or more other content metrics characterizing presence or absence of other image content. Examples of image content that may be characterized using content metrics are described with reference to Table 1 above. The content metric is determined at block 440 to evaluate whether the medical images in the buffer include content of a representative image - in other words, whether the content of the image is such that the medical images are representative of the medical imaging exam. Examples of representative content include presence of one or more abnormalities that should be documented, presence of a target anatomy or object, one or more views of a target anatomy, object, or abnormality, and so forth. The content metric can be determined, for example, using a neural network or other Al model, such as the neural network 300 of FIG. 3. In some implementations, a single model is used to determine the quality metric and the content metric, while in other implementations a quality model is used to determine the quality metric and a content model is used to determine the content metric. In various implementations, one or more models are selected and / or trained to determine the quality metric or the content metric. For example, different models can be selected and used to detect different abnormalities or other image content, such as models selected based on an exam type.

[0083]

[0070] In some implementations, one or more content metrics are determined at block 440 based on an exam type, such as based on whether the medical imaging exam is for evaluating a subject’s lungs, heart, or abdomen.

[0084]

[0071] The process 400 proceeds to block 450, where the content score generated at block 440 is compared to a second threshold value. For example, an image frame-specific and / or overall content metric can be compared to a threshold value to determine whether one or more medical images include content desired or needed for documentation of the medical imaging exam. If the content metric does not satisfy the threshold value, then at least a portion of the medical images can be discarded and the buffer can be updated with new 2024PF00252 medical images for further evaluation (e.g., computation of quality and / or content metrics for the new medical images).

[0085]

[0072] In some implementations, the process 400 includes evaluating the medical images in the buffer based on other characteristics, such as whether the medical images are duplicative or likely to be duplicative of medical images that have already been identified as representative images. For example, if the medical images are determined to satisfy at least one of the first threshold value or the second threshold value (e.g., because the quality and / or content of the medical images is sufficient for documentation), then the medical images can be evaluated to determine whether sufficient time has passed since an earlier set of medical images was identified and stored as representative images for the medical imaging exam. If sufficient time has not passed (e.g., more than one second, five seconds, ten seconds, 30 seconds), then the medical images currently in the buffer may be discarded because they are likely to be duplicative of the earlier-identified images. Additionally or alternatively, a correlation can be determined between content of medical images currently in the buffer and the earlier-identified images to determine whether the content is likely to be duplicative. In some implementations, a first view of the at least a subset of the plurality of medical images is identified, and the first view is compared to a second view of previous representative images. The at least a subset of the plurality of medical images is identified as the representative images when the first view is determined to be different than the second view. For example, the first and second views can be determined to have different image content based at least in part on respective content metrics.

[0086]

[0073] The process 400 proceeds to block 460, where at least a subset of the medical images is identified as representative images for documentation of the medical imaging exam. The subset can be identified based at least in part on the quality metric determined at block 420 and the content metric determined at block 440. For example, the subset can be identified as all or substantially all images in the buffer after discarding any images that fail to satisfy the first threshold value or the second threshold value. When identified, the representative images may be stored in a non-transitory computer-readable medium, which may be outside the buffer. 2024PF00252

[0087]

[0074] In some implementations, the process 400 includes training one or more neural networks or other Al models, such as a quality model and a content model. For example, a training dataset can be generated comprising a plurality of medical images and associated data, such as annotations, labels, quality scores, content scores (e.g., abnormality scores), and so forth. The training dataset can be used to train the one or more neural networks to generate quality metrics, content metrics, labels or other annotations, identify representative images, and so forth.

[0088]

[0075] In some implementations, the process 400 includes automatically generating one or more annotations for the identified subset of the medical images. For example, a label can be generated based on one or more abnormality metrics exceeding the threshold value evaluated at block 450 to indicate that the identified subset of the medical images includes an identified abnormality. Other annotations may include bounding boxes, text, subject information, identification of a target anatomy or object, and so forth.

[0089]

[0076] In some implementations, the process 400 includes automatically formatting the identified subset of the medical images. For example, the identified subset can be stored as and / or used to generate a cineloop. In other examples, the identified subset of the medical images can be automatically formatted as a draft report to be completed by a physician or other user. Additionally or alternatively, the identified subset can be cropped, rotated, or otherwise edited, such as by changing image settings.

[0090]

[0077] In some implementations, identifying the subset of the medical images as representative images includes identifying a first subset having respective abnormality labels and identifying a second subset not having respective abnormality labels.

[0091]

[0078] In various implementations, the process 400 includes providing one or more outputs via a graphical user interface, such as by displaying quality and / or content metrics and displaying the identified subset of medical images for further evaluation by the user. For example, the process 400 can include generating and displaying a graphical user interface to allow the user to select images from the identified subset to be stored as representative medical images. 2024PF00252

[0092]

[0079] In various embodiments, operations of the process 400 can be performed continuously and / or iteratively, such as to update images in the buffer as images are saved or discarded and to evaluate the updated buffer on a continuous or periodic basis. In these and other embodiments, the process 400 or operations thereof can be repeated until a stopping criterion is met. For example, the process 400 may include automatically terminating the process 400 when a sufficient number of representative images and / or cineloops have been identified or saved.

[0093]

[0080] The process 400 can be performed in any order, including performing one or more operations in parallel and / or repeating one or more operations. Additionally, operations can be added to or removed from the process 400 without deviating from the teachings of the present disclosure.

[0094]

[0081] Advantageously, systems and related methods disclosed herein can automatically document medical imaging exams, including automatically identifying representative images based on various factors including image quality and image content (e.g., presence of abnormalities or other image content). At least a portion of the systems and methods is completely automated, such as by training and / or using one or more neural networks or other Al models, such that a user need not manually identify and characterize (e.g., label or annotate) representative images for inclusion in a report or other documentation. The systems and methods provide various technical advantages and improvements, including improved accuracy and efficiency when documenting exams and conservation of technical resources (e.g., by reducing storage required to retain medical images for later review and analysis).

[0095]

[0082] In various examples where components, systems and / or methods are implemented using a programmable device, such as a computer-based system or programmable logic, it should be appreciated that the above-described systems and methods can be implemented using any of various known or later developed programming languages, such as “Python”, “C”, “C++”, “FORTRAN”, “Pascal”, “VHDL” and the like. Accordingly, various storage media, such as magnetic computer disks, optical disks, electronic memories and the like, can be prepared that can contain information that can direct a device, such as a computer, to

[0096] ■ 2024PF00252 implement the above-described systems and / or methods. Once an appropriate device has access to the information and programs contained on the storage media, the storage media can provide the information and programs to the device, thus enabling the device to perform functions of the systems and / or methods described herein. For example, if a computer disk containing appropriate materials, such as a source file, an object file, an executable file or the like, were provided to a computer, the computer could receive the information, appropriately configure itself and perform the functions of the various systems and methods outlined in the diagrams and flowcharts above to implement the various functions. That is, the computer could receive various portions of information from the disk relating to different elements of the above-described systems and / or methods, implement the individual systems and / or methods and coordinate the functions of the individual systems and / or methods described above.

[0097]

[0083] In view of this disclosure it is noted that the various methods and devices described herein can be implemented in hardware, software, and / or firmware. Further, the various methods and parameters are included by way of example only and not in any limiting sense. In view of this disclosure, those of ordinary skill in the art can implement the present teachings in determining their own techniques and needed equipment to affect these techniques, while remaining within the scope of the invention. The functionality of one or more of the processors described herein may be incorporated into a fewer number or a single processing unit (e.g., a CPU) and may be implemented using application specific integrated circuits (ASICs) or general-purpose processing circuits which are programmed responsive to executable instructions to perform the functions described herein.

[0098]

[0084] Although the present system may have been described with particular reference to an ultrasound imaging system, it is also envisioned that the present system can be extended to other medical imaging systems where one or more images are obtained in a systematic manner. Accordingly, the present system may be used to obtain and / or record image information related to, but not limited to renal, testicular, breast, ovarian, uterine, thyroid, hepatic, lung, musculoskeletal, splenic, cardiac, arterial and vascular systems, as well as other imaging applications related to ultrasound-guided interventions. Further, the present 2024PF00252 system may also include one or more programs which may be used with conventional imaging systems so that they may provide features and advantages of the present system. Certain additional advantages and features of this disclosure may be apparent to those skilled in the art upon studying the disclosure, or may be experienced by persons employing the novel systems and methods of the present disclosure. Another advantage of the present systems and method may be that conventional medical image systems can be easily upgraded to incorporate the features and advantages of the present systems, devices, and methods.

[0099]

[0085] Of course, it is to be appreciated that any one of the examples, examples or processes described herein may be combined with one or more other examples, examples and / or processes or be separated and / or performed amongst separate devices or device portions in accordance with the present systems, devices and methods.

[0100]

[0086] Finally, the above discussion is intended to be merely illustrative of the present systems and methods and should not be construed as limiting the appended claims to any particular example or group of examples. Thus, while the present system has been described in particular detail with reference to exemplary examples, it should also be appreciated that numerous modifications and alternative examples may be devised by those having ordinary skill in the art without departing from the broader and intended spirit and scope of the present systems and methods as set forth in the claims that follow. Accordingly, the specification and drawings are to be regarded in an illustrative manner and are not intended to limit the scope of the appended claims.

Claims

2024PF00252CLAIMSWhat is claimed is:

1. A computer-implemented method of automatically documenting a medical imaging exam, the method comprising: storing a plurality of medical images in an image buffer; determining a quality metric for at least a portion of the plurality of medical images; comparing the quality metric to a first threshold value; determining a content metric for the at least a portion of the plurality of medical images, wherein the content metric includes an abnormality score characterizing at least one abnormality in the at least a portion of the plurality of medical images; comparing the content metric to a second threshold value; identifying at least a subset of the plurality of medical images as representative images based on the first threshold value and the second threshold value; and storing the representative images in a non-transitory computer-readable medium outside the buffer.

2. The computer-implemented method of claim 1, wherein at least one of the quality metric or the content metric are determined using a neural network.

3. The computer-implemented method of claim 2, wherein the neural network is trained using a training dataset comprising medical images, associated annotations, and at least one of quality metrics or content metrics.

4. The computer-implemented method of claim 1, further comprising: acquiring the plurality of medical images via an ultrasound imaging system, wherein the at least a subset of the plurality of medical images is identified as the representative images in real time during the medical imaging exam.

5. The computer-implemented method of claim 1, further comprising:2024PF00252 determining a medical exam type for the medical imaging exam based at least in part on an exam preset, wherein the quality metric, the content metric, or both are based at least in part on the medical exam type.

6. The computer-implemented method of claim 1, further comprising: automatically generating at least one annotation for the at least a subset of the plurality of medical images; and automatically formatting the at least a subset of the plurality of medical images.

7. The computer-implemented method of claim 1, wherein identifying the at least a subset of the plurality of medical images as the representative images includes determining an elapsed time after storing previous representative images, and wherein the at least a subset of the plurality of medical images is identified as the representative images when the elapsed time meets or exceeds a threshold time.

8. The computer-implemented method of claim 1, wherein identifying the at least a subset of the plurality of medical images as the representative images includes determining a correlation coefficient between the at least a subset of the plurality of medical images and previous representative images, and wherein the at least a subset of the plurality of medical images is identified as the representative images when the correlation is below a threshold value.

9. The computer-implemented method of claim 1 , wherein identifying the at least a subset of the plurality of medical images as the representative images includes determining a first view of the at least a subset of the plurality of medical images and comparing the first view to a second view of previous representative images, and wherein the at least a subset of the plurality of medical images is identified as the representative images when the first view is different than the second view.2024PF0025210. The computer-implemented method of claim 1 , wherein identifying the at least a subset of the plurality of medical images as the representative images includes identifying a first subset of the plurality of medical images, the first subset having respective abnormality labels, and identifying a second subset of the plurality of medical images, the second subset not having respective abnormality labels.

11. A computing system comprising: at least one processor; and at least one non-transitory memory carrying instructions that, when executed, cause the computing system to: store a plurality of medical images in an image buffer; determine a quality metric for at least a portion of the plurality of medical images; compare the quality metric to a first threshold value; determine a content metric for the at least a portion of the plurality of medical images, wherein the content metric includes an abnormality score characterizing at least one abnormality in the at least a portion of the plurality of medical images; compare the content metric to a second threshold value; identify at least a subset of the plurality of medical images as representative images based on the first threshold value and the second threshold value; and store the representative images in a non-transitory computer-readable medium outside the buffer.

12. The system of claim 11, wherein the image buffer is a first-in first-out (FIFO) buffer representing a predetermined time period for acquisition of at least some of the plurality of medical images.

13. The computing system of claim 11, wherein at least one of the quality metric or the content metric are determined using a neural network.2024PF0025214. The computing system of claim 13, wherein the neural network is trained using a training dataset comprising medical images, associated annotations, and at least one of quality metrics or content metrics.

15. The computing system of claim 11 further comprising an ultrasound imaging probe, wherein the instructions further cause the system to: acquire the plurality of medical images via the ultrasound imaging probe, wherein the at least a subset of the plurality of medical images is identified as the representative images in real time during a medical imaging exam.

16. The computing system of claim 15, wherein the instructions further cause the system to: determine a medical exam type for the medical imaging exam based at least in part on an exam preset, wherein the quality metric, the content metric, or both are based at least in part on the medical exam type.

17. The computing system of claim 11, wherein the instructions further cause the system to: automatically generate at least one annotation for the at least a subset of the plurality of medical images.

18. The computing system of claim 11 , wherein the instructions further cause the computing system to: automatically format the at least a subset of the plurality of medical images.

19. The computing system of claim 11 , wherein the instructions further cause the computing system to: store the at least a subset of the plurality of medical images as a cineloop.2024PF0025220. The computing system of claim 11, wherein identifying the at least a subset of the plurality of medical images as the representative images causes the computing system to identify a first subset of the plurality of medical images, the first subset having respective abnormality labels, and identify a second subset of the plurality of medical images, the second subset not having respective abnormality labels.

Images