System and method of estimating gestational age using blind sweep ultrasound imaging

Abstract

A method for estimating gestational age of a fetus using ultrasound imaging includes performing blind sweeps of a subject to obtain blind sweep ultrasound image data; determining whether the blind sweep ultrasound image data includes standard plane biometry image data of fetal anatomy using a trained first machine learning model for identifying standard planes; when the first machine learning model identifies a standard plane, inputting the standard plane to a trained second machine learning model that determines a standard plane measurement of the fetal anatomy in the identified standard plane, and estimating the gestational age of the fetus from the standard plane measurement; and when the first machine learning model does not identify a standard plane, inputting the blind sweep ultrasound image data to a trained third machine learning model that estimates the gestational age of the fetus directly from the blind sweep ultrasound image data.

Description

2024PF00288- PATENT -SYSTEM AND METHOD OF ESTIMATING GESTATIONAL AGE USING BLIND SWEEP ULTRASOUND IMAGINGFIELD OF THE INVENTION

[0001] The invention relates to the field of antenatal ultrasound imaging, and more specifically to estimating gestational age based on blind sweep ultrasound image data.BACKGROUND OF THE INVENTION

[0002] Antenatal ultrasound screening is routinely conducted early in pregnancy for monitoring the health of the mother and developing fetus. One important objective of antenatal ultrasound screening is estimating gestational age of the fetus. Standard-of-care clinical practice for tracking growth and development of the fetus involves acquisition of standard planes of certain fetal anatomies to perform biometric measurements, and relating the measurements to the gestational age of the fetus. Relating the measurements to the gestational age may be based on known formulas or look-up tables, such as a Hadlock table, for example.

[0003] In the first trimester of pregnancy, the gestational age may be estimated using measurements of crown-rump length (CRL) in specific cross-sectional sagittal views of the long axis of the fetus. In the second and third trimesters, the gestational age may be estimated using measurements of biparietal diameter (BPD), head circumference (HC), femur length (FL) and abdominal circumference (AC) in standard planes acquired of the head, femur and abdomen of the fetus. The accuracy of these estimates decreases as the gestational age increases. For example, accuracy is highest earlier in the first trimester and lowest later in the third trimester. For example, accuracy of the estimated gestational age may be within 5-7 days when the actual gestational age is less than 8 6 / 7 weeks, and more than 21 days when the actual gestation age is greater than 28 weeks.

[0004] Additionally, acquisition of standard planes and subsequent measurements on these standard planes using ultrasound imaging require a high level of clinical expertise in ultrasound imaging. Unfortunately, availability of skilled ultrasound operators capable of making such measurements in low and middle income countries (LMIC) and even in remote / rural areas of2024PF00288- PATENT - developed countries is often scarce. Novice ultrasound operators typically cannot reliably acquire the standard plane images needed to make biometric measurements on the fetal anatomy.

[0005] Attempts have been made to address this problem, and to make gestational age estimation (and other basic pregnancy -related parameters) available to a wider section of the world population. For example, a set of orthogonal ultrasound sweeps may be acquired by a non-expert ultrasound operator, which are referred to as “blind sweeps” since they are not guided by a search for specific anatomic features. Measurements based on conventional deep learning (DL) models may be used to estimate gestational age from the ultrasound images obtained by the blind sweeps. However, gestational age estimation from blind sweeps using measurement-based conventional DL models is unreliable. For example, using conventional DL models, only 55 percent of the blind sweeps contain at least one standard plane (e.g., for HC, BPD, AC or FL) on which biometric measurements are possible. When the threshold is raised, only 22 percent of blind sweeps contain at least two standard planes (i.e., 2 out of HC, BPD, AC or FL standard planes), and only 5 percent contain at least 3 standard planes (i.e., 3 out of HC, BPD, AC or FL standard planes). An alternate approach is to check for the presence of standard planes in the ultrasound images obtained by blind sweeps, and to measure fetal anatomies only when such standard planes are present. Otherwise, the gestational ages cannot be determined.SUMMARY OF THE INVENTION

[0006] In a representative embodiment, a method is provided for estimating gestational age of a fetus using ultrasound imaging. The method includes performing blind sweeps of a subject using an ultrasound probe to obtain blind sweep ultrasound image data from ultrasound images of fetal anatomy of the fetus; inputting the blind sweep ultrasound image data to a trained first machine learning model for determining whether the blind sweep ultrasound image data includes standard planes for performing biometric measurements of the fetal anatomy; when the trained first machine learning model identifies at least one standard plane in the blind sweep ultrasound image data, inputting the identified at least one standard plane to a trained second machine learning model that determines at least one biometric measurement of the fetal anatomy in the at least one identified standard plane, and estimating the gestational age of the fetus from the at least one biometric measurement; when the trained first machine learning model does not2024PF00288- PATENT - identify at least one standard plane in the blind sweep ultrasound image data, inputting the blind sweep ultrasound image data to a trained third machine learning model that estimates the gestational age of the fetus directly from the blind sweep ultrasound image data; and displaying the estimated gestational age of the fetus.

[0007] In another representative embodiment, a system is provided for estimating gestational age of a fetus carried by a subject using ultrasound imaging. The system includes an ultrasound imaging system including an ultrasound controller and an ultrasound probe operable by an operator and configured to obtain blind sweep ultrasound images from a blind sweep of an abdomen of the subject; a display; at least one processor coupled to the ultrasound imaging system and the display; and at least one non-transitory memory coupled to the at least processor. The at least one non-transitory memory stores instructions that, when executed by the at least one processor, cause the at least one processor to receive the blind sweep ultrasound image data from ultrasound images of fetal anatomy of the fetus; input the blind sweep ultrasound image data to a trained first machine learning model for determining whether the blind sweep ultrasound image data includes standard planes for performing biometric measurements of the fetal anatomy; when the trained first machine learning model identifies at least one standard plane in the blind sweep ultrasound image data, input the identified at least one standard plane to a trained second machine learning model that determines at least one biometric measurement of the fetal anatomy in the at least one identified standard plane, and estimate the gestational age of the fetus from the at least one biometric measurement determined by the trained second machine learning model; when the trained first machine learning model does not identify at least one standard plane in the blind sweep ultrasound image data, input the blind sweep ultrasound image data to a trained third machine learning model that estimates the gestational age of the fetus directly from the blind sweep ultrasound image data; and cause the estimated gestational age of the fetus to be displayed on the display.

[0008] In another representative embodiment, a non-transitory computer readable medium stores instructions for estimating gestational age of a fetus carried by a subject using ultrasound imaging. When executed by at least one processor, the instructions cause the at least one processor to receive the blind sweep ultrasound image data from ultrasound images obtained by performing a blind sweep acquisition of fetal anatomy of the fetus; and input the blind sweep2024PF00288- PATENT - ultrasound image data to a trained first machine learning model for determining whether the blind sweep ultrasound image data includes standard planes for performing biometric measurements of the fetal anatomy; when the trained first machine learning model identifies at least one standard plane in the blind sweep ultrasound image data, input the identified at least one standard plane to a trained second machine learning model that determines at least one biometric measurement of the fetal anatomy in the at least one identified standard plane, and estimate the gestational age of the fetus from the at least one biometric measurement determined by the trained second machine learning model; when the trained first machine learning model does not identify at least one standard plane in the blind sweep ultrasound image data, input the blind sweep ultrasound image data to a trained third machine learning model that estimates the gestational age of the fetus directly from the blind sweep ultrasound image data; and cause the estimated gestational age of the fetus to be displayed on the display.BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The example embodiments are best understood from the following detailed description when read with the accompanying drawing figures. It is emphasized that the various features are not necessarily drawn to scale. In fact, the dimensions may be arbitrarily increased or decreased for clarity of discussion. Wherever applicable and practical, like reference numerals refer to like elements.

[0010] FIG. l is a simplified block diagram of a system for estimating gestational age of a fetus using ultrasound imaging, according to a representative embodiment.

[0011] FIG. 2 is a flow diagram of a method for identifying standard planes in blind sweep ultrasound image data using a trained first machine learning model, according to a representative embodiment.

[0012] FIG. 3 is a flow diagram of a method for measuring standard planes in blind sweep ultrasound image data using a trained second machine learning model, according to a representative embodiment.

[0013] FIG. 4 is a flow diagram of a method for directly estimating gestational age in blind sweep ultrasound image data using a trained third machine learning model, according to a representative embodiment.2024PF00288- PATENT -

[0014] FIG. 5 is a flow diagram of a method for estimating gestational age of a fetus using ultrasound imaging, according to a representative embodiment.DETAILED DESCRIPTION OF EMBODIMENTS

[0015] Aspects of the disclosure may be supported by various information technology (IT) backends, including either or both local architectures, either as monoliths, networked, or a combination thereof, and hosted architectures, such as a software as a service (SaaS), platform as a service (PaaS), and / or infrastructure as a service (laaS), or the like. In an example, a supporting infrastructure includes multiple interconnected layers respectively hosting, as an abstraction, various IT processes, services, accounts, and other management components.

[0016] Any of the steps described in relation to examples and / or training described below can be performed by a specific-purpose computer system or general-purpose computer system, or a computer-readable medium, or data carrier system configured to carry out any of the steps described previously. The computer system can include a set of software instructions that can be executed to cause the computer system to perform any of the methods or computer-based functions disclosed herein. The computer system may operate as a standalone device or may be connected, for example using a network, to other computer systems or peripheral devices. As an example, a computer system performs logical processing based on digital signals received via an analogue-to-digital converter.

[0017] In the following detailed description, for the purposes of explanation and not limitation, representative embodiments disclosing specific details are set forth in order to provide a thorough understanding of an embodiment according to the present teachings. Descriptions of known systems, devices, materials, methods of operation and methods of manufacture may be omitted so as to avoid obscuring the description of the representative embodiments. Nonetheless, systems, devices, materials and methods that are within the purview of one of ordinary skill in the art are within the scope of the present teachings and may be used in accordance with the representative embodiments. It is to be understood that the terminology used herein is for purposes of describing particular embodiments only and is not intended to be limiting. The defined terms are in addition to the technical and scientific meanings of the defined terms as commonly understood and accepted in the technical field of the present teachings.2024PF00288- PATENT -

[0018] It will be understood that, although the terms first, second, third, etc. may be used herein to describe various elements or components, these elements or components should not be limited by these terms. These terms are only used to distinguish one element or component from another element or component. Thus, a first element or component discussed below could be termed a second element or component without departing from the teachings of the inventive concept.

[0019] The terminology used herein is for purposes of describing particular embodiments only and is not intended to be limiting. As used in the specification and appended claims, the singular forms of terms “a,” “an” and “the” are intended to include both singular and plural forms, unless the context clearly dictates otherwise. Additionally, the terms “comprises,” “comprising,” and / or similar terms specify the presence of stated features, elements, and / or components, but do not preclude the presence or addition of one or more other features, elements, components, and / or groups thereof. As used herein, the term “and / or” includes any and all combinations of one or more of the associated listed items.

[0020] Unless otherwise noted, when an element or component is said to be “connected to,” “coupled to,” or “adjacent to” another element or component, it will be understood that the element or component can be directly connected or coupled to the other element or component, or intervening elements or components may be present. That is, these and similar terms encompass cases where one or more intermediate elements or components may be employed to connect two elements or components. However, when an element or component is said to be “directly connected” to another element or component, this encompasses only cases where the two elements or components are connected to each other without any intermediate or intervening elements or components.

[0021] The present disclosure, through one or more of its various aspects, embodiments and / or specific features or sub-components, is thus intended to bring out one or more of the advantages as specifically noted below. For purposes of explanation and not limitation, example embodiments disclosing specific details are set forth in order to provide a thorough understanding of an embodiment according to the present teachings. However, other embodiments consistent with the present disclosure that depart from specific details disclosed herein remain within the scope of the appended claims. Moreover, descriptions of well-known apparatuses and methods may be omitted so as to not obscure the description of the example2024PF00288- PATENT - embodiments. Such methods and apparatuses are within the scope of the present disclosure.

[0022] Generally, the various embodiments described herein provide a system and method for estimating gestational age of a fetus using blind sweep ultrasound image data regardless of whether the blind sweeps of the pregnant subject include imaging of standard planes. As stated above, a little over half of blind sweeps end up containing at least one standard plane, which can therefore be leveraged to obtain fetal anatomy measurements. For the remaining blind sweeps, according to the various embodiments, gestational age estimation may be done directly from the acquired blind sweep ultrasound image data using a trained deep learning (DL) machine learning model, despite the absence of standard planes. In other words, the embodiments leverage this redundancy to provide a solution that first checks for the presence of standard planes, and when at least one standard plane exists, estimates the gestational age of the fetus using biometric measurements of the fetal anatomy in the standard planes. When no standard planes are found, the solution reverts to the trained machine learning model and estimates the gestational age of the fetus directly from the blind sweep ultrasound image data. This approach gives the user the option of obtaining the gestational age estimate via a preferred, clinically accepted method when possible, and determining the gestational age estimate using the machine learning model in cases where the clinically accepted method is not possible. In a hybrid solution, the gestational age is estimated using both the biometric measurements from the standard planes and the machine learning model, where the final estimated gestational age of the fetus is some combination (e.g., weighted average) of the two results. The embodiments thus reflect improvement in the technical field of ultrasound imaging, as well as medical treatment and diagnostics.

[0023] FIG. l is a simplified block diagram of a system for estimating gestational age of a fetus using ultrasound imaging, according to a representative embodiment.

[0024] Referring to FIG. 1, system 100 includes a workstation 105 for implementing and / or managing the processes described herein with regard to estimating gestational age of a fetus using ultrasound images from an ultrasound imaging system 140. The workstation 105 includes one or more processors indicated by processor 120, one or more memories indicated by memory 130, a user interface 122 and a display 124. The processor 120 communicates with the ultrasound imaging system 140 through an imaging interface (not shown). The ultrasound imaging system 140 includes an ultrasound controller 143 and an ultrasound probe 145 operable2024PF00288- PATENT - by an operator to obtain ultrasound images of the abdomen of a subject (pregnant patient) 150. The ultrasound probe 145 may be manipulated manually by the ultrasound operator, automatically by a robot under control of a robot controller (not shown), or a combination of both.

[0025] The ultrasound probe 145 may include a 2D matrix array of transducer elements, capable of scanning in two or three dimensions, for example, for emitting ultrasound waves into the body of the subject 150 and receiving echo signals in response. The transducer elements may include capacitive micromachined ultrasonic transducers (CMUTs) or piezoelectric transducers formed of materials such as lead zirconate titanate (PZT) or polyvinylidene difluoride (PVDF), for example, although other types of transducer material may be incorporated without departing from the scope of the present teachings. The transducer array may be coupled to a microbeamformer in the ultrasound probe 145, which controls transmission and reception of signals by the transducer elements.

[0026] The ultrasound probe 145 is connected to the ultrasound controller 143 via a probe cable 147. The ultrasound controller 143 is configured to control the ultrasound imaging process, and includes known elements for performing ultrasound imaging, such as a transmit / receive (T / R) switch configured to switch between transmission and reception modes, and a main beamformer configured to provide final beamforming. One of the functions performed by the ultrasound controller 143 is the direction in which beams are steered and focused. For example, beams may be steered straight ahead from (orthogonal to) the transducer array of the ultrasound probe 145, or at different angles for a wider field of view. Generally, the transmitting of ultrasound signals and the receiving and processing of echo signals in response is known, and therefore additional detail in this regard is not included herein. In various embodiments, all or part of the functionality of the ultrasound controller 143 may be implemented by the processor 120.

[0027] For purposes of discussion, it may be assumed the ultrasound operator of the ultrasound probe 145 is inexperienced and / or lacks proper training for capturing ultrasound images of standard planes that may be used to measure the fetal anatomy with sufficient precision to enable estimation of the gestational age of the fetus, as discussed above. Therefore, the ultrasound probe 145 is used to perform a blind sweep acquisition, including blind sweeps across the subject 150, in order to acquire blind sweep ultrasound images. The blind sweep acquisition may consist of2024PF00288- PATENT - performing six or more predefined freehand motions (e.g., three or more vertical sweeping motions, three or more horizontal sweeping motions, or similar motions) with the ultrasound probe 145 across the subject’s abdomen, referred to as blind sweeps, as would be apparent to one skilled in the art.

[0028] The memory 130 stores instructions executable by the processor 120. When executed, the instructions cause the processor 120 to implement one or more processes for performing a gestational age estimation of a fetus using blind sweep ultrasound images acquired by the ultrasound imaging system 140. The ultrasound images may be provided from the ultrasound imaging system 140 in real-time or near real-time during the scanning procedure, or may be retrieved from storage following the scanning procedure. For purposes of illustration, the memory 130 is shown to include software modules, each of which includes the instructions, executable by the processor 120, corresponding to an associated capability of the system 100.

[0029] The processor 120 is representative of one or more processing devices, and may be implemented by a general purpose computer, a central processing unit (CPU), a digital signal processor (DSP), a graphical processing unit, a computer processor, a microprocessor, a state machine, programmable logic device, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), or combinations thereof, using any combination of hardware, software, firmware, hard-wired logic circuits, or combinations thereof. Any processor or processing unit herein may include multiple processors, parallel processors, or both. Multiple processors may be included in, or coupled to, a single device or multiple devices. The term “processor” as used herein encompasses an electronic component able to execute a program or machine executable instruction. A processor may also refer to a collection of processors within a single computer system or distributed among multiple computer systems, such as in a cloudbased or other multi-site application. Programs have software instructions performed by one or multiple processors that may be within the same computing device or which may be distributed across multiple computing devices.

[0030] The memory 130 may include main memory and / or static memory, where such memories may communicate with each other and the processor 120 via one or more buses. The memory 130 may be implemented by any number, type and combination of random access memory (RAM) and read-only memory (ROM), for example, and may store various types of information,2024PF00288- PATENT - such as software algorithms, artificial intelligence (Al) machine learning models, and computer programs, all of which are executable by the processor 120. The various types of ROM and RAM may include any number, type and combination of computer readable storage media, such as a disk drive, flash memory, an electrically programmable read-only memory (EPROM), an electrically erasable and programmable read only memory (EEPROM), registers, a hard disk, a removable disk, tape, compact disk read only memory (CD-ROM), digital versatile disk (DVD), floppy disk, Blu-ray disk, a universal serial bus (USB) drive, or any other form of storage medium. The memory 130 is a tangible storage medium for storing data and executable software instructions, and is non-transitory during the time software instructions are stored therein. As used herein, the term “non-transitory” is to be interpreted not as an eternal characteristic of a state, but as a characteristic of a state that will last for a period. The term “non-transitory” specifically disavows fleeting characteristics such as characteristics of a carrier wave or signal or other forms that exist only transitorily in any place at any time. The memory 130 may store software instructions and / or computer readable code that enable performance of various functions. The memory 130 may be secure and / or encrypted, or unsecure and / or unencrypted.

[0031] The system 100 may also include a database 112 for storing information that may be used by the various software modules of the memory 130. For example, the database 112 may include image data from previously obtained ultrasound images of the subject 150 and / or of other similarly situated subjects. The stored image data may be used for training an Al machine learning model, such as a neural network model, for example, as discussed below. The database 112 may be implemented by any number, type and combination of RAM and ROM, for example. The various types of ROM and RAM may include any number, type and combination of computer readable storage media, such as a disk drive, flash memory, EPROM, EEPROM, registers, a hard disk, a removable disk, tape, CD-ROM, DVD, floppy disk, Blu-ray disk, USB drive, or any other form of storage medium known in the art. The database 112 comprises tangible storage mediums for storing data and executable software instructions and is non- transitory during the time data and software instructions are stored therein. The database 112 may be secure and / or encrypted, or unsecure and / or unencrypted. For purposes of illustration, the database 112 is shown as a separate storage medium, although it is understood that it may be combined with and / or included in the memory 130, without departing from the scope of the2024PF00288- PATENT - present teachings.

[0032] The processor 120 may include or have access to an artificial intelligence (Al) engine, which may be implemented as software that provides artificial intelligence (e.g., deep learning, neutral network models) and applies machine learning described herein. The Al engine may reside in any of various components in addition to or other than the processor 120, such as the memory 130, an external server, and / or the cloud, for example. When the Al engine is implemented in a cloud, such as at a data center, for example, the Al engine may be connected to the processor 120 via the internet or other communication network using one or more wired and / or wireless connect! on(s). In various embodiments, all or part of the processes provided by first, second and third machine learning models, discussed below, may be implemented by the Al engine, for example. The training and execution of the first, second and third machine learning models cannot practically be performed in the human mind.

[0033] The user interface 122 is configured to provide information and data output by the processor 120, the memory 130 and / or the ultrasound imaging system 140 to the user and / or for receiving information and data input by the user. That is, the user interface 122 enables the user to enter data and to control or manipulate aspects of the processes described herein, and also enables the processor 120 to indicate the effects of the user’s input, which may include control or manipulation of the ultrasound probe 145 when the user / operator. All or a portion of the user interface 122 may be implemented by a graphical user interface (GUI), such as GUI 128 viewable on the display 124, discussed below. The user interface 122 may include one or more interface devices, such as a mouse, a keyboard, a trackball, a joystick, a microphone, a video camera, a touchpad, a touchscreen, voice or gesture recognition captured by a microphone or video camera, for example.

[0034] The display 124 may be a monitor such as a computer monitor, a television, a liquid crystal display (LCD), an organic light emitting diode (OLED), a flat panel display, a solid-state display, or a cathode ray tube (CRT) display, or an electronic whiteboard, for example. The display 124 includes a screen 126 for viewing ultrasound images of the subject 150, along with various features described herein to communicate to the user the degree of image degradation, if any, as well as the GUI 128 to enable the user to interact with the displayed images and features. In an embodiment, the ultrasound imaging system 140 may include a separate dedicated display2024PF00288- PATENT - for acquiring the ultrasound images, where dedicated display is also represented by the display 124.

[0035] Referring to the memory 130, the various modules store sets of data and instructions executable by the processor 120 to estimate gestational age of a fetus using blind sweep ultrasound image data. The memory 130 includes ultrasound image module 131, which is configured to receive and process ultrasound images acquired by the ultrasound probe 145 of the ultrasound imaging system 140 to provide corresponding ultrasound image data. The ultrasound images are obtained from a set of blind sweeps of the subject’s abdomen using the ultrasound probe 145, as discussed above, and may be referred to as blind sweep ultrasound images.

[0036] The blind sweep ultrasound images may be received in real-time or near real-time from the ultrasound imaging system 140, e.g., during a contemporaneous imaging session of the subject 150. The blind sweep ultrasound images may be displayed on the display 124, if desired, although blind sweep ultrasound imaging does not require display. The display of real-time images would enable the operator to visualize to some extent the anatomy of the subject 150 while performing the blind sweeps using the ultrasound probe 145. Alternatively, or in addition, the blind sweep ultrasound images may be previously acquired images, obtained during a previous imaging session of the subject 150 and retrieved from storage (e.g., database 112). The ultrasound image module 131 may also store data associated with the ultrasound images, such as time and date of image acquisition, identification of the ultrasound imaging system 140, and identification of the ultrasound operator who acquired the ultrasound images using the ultrasound imaging system 140.

[0037] Standard plane module 132 is configured to automatically determine whether the blind sweep ultrasound image data acquired using the ultrasound probe 145 have captured one or more standard planes for making biometric measurements of the fetal anatomy. A standard plane is a specific orientation or view used to visualize a specific anatomic structure within the fetus’ body. Each standard plane may be an image frame. Automatically determining whether the blind sweep ultrasound image data includes one or more standard planes involves recognizing landmarks corresponding to key portions of anatomic structures in the ultrasound image data, and classifying the standard planes based on the recognized landmarks associated with the standard planes.2024PF00288- PATENT -

[0038] To make this determination, the standard plane module 132 provides a first machine learning model for identifying standard planes in the blind sweep ultrasound image data. Generally, the first machine learning model receives the blind sweep ultrasound image data from the blind sweeps as input, and provides an indication of whether any standard planes have been identified as output, as discussed below. In addition, or alternatively, the first machine learning model may output the standard planes identified in the blind sweep ultrasound image data whenever the determination is made that the standard planes are present in the blind sweep ultrasound image data. The first machine learning model may be implemented as any suitable type of trainable machine learning model (e.g., deep learning model), such as a convolutional neural network (CNN), an artificial neural network (ANN), a recurrent neural network (RNN), a vision transformer, or a U-net model, for example. Deep learning models refer to neural network models with large numbers of layers and parameters, which directly enable tasks such as classification and regression, for example, as would be apparent to one skilled in the art.

[0039] The first machine learning model is previously trained in a supervised fashion using a first training data set including labelled ultrasound image data from thousands of ultrasound images of fetal anatomies that include standard planes and non-standard planes. Anatomic structures of the fetal anatomies relevant to determining gestational age of fetuses include the head, the femur(s) and the abdomen. Accordingly, the first training data set for training the first machine learning model includes ultrasound image data from ultrasound images showing standard plane landmarks with relevant anatomic structures of fetuses including the head, the femur(s) and / or the abdomen, ultrasound image data from ultrasound images showing nonstandard planes, and labels expressly identifying the standard planes, the relevant anatomic structures in the standard planes, and relevant anatomic structures in the non-standard planes. The non-standard plane data, in particular, is needed to train the first machine learning model to understand where not to activate and learn standard plane information more easily. The labels may be added by experts in ultrasound imaging, such as radiologist and sonographers, for example, or may be labeled automatically through techniques such as ground truth automation, as would be apparent to one skilled in the art. The labeled first training data set may be referred to as ground truth data. The first training data set for training the first machine learning model may be stored in the database 112, for example.2024PF00288- PATENT -

[0040] The first machine learning model is thus trained end-to-end using the first training data set to detect standard planes in the blind sweep ultrasound image data by identifying landmarks, e.g., using object detection, associated with standard planes of the fetal anatomic structures. In an embodiment, the first machine learning model may use bounding box regression loss (e.g., mean squared error loss) as the loss function, for example, during training. After being trained on the first training data set, the first machine learning model is able to process new ultrasound images and predict corresponding standard planes based on the same.

[0041] FIG. 2 is a flow diagram of a method for identifying standard planes in blind sweep ultrasound image data using the trained first machine learning model provided by the standard plane module 132, according to a representative embodiment. That is, during inference, the trained first machine learning model receives as input the blind sweep ultrasound image data from the blind sweeps in block S211. In block S212, the trained first machine learning model searches the blind sweep ultrasound image data for predetermined anatomic landmarks, and in block S213, identifies standard planes corresponding to recognized landmarks (if any). In block S214, the trained first machine learning model outputs an indication of whether any standard planes have been identified. In optional block S215, the trained first machine learning model may also output the standard planes themselves identified in the blind sweep ultrasound image data whenever the determination is made that the standard planes are present in the blind sweep ultrasound image data. In this case, the first machine learning model may indicate the presence of the standard planes simply by outputting the same.

[0042] Biometric measurement module 133 is accessed when the trained first machine learning model of the standard plane module 132 determines that the blind sweep ultrasound image data includes at least one standard plane for making biometric measurements of the fetal anatomy. The biometric measurement module 133 is configured to automatically determine biometric measurements of anatomic structures in the fetal anatomy in the at least one identified standard plane. Automatically determining the biometric measurements involves identifying boundaries of the relevant fetal anatomy and performing measurements across the identified boundaries, discussed below.

[0043] Calculated GA estimate module 134 is also accessed when the trained first machine learning model of the standard plane module 132 determines that the blind sweep ultrasound2024PF00288- PATENT - image data includes at least one standard plane for making biometric measurements of the fetal anatomy. The calculated GA estimate module 134 is configured to estimate the gestational age of the fetus from the biometric measurements provided by the biometric measurement module 133 using an age calculation formula based on predetermined relationships between biometric measurements and gestational age. The age calculation formula may be implemented using established mathematical relationships, tables and / or relational databases (e.g., look-up tables). Examples of the age calculation formula include Hadlock formula, Intergrowth formula, and World Health Organization (WHO) fetal growth formula, although other age calculation formulas may be incorporated without departing from the scope of the present teachings, as would be apparent to one skilled in the art.

[0044] To determine the biometric measurements, the biometric measurement module 133 provides a second machine learning model. Generally, the second machine learning model receives as input the standard planes identified by the first machine learning model, and provides biometric measurements of fetal anatomic structures in each of the standard planes to the calculated GA estimate module 134 as output, as discussed below. The calculated GA estimate module 134 estimates and outputs the gestational age of the fetus using the age calculation formula based on the biometric measurements of the fetal anatomic structures. In an alternative embodiment, the second machine learning model may also include estimating the gestational age using a previously trained age calculation formula. The second machine learning model may be implemented as any suitable type of trainable machine learning model (e.g., deep learning model), such as a CNN, an ANN, an RNN, a vision transformer, or a U-net model, for example.

[0045] The second machine learning model is previously trained in a supervised fashion using a second training data set including labelled ultrasound image data of standard planes from thousands of ultrasound images of fetal anatomies. As mentioned above, fetal anatomic structures relevant to determining gestational age of fetuses include the head, the femur(s) and the abdomen. Accordingly, the second training data set for training the second machine learning model includes ultrasound image data from ultrasound images showing standard plane images of the head, the femur(s), the abdomen and / or cross-sectional sagittal view including boundaries of the same, measurements of the head, the femur(s), the abdomen, and / or the cross-sectional sagittal view of the fetus using the respective boundaries, and labels expressly identifying the2024PF00288- PATENT - respective boundaries and measurements. The manner in which measurements of CRL, BPD, HC, FL and AC of the fetuses may be made is discussed below. The labels may be added by experts in ultrasound imaging, such as radiologist and sonographers, for example, or may be labeled automatically through techniques such as ground truth automation, as would be apparent to one skilled in the art. When the second machine learning model includes estimating the gestational age, the gestational ages corresponding to the measurements are also labeled in the second training data set. The second training data set for training the second machine learning model may be stored in the database 112, for example. In an embodiment, the first and second training data set may be the same training data set, although not necessarily.

[0046] The second machine learning model is thus trained end-to-end to detect and measure anatomic structures in standard planes of the fetal anatomies. In an embodiment, the second machine learning model may use bounding box regression loss (e.g., mean squared error loss) as the loss function, for example. After being trained on the second training data set, the second machine learning model is able to process new standard plane ultrasound images and predict corresponding anatomic structure measurements and gestational ages based on the same.

[0047] FIG. 3 is a flow diagram of a method for measuring standard planes in blind sweep ultrasound image data using the trained second machine learning model provided by the biometric measurement module 133, according to a representative embodiment. That is, during inference, the trained second machine learning model receives as input the standard planes from the blind sweep ultrasound image data output by the first machine learning model in block S311. In block S312, the trained second machine learning model identifies the fetal anatomic structures in the standard plane along with the boundaries (edges) of the anatomic structures, e.g., using edge detection and feature extraction. In block S313, the trained second machine learning model performs biometric measurements of the fetal anatomy in the standard planes using the identified boundaries. In block S314, the trained second machine learning model outputs the biometric measurements to the calculated GA estimate module 134, which estimates the gestational age of the fetus from the biometric measurements using an age calculation formula, as discussed above. As mentioned above, the relevant anatomic structures include the head, the femur(s) and the abdomen of the fetus, and thus the biometric measurements performed and output by the second machine learning model include head circumference, femur length, abdominal circumference and2024PF00288- PATENT - biparietal diameter, and the cross-sectional sagittal view, for example.

[0048] Also as mentioned above, in the first trimester of pregnancy, the gestational age may be estimated using measurements of the CRL in a specific standard plane of the fetus showing a cross-sectional sagittal view . For this, the blind sweep ultrasound image data must include a mid-sagittal view of the fetus, with genital tubercle and fetal spine longitudinally in view. The maximum length from cranium to caudal rump are then be measured as a straight line. In the second and third trimesters, the gestational age may be estimated using measurements of the BPD, the HC, the FL and / or the AC in standard planes acquired of the head, femur and abdomen of the fetus. In particular, the BPD and the HC are measured in a transverse section of the fetus’ head at the level of the thalami and cavum septi pellucidi, while the cerebellar hemispheres is not visible in this scanning plane. The FL is measured with full length of the femur bone perpendicular to the ultrasound beam used to acquire the blind sweep ultrasound image data, excluding the distal femoral epiphysis. The AC is measured in a symmetrical, transverse round section at the skin line, with visualization of the vertebrae and in a standard plane with visualization of the stomach, umbilical vein, and portal sinus. The biometric measurements of the CRL, BPD, HC, FL and / or AC are output to the age calculation formula, which estimates the gestational age of the fetus accordingly, as discussed above.

[0049] Direct gestational age (GA) estimate module 135 is accessed when the trained first machine learning model of the standard plane module 132 determines that the blind sweep ultrasound image data does not include any standard planes for making biometric measurements of the fetal anatomy. The direct GA estimate module 135 is configured to estimate the gestational age of the fetus directly from the blind sweep ultrasound image data. Estimating the gestational age of the fetus directly from the blind sweep ultrasound image data involves extracting features of the relevant fetal anatomy, and estimating the gestational age of the fetus directly from the extracted features, as discussed below. To make this estimation, the direct GA estimate module 135 provides a third machine learning model. Generally, the third machine learning model receives as input the blind sweep ultrasound image data from the blind sweeps, identifies features of fetal structures (anatomic and non-anatomic structures) in the blind sweep ultrasound image data without standard planes, and provides the estimated gestational age of the fetus directly based on the identified features. The third machine learning model is a DL machine2024PF00288- PATENT - learning model that may be implemented as any suitable type of trainable machine learning model (e.g., deep learning model), such as an RNN, a CNN, an ANN, a vision transformer, or a U-net model, for example. In an embodiment, the third machine learning model may be a feature extractor CNN with an attention module regressor (which may be referred to as a neural network regression model or a CNN regression model), for example.

[0050] The third machine learning model is previously trained in a supervised fashion using a third training data set including labelled blind sweep ultrasound image data from thousands of ultrasound images of fetal anatomies. The labelled blind sweep ultrasound image data includes features, such as shapes, textures and sizes of fetal (anatomic and non-anatomic) structures, in the blind sweep ultrasound training images, and corresponding ground truth gestational ages of the fetuses associated with the features. The model third machine learning modifies its weights in such a way that the output converges to the ground truth gestational age, e.g., using the backpropagation method of gradients. Accordingly, the third training data set includes blind sweep ultrasound image data from blind sweep ultrasound training images of fetuses, labels expressly identifying features (e.g., boundaries, textures and sizes) of identifiable structures therein, and ground truth gestational ages associated with the respective features. The labels, including the ground truth gestational ages, may be added by experts in ultrasound imaging, such as radiologist and sonographers, for example, or may be labeled automatically through techniques such as ground truth automation, as would be apparent to one skilled in the art. The ground truth gestational ages may be determined by the experts from separate sets of images and ultrasound sweeps of the same fetuses, for example. There is no need to measure or label specific fetal anatomies in the ultrasound training images for the third training data set. The third training dataset thus consists of the blind sweep ultrasound image data and the ground truth gestational ages. The training data set may be stored in the database 112, for example.

[0051] The third machine learning model is thus trained end-to-end to estimate the gestational age of the fetus directly from the blind sweep ultrasound image data. In an embodiment, the third machine learning model may be trained with mean absolute error loss (i.e., LI loss) as the loss function, for example. After being trained on the training data set, the third machine learning model is able to process new ultrasound images and predict the gestational age based on the same.2024PF00288- PATENT -

[0052] FIG. 4 is a flow diagram of a method for directly estimating gestational age in blind sweep ultrasound image data using the trained third machine learning model provided by the direct GA estimate module 135, according to a representative embodiment. That is, during inference, the trained third machine learning model receives as input the blind sweep ultrasound image data acquired by the ultrasound probe 145 during imaging of the subject 150 in block S411. In an embodiment, preprocessing may be performed on the blind sweep ultrasound image data by a preprocessing module (not shown) prior to being received by the trained third machine learning model. The preprocessing module performs image optimization operations, such as image resizing, enhancement and the like, to enhance image quality and reduce noise, as would be apparent to one skilled in the art. In block S412, the third machine learning model extracts (identifies) features from the blind sweep ultrasound image data, where the features may include texture, shape, size and other relevant characteristics of identifiable portions of anatomic and non-anatomic structures. For example, the trained third machine learning model may identify boundaries (edges) of features in the blind sweep ultrasound images, e.g., using edge detection and feature extraction. The features may be identified by convolutional kernels, where the identified features may include texture, shape and size, for example. The features are identified, without standard planes, based on similarities to the known (ground truth) fetal structures in ultrasound image data, discussed above. In block S413, the third machine learning model estimates (predicts) the gestational age of the fetus directly based on these features calculated by the convolutional kernels. Accordingly, the third machine learning model performs direct regression from the blind sweep ultrasound image data to estimated gestational age, and therefore may be referred to as an end-to-end model. In block S414, the trained third machine learning model outputs the estimated gestational age of the fetus.

[0053] In an embodiment, the gestational age of the fetus is estimated using both techniques (assuming at least one standard plane has been identified): estimated gestational age based on blind sweep ultrasound image data with standard planes provided by the biometric measurement module 133 and the calculated GA estimate module 134, and estimated gestational age based on blind sweep ultrasound image data provided by the direct GA estimate module 135. In this case, the user is able to select one or the other estimated gestational age, or combine the results for a hybrid estimated gestational age, as discussed below.2024PF00288- PATENT -

[0054] Further, when estimated gestation ages are available using both techniques, results analysis module 136 of the memory 130 provides additional features to help the user plan. In various embodiments, the results analysis module 136 may be configured to provide analysis tools, such as confidence scores, comparisons and recommendations of the estimated gestational ages, for example. The results analysis module 136 provides a first confidence score for the estimated gestational age based on blind sweep ultrasound image data with standard planes using the trained second machine learning model, and a second confidence score for the estimated gestation age based on blind sweep ultrasound image data without standard planes using the trained third machine learning model. The first and second confidence scores indicate the level of certainty that the corresponding estimated gestational ages are accurate.

[0055] The results analysis module 136 may provide comparisons by controlling the display 124 and / or the GUI 128 to display the two estimated gestational ages, along with differences and similarities between them. This enables numerical comparisons of the estimated gestation age results.

[0056] The results analysis module 136 also may provide a recommendation with regard to which estimated gestational age to use based on various factors, such as quality of the blind sweep ultrasound images and accuracy of the measurements by the second and third machine learning models, respectively, for example. The accuracy of the respective measurements may be estimated based on various considerations, such as the number of standard planes that are found. For example, if three standard planes are found, then there is a high likelihood that the estimated gestation age provided by the biometric measurement module 133 and the calculated GA estimate module 134 will be accurate, with a high confidence level. Conversely, if only one standard plane is found, the estimated gestation age provided by the biometric measurement module 133 and the calculated GA estimate module 134 is assumed to be less accurate, and therefore has a lower confidence level. Another consideration may be expected gestational age based on factors other than imaging, such as the date of the last menstrual period (LMP). Using this expected gestational age, the estimated gestation age provided by the biometric measurement module 133 and the calculated GA estimate module 134 is more accurate in lower trimester pregnancies (first trimester and early second trimester) and less accurate in higher trimester pregnancies (late second trimester and the third trimester). In the latter case, the recommendation2024PF00288- PATENT - may be to use the result from the direct GA estimate module 135. The user may then choose which estimate to use, thereby increasing the accuracy and reliability of the estimated gestational age, while providing the user with more flexibility and control over the estimation process.

[0057] In an embodiment, when at least one standard plane is found in the blind sweep ultrasound image data, each standard plane image identified by the first machine learning model may be shown on the display 124 via the GUI 128, alongside corresponding measurements (e.g., circumference / diameter) determined by the second machine learning model and the estimated gestational age determined by the age calculation formula. The user (and / or a remote expert user) may be given a choice to edit contours of the measurements (e.g., end points of lines for calculating diameters and / or points on circumference curves) via the GUI 128 and / or the user interface 122. The estimated gestational age is then automatically updated in response to the user edits to the contours by returning the edited contours to the age calculation formula.

[0058] Also, in addition to the standard plane-based estimated gestational age output by the age calculation formula, the display 124 may also show the gestational age output estimated by the third machine learning model alongside the standard plane-based estimated gestational age. The user may then be given the choice to select one of two estimated gestational ages, or an average of the two estimated gestational ages, or other such combination, via the GUI 128 and / or the user interface 122. The user may also be given the option to recompute the gestational age estimated by the third machine learning model, e.g., in case of significant disagreement between the two estimated gestational ages) using a different subset of blind sweep ultrasound images or using a different third machine learning model, for example.

[0059] FIG. 5 is a flow diagram of a method of estimating gestational age of a fetus, according to a representative embodiment. The method may be implemented at least in part using instructions stored in memory 130 and executable by the processor 120 in the system 100, for example.

[0060] Referring to FIG. 5, the method for estimating the gestational age of the includes obtaining blind sweep ultrasound image data from ultrasound images of fetal anatomy of the fetus in block S511 by performing a blind sweep of a subject using an ultrasound probe.

[0061] In block S512, it is determined whether the blind sweep ultrasound image data includes standard plane(s) data of the fetal anatomy. This determination is made by inputting the blind2024PF00288- PATENT - sweep ultrasound image data acquired by the ultrasound probe to a trained first machine learning model for identifying standard planes in the blind sweep ultrasound image data, as discussed above.

[0062] When it is determined that the trained first machine learning model identifies at least one standard plane in the blind sweep ultrasound image data (block S512: yes), the identified at least one standard plane is / are input to a trained second machine learning model in block S513, where the trained second machine learning model identifies and determines at least one standard plane measurement of the fetal anatomy in the at least one identified standard plane, as discussed above. The gestational age of the fetus is then estimated from the Hadlock formula, the Intergrowth formula, or the WHO fetal growth formula, for example.

[0063] When it is determined in block S512 that the trained first machine learning model does not identify at least one standard plane in the blind sweep ultrasound image data (block S512: no), the blind sweep ultrasound image data is input to a trained third machine learning model in block S515. The trained third machine learning model estimates the gestational age of the fetus directly from the blind sweep ultrasound image data, as discussed above.

[0064] In optional block S516, following block S514, it may be determined whether the user would like to additionally estimate the gestational age of the fetus directly from the blind sweep data using the third machine learning model, even though standard plane measurements are otherwise available. That is, block S516 offers a dual approach in which the gestational age of the fetus is estimated using two different techniques: estimated gestational age based on standard plane measurements output by the second machine learning model and the age calculation formula, and estimated gestational age based directly on blind sweep ultrasound image data output by the third machine learning model. When the additional gestational age estimate is not desired (block S516: No), the method proceeds to block S517, discussed below. When the additional gestational age estimate is desired (block S516: Yes), the method proceeds to block S515, in which the blind sweep ultrasound image data is input to the trained third machine learning model to estimate the gestational age of the fetus directly from the blind sweep ultrasound image data, as discussed above.

[0065] When the additional gestational age estimate is provided, the user may select either the estimated gestational age based on standard plane measurements output by the second machine2024PF00288- PATENT - learning model and age calculation formula or the estimated gestational age based directly on the blind sweep ultrasound image data output by the third machine learning model. Alternatively, the estimated gestational age from the third machine learning model may be combined with the estimated age determined by the age calculation formula to provide a hybrid estimated gestational age of the fetus. For example, the gestational age from the third machine learning model and the gestational age determined by the age calculation formula may be averaged or weighted averaged to determine a hybrid estimated gestational age. For example, since the age calculation formula is generally considered to be more clinically acceptable than the third machine learning model, the second machine learning model / age calculation formula result and the third machine learning model result may be weighted at a ratio of 2 to 1 or 3 to 1. Which gestational age is more accurate depends on various factors, such as whether and how many standard plane(s) is / are present, and what the stage of the pregnancy is, for example, so other ratios reflecting such determinations may also be contemplated.

[0066] The user also has the option to combine the two estimates in a way that makes sense for a specific use case. For example, the user may use the estimated gestational age based on standard plane measurements output by the second machine learning model and age calculation formula as a primary estimate and the estimated gestation age based directly on the blind sweep ultrasound image data output by the third machine learning model as a backup or secondary estimate. The user may also compare both age estimates with a third estimate based on the last menstrual period (LMP) to further build confidence in one or the other estimated gestational age. When the user is not satisfied with either of the two estimates, the method may be repeated to recompute one or both estimates using a different subset of blind sweep ultrasound images or using a different third machine learning model.

[0067] In block S517, the estimated gestational age of the fetus is displayed on a display. In an embodiment, the gestational age may be displayed along with blind sweep ultrasound images showing portions of the fetal anatomy on which the gestational age estimate is based. The estimated gestational age of the fetus may also be stored and / or output in a report, depending on a user’s preferences.

[0068] In accordance with various embodiments of the present disclosure, the methods described herein may be implemented using a hardware computer system that executes software programs2024PF00288- PATENT - stored on non-transitory storage mediums. Further, in an exemplary, non-limited embodiment, implementations can include distributed processing, component / object distributed processing, and parallel processing. Virtual computer system processing may implement one or more of the methods or functionalities as described herein, and a processor described herein may be used to support a virtual processing environment.

[0069] Although evaluating quality of an ultrasound imaging system has been described with reference to exemplary embodiments, it is understood that the words that have been used are words of description and illustration, rather than words of limitation. Changes may be made within the purview of the appended claims, as presently stated and as amended, without departing from the scope and spirit of the embodiments. Also, although evaluating quality of an ultrasound imaging system has been described with reference to particular means, materials and embodiments, it is not intended to be limited to the particulars disclosed; rather evaluating quality of an ultrasound imaging system extends to all functionally equivalent structures, methods, and uses such as are within the scope of the appended claims.

[0070] The illustrations of the embodiments described herein are intended to provide a general understanding of the structure of the various embodiments. The illustrations are not intended to serve as a complete description of all of the elements and features of the disclosure described herein. Many other embodiments may be apparent to those of skill in the art upon reviewing the disclosure. Other embodiments may be utilized and derived from the disclosure, such that structural and logical substitutions and changes may be made without departing from the scope of the disclosure. Additionally, the illustrations are merely representational and may not be drawn to scale. Certain proportions within the illustrations may be exaggerated, while other proportions may be minimized. Accordingly, the disclosure and the figures are to be regarded as illustrative rather than restrictive.

[0071] One or more embodiments of the disclosure may be referred to herein, individually and / or collectively, by the term “invention” merely for convenience and without intending to voluntarily limit the scope of this application to any particular invention or inventive concept. Moreover, although specific embodiments have been illustrated and described herein, it should be appreciated that any subsequent arrangement designed to achieve the same or similar purpose may be substituted for the specific embodiments shown. This disclosure is intended to cover any2024PF00288- PATENT - and all subsequent adaptations or variations of various embodiments. Combinations of the above embodiments, and other embodiments not specifically described herein, will be apparent to those of skill in the art upon reviewing the description.

[0072] The Abstract of the Disclosure is provided to comply with 37 C.F.R. § 1.72(b) and is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims. In addition, in the foregoing Detailed Description, various features may be grouped together or described in a single embodiment for the purpose of streamlining the disclosure. This disclosure is not to be interpreted as reflecting an intention that the claimed embodiments require more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive subject matter may be directed to less than all of the features of any of the disclosed embodiments. Thus, the following claims are incorporated into the Detailed Description, with each claim standing on its own as defining separately claimed subject matter.

[0073] The preceding description of the disclosed embodiments is provided to enable any person skilled in the art to practice the concepts described in the present disclosure. As such, the above disclosed subject matter is to be considered illustrative, and not restrictive, and the appended claims are intended to cover all such modifications, enhancements, and other embodiments which fall within the true spirit and scope of the present disclosure. Thus, to the maximum extent allowed by law, the scope of the present disclosure is to be determined by the broadest permissible interpretation of the following claims and their equivalents and shall not be restricted or limited by the foregoing detailed description.

Claims

2024PF00288- PATENT -CLAIMS:

1. A method for estimating gestational age of a fetus using ultrasound imaging, the method comprising: performing blind sweeps of a subject using an ultrasound probe to obtain blind sweep ultrasound image data from ultrasound images of fetal anatomy of the fetus; inputting the blind sweep ultrasound image data to a trained first machine learning model for determining whether the blind sweep ultrasound image data includes standard planes for performing biometric measurements of the fetal anatomy; when the trained first machine learning model identifies at least one standard plane in the blind sweep ultrasound image data, inputting the identified at least one standard plane to a trained second machine learning model that determines at least one biometric measurement of the fetal anatomy in the at least one identified standard plane, and estimating the gestational age of the fetus from the at least one biometric measurement; when the trained first machine learning model does not identify at least one standard plane in the blind sweep ultrasound image data, inputting the blind sweep ultrasound image data to a trained third machine learning model that estimates the gestational age of the fetus directly from the blind sweep ultrasound image data; and displaying the estimated gestational age of the fetus.

2. The method of claim 1, wherein the gestational age of the fetus is estimated from the at least one biometric measurement using an age calculation formula.

3. The method of claim 2, wherein the age calculation formula comprises a Hadlock formula, an Intergrowth formula, or a World Health Organization (WHO) fetal growth formula.

4. The method of claim 1, wherein the blind sweep ultrasound image data shows at least one of a head, a femur, an abdomen, or a cross-sectional sagittal view of the fetus in the at least one standard plane, and2024PF00288- PATENT - wherein the at least one biometric measurement of the fetal anatomy determined by the trained second machine learning model comprises at least one of a head circumference of the head, a femur length of the femur, an abdominal circumference of the abdomen, a biparietal diameter of the head, or a crown-rump length.

5. The method of claim 1, wherein the trained first machine learning model comprises a first neural network model trained using a first training data set comprising labeled standard planes in ultrasound image data and labeled non-standard planes in ultrasound image data.

6. The method of claim 1, wherein the trained second machine learning model comprises a second neural network model trained using a second training data set comprising labeled biometric measurements of a plurality of types of fetal anatomy shown in standard planes of the blind sweep ultrasound image data.

7. The method of claim 1, wherein the trained third machine learning model comprises a neural network regression model trained using a third training data set comprising blind sweep ultrasound image data of fetuses and known gestational ages of the fetuses corresponding to the blind sweep ultrasound image data.

8. The method of claim 1, wherein the trained third machine learning model estimates the gestational age of the fetus directly from the blind sweep ultrasound image data by: extracting information regarding a plurality of features from the blind sweep ultrasound image data using a neural network, wherein the blind sweep ultrasound image data is preprocessed to enhance image quality and reduce noise, wherein the plurality of features include identifiable portions of anatomic structures and the information includes at least texture and shape of the features in the ultrasound images; and predicting the gestational age of the fetus based on the extracted information using a neural network regression model.2024PF00288- PATENT -9. The method of claim 1, wherein when the trained first machine learning model identifies at least one standard plane in the blind sweep ultrasound image data, the method further comprises: inputting the blind sweep ultrasound image data to the trained third machine learning model to estimate another gestational age of the fetus directly from the blind sweep ultrasound image data; and selecting one of the gestational age estimated by the trained second machine learning model or the another gestational age estimated by the trained third machine learning model as the estimated gestational age of the fetus.

10. The method of claim 1, wherein when the trained first machine learning model identifies at least one standard plane in the blind sweep ultrasound image data, the method further comprises: inputting the blind sweep ultrasound image data to the trained third machine learning model to estimate another gestational age of the fetus directly from the blind sweep ultrasound image data; and combining weighted values of the gestational age estimated by the trained second machine learning model and the another gestational age estimated by the trained third machine learning model as the estimated gestational age of the fetus.

11. A system for estimating gestational age of a fetus carried by a subject using ultrasound imaging, the system comprising: an ultrasound imaging system includes an ultrasound controller and an ultrasound probe operable by an operator and configured to obtain blind sweep ultrasound images from blind sweeps of an abdomen of the subject; a display; at least one processor coupled to the ultrasound imaging system and the display; and at least one non-transitory memory coupled to the at least one processor, the at least one non-transitory memory storing instructions that, when executed by the at least one processor, cause the at least one processor to:2024PF00288- PATENT - receive blind sweep ultrasound image data from the blind sweep ultrasound images of fetal anatomy of the fetus; input the blind sweep ultrasound image data to a trained first machine learning model for determining whether the blind sweep ultrasound image data includes standard planes for performing biometric measurements of the fetal anatomy; when the trained first machine learning model identifies at least one standard plane in the blind sweep ultrasound image data, input the identified at least one standard plane to a trained second machine learning model that determines at least one biometric measurement of the fetal anatomy in the at least one identified standard plane, and estimate the gestational age of the fetus from the at least one biometric measurement determined by the trained second machine learning model; when the trained first machine learning model does not identify at least one standard plane in the blind sweep ultrasound image data, input the blind sweep ultrasound image data to a trained third machine learning model that estimates the gestational age of the fetus directly from the blind sweep ultrasound image data; and cause the estimated gestational age of the fetus to be displayed on the display.

12. The system of claim 11, wherein the gestational age of the fetus is estimated from a Hadlock formula, an Intergrowth formula, or a World Health Organization (WHO) fetal growth formula.

13. The system of claim 11, wherein the blind sweep ultrasound image data shows at least one of a head, a femur, an abdomen, or a cross-sectional sagittal view of the fetus in the at least one standard plane, and wherein the at least one biometric measurement of the fetal anatomy determined by the trained second machine learning model comprises at least one of a head circumference of the head, a femur length of the femur, an abdominal circumference of the abdomen, a biparietal diameter of the head, or a crown-rump length.292024PF00288- PATENT -14. The system of claim 11, wherein the trained first machine learning model comprises a first neural network model trained using a first training data set comprising labeled standard planes in ultrasound image data and labeled non-standard planes in ultrasound image data.

15. The system claim 11, wherein the trained second machine learning model comprises a second neural network model trained using a second training data set comprising labeled biometric measurements of a plurality of types of fetal anatomy shown in standard planes of the ultrasound image data.

16. The system of claim 11, wherein the trained third machine learning model comprises a neural network regression model trained using a third training data set comprising blind sweep ultrasound image data of fetuses and known gestational ages of the fetuses corresponding to the blind sweep ultrasound image data.

17. The system of claim 11, wherein the trained third machine learning model estimates the gestational age of the fetus directly from the blind sweep ultrasound image data by: extracting information regarding a plurality of features from the blind sweep ultrasound image data using a neural network, wherein the plurality of features include identifiable portions of anatomic structures and the information includes at least texture and shape of the features in the blind sweep ultrasound images; and predicting the gestational age of the fetus based on the extracted information using a neural network regression model.

18. The system of claim 11, wherein when the trained first machine learning model identifies at least one standard plane in the blind sweep ultrasound image data, the instructions further cause the at least one processor to: input the blind sweep ultrasound image data to the trained third machine learning model to estimate another gestational age of the fetus directly from the blind sweep ultrasound image data; and302024PF00288- PATENT - select one of the gestational age estimated by the trained second machine learning model or the another gestational age estimated by the trained third machine learning model as the estimated gestational age of the fetus.

19. The system of claim 11, wherein when the trained first machine learning model identifies at least one standard plane in the blind sweep ultrasound image data, the instructions further cause the at least one processor to: input the blind sweep ultrasound image data to the trained third machine learning model to estimate another gestational age of the fetus directly from the blind sweep ultrasound image data; and combine weighted values of the gestational age estimated by the trained second machine learning model and the another gestational age estimated by the trained third machine learning model as the estimated gestational age of the fetus.

20. A non-transitory computer readable medium storing instructions for estimating gestational age of a fetus carried by a subject using ultrasound imaging that, when executed by at least one processor, cause the at least one processor to: receive blind sweep ultrasound image data from ultrasound images obtained by performing a blind sweep acquisition of fetal anatomy of the fetus; input the blind sweep ultrasound image data to a trained first machine learning model for determining whether the blind sweep ultrasound image data includes standard planes for performing biometric measurements of the fetal anatomy; when the trained first machine learning model identifies at least one standard plane in the blind sweep ultrasound image data, input the identified at least one standard plane to a trained second machine learning model that determines at least one biometric measurement of the fetal anatomy in the at least one identified standard plane, and estimate the gestational age of the fetus from the at least one biometric measurement determined by the trained second machine learning model; when the trained first machine learning model does not identify at least one standard plane in the blind sweep ultrasound image data, input the blind sweep ultrasound image data to a312024PF00288- PATENT - trained third machine learning model that estimates the gestational age of the fetus directly from the blind sweep ultrasound image data; and cause the estimated gestational age of the fetus to be displayed on the display.32

Images