System and method for providing clinical output for a subject based on raw ultrasound data

The system processes raw ultrasound data using machine learning to bypass conventional imaging losses, enabling accurate clinical outputs and enhanced imaging by extracting biophysical attributes, addressing the limitations of conventional ultrasound systems.

WO2026239382A1PCT designated stage Publication Date: 2026-11-19SHEBA IMPACT LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/IL2026/050395
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-12
Filing Date
2026-05-12
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Conventional ultrasound imaging systems lose critical signal characteristics and information during the imaging pipeline, limiting quantitative ultrasound techniques and AI applications, which rely on post-beamformed data or reconstructed images, and are difficult to integrate into standard workflows.

Method used

A system and method that processes raw ultrasound data using machine learning models to provide clinical outputs by bypassing the information-loss constraints of conventional imaging architectures, utilizing Self-Supervised Learning (SSL) paradigms to extract biophysical attributes and generate enhanced ultrasound representations.

Benefits of technology

Enables accurate clinical outputs and enhanced ultrasound imaging by preserving critical signal information, improving SNR, spatial resolution, and recovering subtle tissue boundaries, while providing clinical findings and risk assessments directly from raw data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure IL2026050395_19112026_PF_FP_ABST
    Figure IL2026050395_19112026_PF_FP_ABST
Patent Text Reader

Abstract

The presently disclosed subject matter discloses systems and methods for ultrasound diagnostics. The system obtains raw ultrasound data associated with at least one anatomical region or compartment of the subject; provide the raw ultrasound data to a first machine learning model capable of receiving given raw ultrasound data and providing at least one vector representation of the raw ultrasound data; and, based on a vector representation of the obtained raw ultrasound data, provided by the first machine learning model, provide at least one clinical output. Additionally, the presently disclosed subject matter discloses systems and methods for ultrasound enhancement.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] SYSTEM AND METHOD FOR PROVIDING CLINICAL OUTPUT FOR A SUBJECT BASED ON RAW ULTRASOUND DATA TECHNICAL FIELD

[0002] The present disclosure generally relates to the field of ultrasound diagnostics, and more particularly, to systems and methods for providing clinical output for a subj ect based on raw ultrasound data. In addition, the present disclosure generally relates to the field of image enhancement, and more particularly, to system sand methods for ultrasound image enhancement by leveraging raw ultrasound data.

[0003] BACKGROUND

[0004] Ultrasound imaging is a widely utilized medical modality due to its non-invasive nature, real-time visualization capabilities, and relative cost-effectiveness. In a conventional ultrasound imaging architecture, acoustic pulses are transmitted into a subject, and the resulting echoes are received by elements of a transducer array. These received signals are typically processed through a rigid sequence of operations, the "imaging pipeline", which may include beamforming, filtering, demodulation, envelope detection, time-gain compensation, log compression, and scan conversion. The final output is usually a human-interpretable representation, such as a B-mode image, M-mode trace, or Doppler measurement.

[0005] While the conventional imaging pipeline is optimized for subjective clinical visualization, it inherently involves significant data reduction and loss of information. For example, processes such as envelope detection and log compression are non-linear transformations that prioritize the dynamic range required for human perception over signal integrity. Consequently, critical signal characteristics, including phase information, frequency-dependent attenuation, scatterer microstructure details, and element-specific coherence, are often modified, averaged, or eliminated before the data reaches the clinician.

[0006] Current Quantitative Ultrasound (QUS) techniques attempt to retrieve objective parameters, such as attenuation or backscatter coefficients, to assist in tissue characterization. However, many QUS implementations rely on handcrafted features derived from post-beamformed data or reconstructed images, which are already limitedby the aforementioned imaging pipeline. Furthermore, many QUS and elastography methods require specialized hardware, dedicated pulse sequences (e.g., acoustic radiation force impulses), or specific clinical protocols that are difficult to integrate into standard workflows.

[0007] Recent advancements in artificial intelligence (Al) and machine learning have been applied to ultrasound, but these systems are frequently "image-only" models. Because these models operate on scan-converted pixel data, they cannot recover the fundamental physical information lost during the initial stages of image formation. Such systems are therefore limited by the artifacts and information bottlenecks of the conventional imaging chain.

[0008] Accordingly, there exists a technical need for systems and methods that can provide clinical outputs by directly leveraging raw ultrasound data, thereby bypassing the information-loss constraints of conventional imaging architectures.

[0009] GENERAL DESCRIPTION

[0010] In accordance with a first aspect of the presently disclosed subject matter, there is provided a system for providing clinical output for a subject based on raw ultrasound data, the system comprising a processing circuitry configured to: obtain raw ultrasound data associated with at least one anatomical region or compartment of said subject; provide said raw ultrasound data to a first machine learning model capable of receiving given raw ultrasound data and providing at least one given vector representation of said raw ultrasound data; and, based on a vector representation of the obtained raw ultrasound data, provided by said first machine learning model, provide at least one clinical output for said subject.

[0011] In some cases, the at least one clinical output includes one or more of: clinical finding, clinical diagnosis, precision treatment insight, prediction regarding a condition development, or combination thereof.

[0012] In some cases, the raw ultrasound data is associated with an anatomical Region of Interest (ROI) defined on an interpretable ultrasound image depicting said anatomical region or compartment, and wherein prior to providing said raw ultrasound data to said machine learning model, said processing circuitry is further configured to: perform an inverse ROI mapping on said interpretable ultrasound image to identify a specific portion of said raw ultrasound data corresponding to said anatomical ROI.In some cases, the vector representation represents physical and / or structural attributes associated with said at least one anatomical region or compartment, and wherein said providing of at least one clinical output is based on said attributes.

[0013] In some cases, the processing circuitry is further configured to provide said raw ultrasound data to a second machine learning model capable of generating at least one multi-dimensional embedding vector representing at least one biophysical attribute of said at least one anatomical region or compartment, based on given raw ultrasound data, and wherein said providing of said at least one clinical output is based on said biophysical attribute.

[0014] In some cases, the processing circuitry is further configured to obtain clinical data associated with said subject, and wherein said providing of said at least one clinical output is further based on said clinical data.

[0015] In some cases, the processing circuitry is further configured to concatenate said at least one vector representation, said at least one multi-dimensional embedding vector, and / or said clinical data, to form a unified representation, and provide said unified vector representation to a designated machine learning model capable of receiving a given unified vector representation and providing a given clinical output for said subject.

[0016] In some cases, the raw ultrasound data includes at least one of: channel data, pre-beamformed data, beamformed data, radiofrequency (RF) data, or combinations thereof.

[0017] In some cases, the raw ultrasound data includes at least one of: channel data, pre-beamformed data or any combination thereof.

[0018] In some cases, the first machine learning model is trained utilizing a non-generative and / or generative Self-Supervised Learning (SSL) paradigm.

[0019] In some cases, the non-generative SSL paradigm forces said machine learning model to predict latent abstract space and / or feature embedding rather than reconstructing exact pixel-level and / or signal intensities.

[0020] In some cases, the at least one anatomical region or compartment comprises a uterine cervix or an endocervical canal, and wherein said objective clinical output comprises a Spontaneous Preterm Birth (sPTB) risk assessment.

[0021] In some cases, the physical and / or structural attributes include at least one of: an attenuation coefficient, one or more Nakagami parameters, a speed of sound, a backscattering coefficient, scattering microstructure statistics, tissue homogeneity, tissueanisotropy, echogenicity, acoustic impedance, density, compressibility, conductivity, elasticity, viscosity, and stiffness.

[0022] In some cases, the anatomical region or compartment includes a tissue, and wherein said clinical output comprises identifying a tissue condition selected from a group consisting of: fibrosis, sclerosis, edema, congestion, collagen concentration, fat concentration, protein concentration, cell concentration, water content, inflammation, and osmolarity.

[0023] In some cases, any machine learning model's performance is evaluated using a scoring mechanism that measures its predictive accuracy against clinical ground truth data.

[0024] In some cases, the at least one clinical output is not preserved in a scan-converted ultrasound image derived from the raw ultrasound data.

[0025] In some cases, the processing circuitry is further configured to provide an enhanced ultrasound representation of said at least one anatomical region or compartment of said subject, wherein said representation is associated with at least one enhancement embedding vector, and wherein said processing circuitry is configured to integrate said enhancement embedding vector with said vector representation to provide said at least one clinical output.

[0026] In accordance with a second aspect of the presently disclosed subject matter, there is provided a system for providing clinical output for a subject based on raw ultrasound data, the system comprising a processing circuitry configured to: obtain raw ultrasound data associated with at least one anatomical region or compartment of said subject; analyze said raw ultrasound data to acquire one or more biophysical outputs associated with said at least one anatomical region or compartment of said subject; and, based on said acquired biophysical outputs, provide at least one clinical output for said subject.

[0027] In some cases, the processing circuitry is further configured to obtain clinical data associated with said subject, and wherein said providing of said at least one clinical output is further based on said clinical data.

[0028] In accordance with a third aspect of the presently disclosed subject matter, there is provided a system for ultrasound image enhancement, the system comprising a processing circuitry configured to: obtain raw ultrasound data associated with a subject; provide said raw ultrasound data to an image enhancement machine learning model capable of receiving given raw ultrasound data and providing at least one rnulti-dimensional embedding vector representing physical, structural, and / or clinical attributes associated with said raw ultrasound data, wherein said model is trained utilizing a generative Self-Supervised Learning (SSL) paradigm; and, based on at least one multidimensional embedding vector provided by said machine learning model, provide an enhanced ultrasound representation of said raw ultrasound data, wherein said enhanced ultrasound representation comprises reconstructed channel data or reconstructed pre-beamformed data.

[0029] In some cases, the raw ultrasound data strictly consists of channel data, and wherein said enhanced ultrasound representation comprises reconstructed channel data.

[0030] In some cases, the raw ultrasound data strictly consists of channel data, and wherein said enhanced ultrasound representation comprises reconstructed pre-beamformed data.

[0031] In some cases, providing said enhanced ultrasound representation comprises improving at least one of: a signal -to-noise ratio (SNR), a spatial resolution, a temporal resolution, a speckle reduction while preserving structural edges, or a recovery of subtle tissue boundaries.

[0032] In accordance with a fourth aspect of the presently disclosed subject matter, there is provided a method for providing clinical output for a subject based on raw ultrasound data comprising: obtaining raw ultrasound data associated with at least one anatomical region or compartment of said subject; providing said raw ultrasound data to a first machine learning model capable of receiving given raw ultrasound data and providing at least one given vector representation of said raw ultrasound data; and, based on a vector representation of the obtained raw ultrasound data, provided by said first machine learning model, providing at least one clinical output for said subject.

[0033] In some cases, the at least one clinical output includes one or more of: clinical finding, clinical diagnosis, precision treatment insight, prediction regarding a condition development, or combination thereof.

[0034] In some cases, the raw ultrasound data is associated with an anatomical Region of Interest (ROI) defined on an interpretable ultrasound image depicting said anatomical region or compartment, and wherein prior to providing said raw ultrasound data to said machine learning model, said method further comprises: performing an inverse ROI mapping on said interpretable ultrasound image to identify a specific portion of said raw ultrasound data corresponding to said anatomical ROI.In some cases, the vector representation represents physical and / or structural attributes associated with said at least one anatomical region or compartment, and wherein said providing of at least one clinical output is based on said attributes.

[0035] In some cases, the method further comprises providing said raw ultrasound data to a second machine learning model capable of generating at least one multi-dimensional embedding vector representing at least one biophysical attribute of said at least one anatomical region or compartment, based on given raw ultrasound data, and wherein said providing of said at least one clinical output is based on said biophysical attribute.

[0036] In some cases, the method further comprises obtaining clinical data associated with said subject, and wherein said providing of said at least one clinical output is further based on said clinical data.

[0037] In some cases, the method further comprises concatenating said at least one vector representation, said at least one multi-dimensional embedding vector, and / or said clinical data, to form a unified representation, and providing said unified vector representation to a designated machine learning model capable of receiving a given unified vector representation and providing a given clinical output for said subject.

[0038] In some cases, the raw ultrasound data includes at least one of: channel data, pre-beamformed data, beamformed data, radiofrequency (RF) data, or combinations thereof.

[0039] In some cases, the raw ultrasound data includes at least one of: channel data, pre-beamformed data or any combination thereof.

[0040] In some cases, the first machine learning model is trained utilizing a non-generative and / or generative Self-Supervised Learning (SSL) paradigm.

[0041] In some cases, the non-generative SSL paradigm forces said machine learning model to predict latent abstract space and / or feature embedding rather than reconstructing exact pixel-level and / or signal intensities.

[0042] In some cases, the at least one anatomical region or compartment comprises a uterine cervix or an endocervical canal, and wherein said objective clinical output comprises a Spontaneous Preterm Birth (sPTB) risk assessment.

[0043] In some cases, the physical and / or structural attributes include at least one of: an attenuation coefficient, one or more Nakagami parameters, a speed of sound, a backscattering coefficient, scattering microstructure statistics, tissue homogeneity, tissue anisotropy, echogenicity, acoustic impedance, density, compressibility, conductivity, elasticity, viscosity, and stiffness.In some cases, the anatomical region or compartment includes a tissue, and wherein said clinical output comprises identifying a tissue condition selected from a group consisting of: fibrosis, sclerosis, edema, congestion, collagen concentration, fat concentration, protein concentration, cell concentration, water content, inflammation, and osmolarity.

[0044] In some cases, any machine learning model's performance is evaluated using a scoring mechanism that measures its predictive accuracy against clinical ground truth data.

[0045] In some cases, the at least one clinical output is not preserved in a scan-converted ultrasound image derived from the raw ultrasound data.

[0046] In some cases, the method further comprises providing an enhanced ultrasound representation of said at least one anatomical region or compartment of said subject, wherein said representation is associated with at least one enhancement embedding vector, and wherein said method further comprises integrating said enhancement embedding vector with said vector representation to provide said at least one clinical output.

[0047] In accordance with a fifth aspect of the presently disclosed subject matter, there is provided a method for providing clinical output for a subject based on raw ultrasound data comprising: obtaining raw ultrasound data associated with at least one anatomical region or compartment of said subject; analyzing said raw ultrasound data to acquire one or more biophysical outputs associated with said at least one anatomical region or compartment of said subject; and, based on said acquired biophysical outputs, providing at least one clinical output for said subject.

[0048] In some cases, the processing circuitry is further configured to obtain clinical data associated with said subject, and wherein said providing of said at least one clinical output is further based on said clinical data.

[0049] In accordance with a sixth aspect of the presently disclosed subject matter, there is provided a method for ultrasound image enhancement comprising: obtaining raw ultrasound data associated with a subject; providing said raw ultrasound data to an image enhancement machine learning model capable of receiving given raw ultrasound data and providing at least one multi-dimensional embedding vector representing physical, structural, and / or clinical attributes associated with said raw ultrasound data, wherein said model is trained utilizing a generative Self-Supervised Learning (SSL) paradigm; and,based on at least one multi-dimensional embedding vector provided by said machine learning model, providing an enhanced ultrasound representation of said raw ultrasound data, wherein said enhanced ultrasound representation comprises reconstructed channel data or reconstructed pre-beamformed data.

[0050] In some cases, the raw ultrasound data strictly consists of channel data, and wherein said enhanced ultrasound representation comprises reconstructed channel data.

[0051] In some cases, the raw ultrasound data strictly consists of channel data, and wherein said enhanced ultrasound representation comprises reconstructed pre-beamformed data.

[0052] In some cases, providing said enhanced ultrasound representation comprises improving at least one of: a signal -to-noise ratio (SNR), a spatial resolution, a temporal resolution, a speckle reduction while preserving structural edges, or a recovery of subtle tissue boundaries.

[0053] In accordance with a seventh aspect of the presently disclosed subject matter, there is provided a non-transitory computer readable storage medium having computer readable program code embodied therewith, the computer readable program code, executable by at least one processing circuitry of a computer to perform a method for providing clinical output for a subject based on raw ultrasound data comprising: obtaining raw ultrasound data associated with at least one anatomical region or compartment of said subject; providing said raw ultrasound data to a first machine learning model capable of receiving given raw ultrasound data and providing at least one given vector representation of said raw ultrasound data; and, based on a vector representation of the obtained raw ultrasound data, provided by said first machine learning model, providing at least one clinical output for said subject.

[0054] In accordance with a eighth aspect of the presently disclosed subject matter, there is provided a non-transitory computer readable storage medium having computer readable program code embodied therewith, the computer readable program code, executable by at least one processing circuitry of a computer to perform a method for providing clinical output for a subject based on raw ultrasound data comprising: obtaining raw ultrasound data associated with at least one anatomical region or compartment of said subject; analyzing said raw ultrasound data to acquire one or more biophysical outputs associated with said at least one anatomical region or compartment of said subject; and, based on said acquired biophysical outputs, providing at least one clinical output for said subject.In accordance with a ninth aspect of the presently disclosed subject matter, there is provided a non-transitory computer readable storage medium having computer readable program code embodied therewith, the computer readable program code, executable by at least one processing circuitry of a computer to perform a method for ultrasound image enhancement comprising: obtaining raw ultrasound data associated with a subject; providing said raw ultrasound data to an image enhancement machine learning model capable of receiving given raw ultrasound data and providing at least one multidimensional embedding vector representing physical, structural, and / or clinical attributes associated with said raw ultrasound data, wherein said model is trained utilizing a generative Self-Supervised Learning (SSL) paradigm; and, based on at least one multidimensional embedding vector provided by said machine learning model, providing an enhanced ultrasound representation of said raw ultrasound data, wherein said enhanced ultrasound representation comprises reconstructed channel data or reconstructed pre-beamformed data.

[0055] BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to understand the presently disclosed subject matter and to see how it may be carried out in practice, the subj ect matter will now be described, by way of non-limiting examples only, with reference to the accompanying drawings, in which:

[0057] Fig. l is a schematic illustration of a conventional ultrasound imaging pipeline in which a system for providing a clinical output for a subject based on raw ultrasound data, in accordance with the presently disclosed subject matter, operates;

[0058] Fig- 2 is an exemplary illustration of different ultrasound data representations corresponding to different stages of the pipeline illustrated in Fig. 1, in accordance with the presently disclosed subject matter;

[0059] Fig- 3 is a block diagram schematically illustrating one example of a system for providing clinical output for a subject based on raw ultrasound data, in accordance with the presently disclosed subject matter;

[0060] Fig- 4 is an exemplary flowchart illustrating an example of a sequence of operations carried out by a system for providing clinical output for a subject based on raw ultrasound data, in accordance with the presently disclosed subject matter;Fig- 5 is a block diagram schematically illustrating one example of a system for ultrasound image enhancement, in accordance with the presently disclosed subject matter; and,

[0061] Fig- 6 is an exemplary flowchart illustrating an example of a sequence of operations carried out by a system for ultrasound image enhancement, in accordance with the presently disclosed subject matter.

[0062] DETAILED DESCRIPTION

[0063] In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the presently disclosed subject matter. However, it will be understood by those skilled in the art that the presently disclosed subject matter may be practiced without these specific details. In other instances, well-known methods, procedures, and components have not been described in detail so as not to obscure the presently disclosed subject matter.

[0064] In the drawings and descriptions set forth, identical reference numerals indicate those components that are common to different embodiments or configurations.

[0065] Unless specifically stated otherwise, as apparent from the following discussions, it is appreciated that throughout the specification discussions utilizing terms such as “obtaining“„ “preforming“, “analyzing”, “concatenating”, “providing”, “generating”, “training”, or the like, include action and / or processes of a computer that manipulate and / or transform data into other data, the data represented as physical quantities, e.g., such as electronic quantities, and / or the data representing the physical objects. The terms “computer”, “processor”, “processing resource”, “processing circuitry”, and “controller” should be expansively construed to cover any kind of electronic device with data processing capabilities, including, by way of non-limiting example, a personal desktop / laptop computer, a server, a computing system, a communication device, a smartphone, a tablet computer, a smart television, a processor (e.g. digital signal processor (DSP), a microcontroller, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), etc.), a group of multiple physical machines sharing performance of various tasks, virtual servers co-residing on a single physical machine, any other electronic computing device, and / or any combination thereof.

[0066] The operations in accordance with the teachings herein may be performed by a computer specially constructed for the desired purposes or by a general-purpose computerspecially configured for the desired purpose by a computer program stored in a non-transitory computer readable storage medium. The term "non-transitory" is used herein to exclude transitory, propagating signals, but to otherwise include any volatile or nonvolatile computer memory technology suitable to the application.

[0067] As used herein, the phrase "for example," "such as", "for instance" and variants thereof describe non-limiting embodiments of the presently disclosed subject matter. Reference in the specification to "one case", "some cases", "other cases" or variants thereof means that a particular feature, structure or characteristic described in connection with the embodiment s) is included in at least one embodiment of the presently disclosed subject matter. Thus, the appearance of the phrase "one case", "some cases", "other cases" or variants thereof does not necessarily refer to the same embodiment s).

[0068] It is appreciated that, unless specifically stated otherwise, certain features of the presently disclosed subject matter, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the presently disclosed subject matter, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable sub-combination.

[0069] In embodiments of the presently disclosed subject matter, fewer, more and / or different stages than those shown in Figs. 4 and 6 may be executed. In embodiments of the presently disclosed subject matter one or more stages illustrated in Figs. 4 and 6 may be executed in a different order and / or one or more groups of stages may be executed simultaneously. Each module in Figs. 3 and 5 can be made up of any combination of software, hardware and / or firmware that performs the functions as defined and explained herein. The modules in Figs.3 and 5 may be centralized in one location or dispersed over more than one location. In other embodiments of the presently disclosed subject matter, the system may comprise fewer, more, and / or different modules than those shown Figs.

[0070] 3 and 5.

[0071] Any reference in the specification to a method should be applied mutatis mutandis to a system capable of executing the method and should be applied mutatis mutandis to a non-transitory computer readable medium that stores instructions that once executed by a computer result in the execution of the method.

[0072] Any reference in the specification to a system should be applied mutatis mutandis to a method that may be executed by the system and should be applied mutatis mutandisto a non-transitory computer readable medium that stores instructions that may be executed by the system.

[0073] Any reference in the specification to a non-transitory computer readable medium should be applied mutatis mutandis to a system capable of executing the instructions stored in the non-transitory computer readable medium and should be applied mutatis mutandis to method that may be executed by a computer that reads the instructions stored in the non-transitory computer readable medium.

[0074] Any reference in the specification to “ultrasound raw data” should be interpreted broadly to include ultrasound data at any stage of the ultrasound processing pipeline, up to the final image generation. This encompasses, without limitation, channel data, pre-beamformed data, beamformed data, RF data, IQ data, as well as any intermediate or partially processed forms thereof, whether alone or in combination, unless expressly stated otherwise. In some embodiments, “ultrasound raw data” may refer only to channel data and / or pre-beamformed data.

[0075] Any reference in the specification to a “clinical output” should be interpreted broadly as encompassing, without limitation, any output, indication, determination, assessment, or inference having clinical relevance. Such clinical output may include, for example, a clinical finding, a clinical diagnosis, a precision treatment insight, a prediction regarding development, progression, likelihood, severity, or risk of a condition, or any combination thereof.

[0076] Bearing this in mind, attention is drawn to Fig. 1, showing a schematic illustration of a conventional ultrasound imaging pipeline 100 in which a system for providing a clinical output for a subject based on raw ultrasound data, also interchangeably referred to herein as a “system”, in accordance with the presently disclosed subject matter, operates.

[0077] As shown in Fig. 1, pipeline 100 includes an ultrasound imaging device, denoted 102, configured to acquire ultrasound data from at least a body part of a subject. The subject may be, for example, a human or animal subject, and the body part may include any tissue, organ, anatomical structure, lesion, vessel, cavity, or other biological region capable of being imaged using ultrasound.

[0078] Ultrasound imaging device 102 may include, for example, an ultrasound machine, ultrasound scanner, ultrasound probe, transducer array, handheld ultrasound device, portable ultrasound device, or any other ultrasound acquisition apparatus configured totransmit acoustic signals toward a target region and receive acoustic echo signals returned therefrom.

[0079] The received acoustic echo signals may be initially obtained as raw ultrasound data, such as channel data, denoted 104. Channel data 104 may correspond to received acoustic signals from respective transducer elements, channels, scan lines, transmissions, receive paths, receive apertures, transmit events, or any combination thereof. In addition, channel data 104 may include radio-frequency (RF) data, in-phase / quadrature (IQ) data, pre-beamformed data, element-domain data, channel -domain data, aperture-domain data, transmission-domain data, received echo data, sampled acoustic data, digitized acoustic data, stacked transmission data, or any other raw or partially processed ultrasound data obtained upstream of conventional image formation.

[0080] In a conventional ultrasound imaging pipeline, channel data 104 may be subjected to a beamforming process, thereby generating beamformed data, denoted 106. The beamforming process may include, for example, delay-and-sum beamforming, dynamic receive focusing, transmit focusing, receive focusing, apodization, compounding, filtering, interpolation, phase correction, time-delay correction, or any other beamforming operation configured to combine signals received from multiple transducer elements or channels so as to form spatially localized ultrasound data.

[0081] Following beamforming, beamformed data 106 may be subjected to signal and / or image processing, thereby generating a scan -converted ultrasound image, denoted 108. Such processing may include, for example, envelope detection, log compression, gain adjustment, time-gain compensation, dynamic range compression, speckle reduction, filtering, interpolation, coordinate transformation, scan conversion, grayscale mapping, contrast adjustment, image formatting, or any combination thereof. Scan-converted ultrasound image 108 may be, for example, a B-mode image or an M-mode image configured for display to a clinician, physician, operator, technician, or other user.

[0082] As illustrated in Fig. 2, different ultrasound data representations may correspond to different stages of pipeline 100. More particularly, a first representation, denoted 202, may be a scan-converted B-mode image corresponding to, or derived from, scan-converted ultrasound image 108; a second representation, denoted 204, may be beamformed data, for example log-compressed envelope data, corresponding to, or derived from, beamformed data 106; and a third representation, denoted 206, may bechannel data, for example stacked transmission data, corresponding to, forming part of, or being derived from, channel data 104.

[0083] As shown in Fig. 2, representations 202, 204 and 206 may represent the same underlying anatomical region or clinical region of interest in different data domains. Row “a” shows the respective ultrasound data representations without segmentation or annotation, whereas row “b” shows corresponding segmented or annotated regions overlaid on, mapped to, or otherwise associated with the respective data representations. Such regions may correspond to tissue regions, anatomical structures, lesion regions, pathological regions, clinically relevant regions of interest, or other regions associated with a clinical finding, diagnosis, image enhancement operation, training label, groundtruth label, or reference indication.

[0084] As evident from Figs. 1 and 2, conventional ultrasound imaging systems transform raw acoustic information into a final image through multiple processing stages. While such processing may provide a visually interpretable image, it may also reduce, compress, average, transform, or otherwise alter information present in the raw ultrasound data. Accordingly, certain tissue-dependent, pathology-dependent, subject-dependent, or diagnostically relevant acoustic features may be less apparent, or may not be preserved in full, in the final scan-converted image. For instance, channel data 104 and / or beamformed data 106 may include channel -dependent, transmission-dependent, phase-related, coherence-related, attenuation-related, scattering-related, spatial, textural, or other information-bearing features that may be unavailable, less pronounced, or less reliable in scan-converted ultrasound image 108. Such information may be clinically relevant where a diagnostic indication is encoded in acoustic patterns or signal relationships distributed across multiple channels, transmissions, or beamformed samples.

[0085] To enable improved utilization of information contained in raw ultrasound data and / or intermediate ultrasound data, and to reduce dependency on information that remains visible or preserved only in conventional scan-converted ultrasound images, the system and method for providing a clinical output for a subject based on raw ultrasound data of the presently disclosed subject matter perform the operations described herein with reference to Fig. 4.

[0086] Before turning to Fig. 4, attention is now drawn to Fig. 3, depicting a block diagram schematically illustrating one example of the system for providing clinical outputfor a subject based on raw ultrasound data 300, in accordance with the presently disclosed subject matter.

[0087] In accordance with the presently disclosed subject matter, the system for providing clinical output for a subject based on raw ultrasound data 300 (also interchangeably referred to herein as “system 300”) may comprise a network interface 306. The network interface 306 (e.g., a network card, a Wi-Fi client, 3G / 4G client, or any other component), enables system 300 to communicate over a network with external systems and handles inbound and outbound communications from such systems. For example, system 300 may receive, through network interface 306, raw ultrasound data associated with at least one anatomical region or compartment of a subject.

[0088] System 300 may further comprise or be otherwise associated with a data repository 304 (e.g., a database, a storage system, a memory including Read Only Memory - ROM, Random Access Memory - RAM, or any other type of memory, etc.) configured to store data. Some examples of data that may be stored in the data repository 304 include:

[0089] • Raw ultrasound data associated with at least one anatomical region or compartment of one or more subjects;

[0090] • One or more first machine learning models capable of receiving given raw ultrasound data and outputting at least one vector representation of the raw ultrasound data;

[0091] • One or more vector representations of raw ultrasound data;

[0092] • One or more second machine learning models configured to generate multidimensional embedding vectors representing at least one biophysical attribute of at least one anatomical structure;

[0093] • One or more clinical outputs;

[0094] • One or more anatomical Region of Interests (ROIs) defined on one or more interpretable ultrasound images depicting one or more anatomical regions or compartments; etc.

[0095] Data repository 304 may be further configured to enable retrieval and / or update and / or deletion of the stored data. It is to be noted that in some cases, data repository 304 may be distributed, while the system 300 has access to the information stored thereon,e.g., via a wired or wireless network to which system 300 is able to connect (utilizing its network interface 306).

[0096] System 300 further comprises processing circuitry 302. Processing circuitry 302 may be one or more processing units (e.g., central processing units), microprocessors, microcontrollers (e.g., microcontroller units (MCUs)) or any other computing devices or modules, including multiple and / or parallel and / or distributed processing units, which are adapted to independently or cooperatively process data for controlling relevant system 300 resources and for enabling operations related to system’s 300 resources.

[0097] The processing circuitry 302 comprises a clinical output module 308, configured to perform a clinical output process, as further detailed herein, inter alia with reference to Fig. 4.

[0098] Turning to Fig.4 there is shown a flowchart illustrating one example of operations carried out by the system for providing clinical output for a subject based on raw ultrasound data 300, in accordance with the presently disclosed subject matter.

[0099] Accordingly, the system for providing clinical output for a subject based on raw ultrasound data 300 (also interchangeably referred to hereafter as “system 300”) may be configured to perform a clinical output process 400, e.g., using clinical output module 308.

[0100] For this purpose, system 300 obtains raw ultrasound data associated with at least one anatomical region or compartment of the subject (block 402). In some embodiments, the raw ultrasound data may include at least one of: channel data, pre-beamformed data, beamformed data, radiofrequency (RF) data, or any combination thereof. For example, the channel data may comprise signals received from individual transducer elements before beamforming; the beamformed data may comprise signals spatially combined to form scan lines, image frames, or other ultrasound representations; and the RF data may comprise acquired echo signals in an un-demodulated or minimally processed form, thereby preserving amplitude, phase, and / or timing information. In some embodiments, the raw ultrasound data may further include intermediate or partially processed data derived from any of the foregoing, and may be represented in analog form, digital form, or both.

[0101] In some embodiments, the raw ultrasound data may be acquired in real-time via various hardware platforms, including for example, a point-of-care ultrasound (POCUS) device, a cart-based ultrasound imaging system, a handheld or probe-based ultrasounddevice operably coupled to a computing device, a wearable or patch-based ultrasound system, or a remotely operated or robotic ultrasound platform. Alternatively, or additionally, the raw ultrasound data may be retrieved from a digital archive, such as a cloud-based storage system or a Picture Archiving and Communication System (PACS).

[0102] In some embodiments, the raw ultrasound data may be associated with an anatomical Region of Interest (RO I) defined on an interpretable ultrasound image (e.g., a B-mode image, an M-mode image, a Doppler image, an elastography image, a contrast-enhanced ultrasound image, a three-dimensional ultrasound image, or any other ultrasound image representation suitable for clinical interpretation) depicting the anatomical region or compartment. The anatomical ROI may be identified manually by a user, by an automated or semi -automated segmentation algorithm, by a machine-learning model, or by an image registration process. Furthermore, the anatomical ROI may correspond to a discrete, localized sub-region of the image, or alternatively, it may encompass substantially the entirety of the interpretable portion of the ultrasound image.

[0103] In some embodiments, the anatomical ROI may be defined in a hierarchical or multi-scale manner. In such embodiments, a primary ROI may encompass a broad anatomical structure (e.g., the maternal uterine cervix), while one or more secondary ROIs (sub-regions) are defined therein to encompass specific internal compartments or features (e.g., the endocervical canal). System 300 may be configured to perform concurrent inverse mapping for the primary and secondary ROIs, ensuring that the first machine learning model can extract and correlate biophysical signatures across different anatomical scales within the same raw acoustic signal domain.

[0104] It is to be understood that while the anatomical ROI may be represented by a reference or metadata (such as spatial coordinates, a bounding box, a contour, or a segmentation mask), the present disclosure is not limited to such indirect representations. Alternatively or additionally, obtaining the raw ultrasound data associated with the anatomical ROI may comprise receiving the actual ROI image itself, for example, a cropped, segmented, or otherwise isolated image portion of the interpretable ultrasound image that explicitly depicts the specific anatomical structure.

[0105] Once obtained, system 300 provides the raw ultrasound data to a first machine learning model capable of receiving given raw ultrasound data and providing at least one given vector representation of the raw ultrasound data (block 404). By doing so, the system effectively reduces the high-dimensional raw acoustic signals into acomputationally manageable embedding format while preserving the rich, underlying properties that are critical for accurate diagnosis.

[0106] In some embodiments, prior to inputting the raw ultrasound data to the first machine learning model, the system may be configured to utilize the ROI identification (whether provided as metadata or as an actual image) to perform an inverse ROI mapping. This mapping identifies and isolates a specific portion of the raw ultrasound data corresponding to the anatomical ROI. This inverse ROI mapping is particularly critical when utilizing raw ultrasound channel data, which is massive, as focusing on a clinically relevant ROI significantly reduces memory load and computational complexity.

[0107] Technically, the inverse ROI mapping may involve applying a coordinate transformation that reverses the scan conversion process, thereby identifying the specific acoustic scan lines and the specific range of time-of-flight samples (indices) within those lines that contributed to the identified ROI. In embodiments where the raw data comprises pre-beamformed channel data, the inverse mapping may further involve calculating the time-delays and element-specific apertures used during the beamforming process to isolate the specific channel-level signals corresponding to the physical tissue volume represented by the ROI, thereby allowing the system to produce representations that possess higher diagnostic power for the specific clinical task involved.

[0108] In some embodiments, the first machine learning model may be trained utilizing a supervised, semi -supervised, or unsupervised machine learning approach. Where an unsupervised approach is utilized, the model may be trained employing a non-generative and / or generative Self-Supervised Learning (SSL) paradigm. In supervised configurations, the model is trained directly on raw ultrasound data, such as raw channel or radiofrequency (RF) data, using clinical ground truth labels while optionally incorporating auxiliary clinical and physical information. In self-supervised embodiments, the model leverages various latent representation approaches based on the reconstruction of specific acoustic data modalities. These reconstruction tasks include, but are not limited to: reconstructing channel data to channel data, or reconstructing channel data to beamformed data or B-mode images. In such paradigms, a non-generative approach forces said machine learning model to predict latent abstract space and / or feature embedding, an abstract representation, rather than reconstructing exact pixel -level and / or signal intensities. While a generative approach tries to predict missing parts by reconstruction of pixels.This predictive approach allows the model to eliminate unnecessary details and noise, thereby extracting highly semantic, domain-specific biophysical features and structural signatures directly from unlabeled data. These learned embeddings provide a robust, data-driven characterization of the tissue's acoustic response, which can then be applied to specific downstream clinical tasks.

[0109] Based on a vector representation of the obtained raw ultrasound data, provided by the first machine learning model, system 300 provides at least one clinical output for the subject (block 406). In some embodiments, the at least one clinical output may include one or more of clinical finding (e.g., clinical measurements, etc.), clinical diagnosis, precision treatment insight, prediction regarding a condition development, or combination thereof.

[0110] In some embodiments, the vector representation may represent physical and / or structural attributes associated with the at least one anatomical region or compartment (e.g., Nakagami parameters, a speed of sound, a backscattering coefficient, scattering microstructure statistics, tissue homogeneity, tissue anisotropy, echogenicity, acoustic impedance, density, compressibility, conductivity, elasticity, viscosity, stiffness, and the like), such that system 300 may provide the at least one clinical output based on these attributes.

[0111] In some embodiments, system 300 may be further configured to provide the raw ultrasound data to a second machine learning model capable of generating at least one multi-dimensional embedding vector representing at least one biophysical attribute of the at least one anatomical region or compartment, based on given raw ultrasound data. In such embodiments, the providing step is further based on the at least one multidimensional embedding vector.

[0112] In some embodiments, system 300 may be further configured to concatenate the at least one vector representation and the at least one multi-dimensional embedding vector to form a unified multi-dimensional vector representation, and provide the unified multidimensional vector representation to a designated machine learning model capable of receiving a given unified multi-dimensional vector representation and outputting a given objective clinical output for the subject.

[0113] In some embodiments, system 300 may be further configured to obtain clinical data associated with the subject, and the providing step may be further based on the clinical data. In such embodiments, as mentioned hereinbefore, system 300 may befurther configured to concatenate the at least one vector representation and the clinical data to form a unified multi-dimensional vector representation, and provide the unified multi-dimensional vector representation to the beforementioned designated machine learning model capable of receiving a given unified multi-dimensional vector representation and outputting a given objective clinical output for the subject.

[0114] By combining explicit physical parameters, such as clinical data and / or physical and / or structural attributes which provide a grounded, interpretable baseline of acoustic properties, with implicit biophysical signatures, which capture complex, non-linear signal patterns, the unified vector provides a more comprehensive digital characterization of the tissue state than either feature set could provide in isolation.

[0115] In some embodiments, the designated machine learning model may be trained to perform a specific downstream clinical task defined by a clinician. To achieve this, the model may be trained utilizing machine learning methods tailored to the required medical objective, including regression for continuous predictions and quantitative measurements (e.g., predicting the exact timeframe until delivery or measuring cervical length), classification for binary decisions and risk assessments, or segmentation for identifying specific anatomical regions.

[0116] Furthermore, any machine learning model's performance may be evaluated using a scoring mechanism that measures its predictive accuracy against clinical ground truth data. For example, this score may continuously evaluate the model's accuracy regarding a clinical finding, clinical diagnosis, precision treatment insight, prediction regarding a condition development, or combination thereof. This score continuously evaluates the model's predictive precision (e.g., comparing the predicted days to delivery against the actual delivery date), thereby quantifying the model's ability to extract, correlate, and assess objective clinical features that are not visible or readily apparent to a clinician in standard visual imaging.

[0117] In some embodiments, said at least one clinical output is not preserved in a scan-converted ultrasound image derived from the raw ultrasound data.

[0118] It is to be noted that while concatenation is described as one embodiment for forming this unified representation, the disclosure further contemplates other fusion techniques, such as weighted summation, element-wise multiplication, or the use of an attention-based integration layer (e.g., a transformer-based fusion module) to prioritize specific features based on signal quality or anatomical context. Furthermore, the inputprovided to the machine learning models is not limited solely to acoustic data. The input may be composed of one, two, or three different types of inputs, leveraging clinical structural data and non- structural clinical data alongside the raw ultrasound data, thereby enriching the predictive power of the task involved.

[0119] In some embodiments, the anatomical region or compartment includes a tissue, and the objective clinical output comprises identifying a tissue condition of selected from a group consisting of: fibrosis, sclerosis, edema, congestion, collagen concentration, fat concentration, protein concentration, cell concentration, water content, inflammation, and osmolarity.

[0120] a subject utilizing an inherently explainable diagnostic pathway based on raw ultrasound data. In such embodiments, the processing circuitry is configured to obtain raw ultrasound data associated with at least one anatomical region or compartment of the subject. Rather than predicting a clinical output directly from the raw data in a "blackbox" manner, the processing circuitry first analyzes the raw ultrasound data to acquire one or more biophysical outputs associated with the at least one anatomical region or compartment of the subject. These biophysical outputs serve as an objective, intermediate representation of the actual tissue properties. Subsequently, based on the acquired biophysical outputs, the processing circuitry provides at least one clinical output for the subject. By establishing the biophysical outputs as a mandatory intermediate analytical step, the system preserves a high level of clinical explainability, allowing medical professionals to directly trace the derived clinical output (e.g., a diagnosis or a risk assessment) back to the specific biophysical and structural nature of the evaluated tissue. In some embodiments, the processing circuitry may be further configured to obtain clinical data associated with the subject, wherein the providing of the at least one clinical output is further based on the clinical data in addition to the acquired biophysical outputs.

[0121] In some embodiments, the physical and / or structural attributes may be acquired using one or more analytical modules or processing pipelines. These analytical modules may include, but are not limited to, Quantitative Ultrasound (QUS) modules, signal processing modules, image processing modules, machine learning models, deep learning models, or any combinations thereof. For example, the analytical modules may utilize conventional handcrafted signal processing techniques, advanced data-driven deep learning methodologies, or an integration of quantitative analysis with deep learning, applied directly to the raw, non-processed pre-beamformed (channel) ultrasound data orpost-beamformed radiofrequency (RF) data. By applying these analytical techniques, the system can estimate and extract physical properties of the biological reflectors within the tissue, deriving parameters such as, but not limited to, an attenuation coefficient, a backscatter coefficient, a speed of sound, density, scattering microstructure statistics, tissue anisotropy, tissue water content, thermal conductivity, and electric conductivity.

[0122] In some embodiments, instead of providing the raw ultrasound data to the first machine learning model, system 300 may execute one or more analytical modules (or processing pipelines), including, but not limited to, signal processing modules, image processing modules, quantitative ultrasound (QUS) modules, or combinations thereof, to assess and derive physical and / or structural attributes directly or indirectly from the raw ultrasound data.

[0123] In a first embodiment, system 300 may be configured to input at least a portion of the raw ultrasound data directly into a designated analytical module configured to output the objective clinical output, without necessarily separating the processing into distinct physical and embedding channels. In a second embodiment, system 300 may be configured to provide at least a portion of the obtained raw ultrasound data to an analytical module capable of receiving given raw ultrasound data and assessing and / or measuring one or more given structural and / or physical attributes directly from the raw ultrasound data, and subsequently derive the objective clinical output strictly based on the measured structural and physical attributes. In a third embodiment, system 300 may be configured to assess and measure one or more tissue and / or fluid clinical attributes directly from the raw ultrasound data, e.g., utilizing a designated analytical module, and subsequently derive the objective clinical output based on the tissue and fluid clinical attributes. In a fourth embodiment, system 300 may employ a cascading analytical pipeline. In such an embodiment, system 300 provides at least a portion of the obtained raw ultrasound data to an analytical module capable of receiving given raw ultrasound data and assessing and / or measuring one or more given structural and / or physical attributes directly from the raw ultrasound data; assesses one or more tissue and / or fluid clinical attributes based on inferred results of the measured structural and physical attributes; and based on the assessed tissue and fluid clinical attributes, derives an objective clinical output for the subject.

[0124] While the aforementioned embodiments describe a separation between specific machine learning models for generating vector representations, embedding vectors, andclinical outputs, it is to be understood that in some embodiments, the processing circuitry may execute an end-to-end single approach. In such an end-to-end architecture, a single unified machine learning model (e.g., a deep neural network) is configured to receive the raw ultrasound data and directly output the objective clinical finding or diagnosis, without necessarily separating the embedding representation phase from the clinical output task.

[0125] The versatility of this analytical framework enables a vast array of clinical and diagnostic applications across diverse medical specialties. In various embodiments, the anatomical structure under assessment and the corresponding diagnostic outputs are categorized as follows:

[0126] Maternal, Fetal, and Reproductive Health

[0127] In some embodiments, the anatomical structure comprises a maternal uterine cervix or an endocervical canal, and the clinical finding comprises a spontaneous preterm birth (sPTB) risk assessment (e.g., predicting a timeframe until delivery) or a probability of successful labor induction. In other embodiments, the system assesses the placenta to identify risks of fetal growth restriction (FGR), preeclampsia, or intrauterine fetal death, and to provide evaluations of fetal anemia.

[0128] In further embodiments, the system evaluates fetal organs, such as fetal lungs (for lung maturity) or the fetal liver (for anemia assessment or detection of primary biliary cirrhosis). The system may further analyze amniotic fluid to detect meconium, evaluate lung maturation, or identify peptides and proteins related to the fetal digestive system. For fertility applications, the system may correlate ovarian follicular fluid with oocyte quality or analyze endometrial signals by hormonal phase (e.g., proliferative or secretory) to assess receptivity for embryo implantation. Additionally, the system may detect endometrial lesions (e.g., endometriosis) within the ovaries, fallopian tubes, peritoneum, bowel, bladder, diaphragm, spleen, or lungs

[0129] Internal Organs and Endocrine Systems

[0130] In some embodiments, the anatomical structure comprises a heart or a portion thereof, and the clinical output comprises one or more of cardiac fibrosis quantification, diffuse myocardial fibrosis assessment, extracellular volume estimation, myocardial amyloid infiltration detection, myocardial iron overload assessment, fatty infiltration of the myocardium, myocarditis detection or grading, ischemic versus non-ischemiccardiomyopathy differentiation, subclinical myocardial injury assessment, chemotherapy-induced cardiotoxicity monitoring (e.g., associated with anthracycline or trastuzumab therapy), radiation-induced myocardial injury assessment, atrial fibrosis burden quantification (e.g., for predicting atrial fibrillation ablation success or thromboembolic risk), or right ventricular tissue characterization. In further embodiments, the anatomical structure comprises a pericardium or a cardiac valve, and the clinical output comprises one or more of: differentiation between transudative and exudative pericardial effusion, constrictive versus restrictive physiology assessment, pericardial thickening evaluation, valvular calcification grading, differentiation between rheumatic and degenerative valvular disease, or characterization of valvular vegetations associated with suspected endocarditis.

[0131] In some embodiments, the anatomical structure comprises a liver or a portion thereof, and the clinical output comprises one or more of: liver fibrosis staging (e.g., MET A VIR-equi valent scoring), steatosis grading, lobular inflammation grading, hepatocyte ballooning assessment, comprehensive non-alcoholic or metabolic-associated steatohepatitis (NAFLD, MASLD, MASH) characterization, differentiation of fibrosis etiology (e.g., viral, alcoholic, autoimmune, cholestatic, or metabolic), portal hypertension assessment, focal hepatic lesion characterization (e.g., differentiation among hepatocellular carcinoma, metastasis, hemangioma, focal nodular hyperplasia, and hepatic adenoma), hepatic iron quantification, hepatic copper deposition assessment (e.g., associated with Wilson's disease), glycogen storage disorder characterization, or post-transplant graft assessment including rejection and ischemia-reperfusion injury monitoring. In further embodiments, the anatomical structure comprises a gallbladder or a biliary tract, and the clinical output comprises one or more of: gallbladder wall inflammation grading, differentiation between acute and chronic cholecystitis, gallbladder polyp characterization (e.g., neoplastic versus non-neoplastic), biliary sludge composition assessment, cholesterolosis or adenomyomatosis characterization, or bile duct wall characterization for differentiation among cholangiocarcinoma, primary sclerosing cholangitis, and benign biliary stricture.

[0132] In some embodiments, the anatomical structure comprises a kidney or a portion thereof, and the clinical output comprises one or more of: prediction of end-stage renal failure, differentiation of acute kidney injury etiologies (e.g., pre-renal, intrinsic, or post-renal), chronic kidney disease staging via cortical fibrosis, tubulointerstitial scarring, orglomerulosclerosis quantification, renal allograft assessment (e.g., cellular rejection, antibody-mediated rejection, BK nephropathy, or recurrent disease), renal mass characterization (e.g., differentiation among clear cell, papillary, and chromophobe renal cell carcinoma, oncocytoma, and angiomyolipoma), characterization of medical renal disease patterns (e.g., diabetic nephropathy, hypertensive nephrosclerosis, or glomerulonephritis), polycystic kidney disease cyst characterization, or hydronephrosis grading with associated parenchymal viability assessment.

[0133] In some embodiments, the anatomical structure comprises a pancreas or a portion thereof, and the clinical output comprises one or more of pancreatic endocrine function assessment, pancreatic exocrine function assessment, chronic pancreatitis fibrosis staging, differentiation of pancreatic cystic lesions (e.g., among intraductal papillary mucinous neoplasm, mucinous cystic neoplasm, serous cystic neoplasm, and pseudocyst), differentiation among pancreatic ductal adenocarcinoma, autoimmune pancreatitis, and focal chronic pancreatitis, pancreatic steatosis grading, islet cell mass estimation (e.g., for diabetes risk stratification or pre-transplant assessment), or acute pancreatitis severity grading and necrosis detection.

[0134] In some embodiments, the anatomical structure comprises a spleen or a portion thereof, and the clinical output comprises one or more of splenic cyst characterization, hemangioma characterization, splenic stiffness or congestion assessment as a portal hypertension surrogate, splenic infiltration assessment in hematologic malignancies (e.g., lymphoma or leukemia), characterization of infectious splenic involvement (e.g., granulomatous disease or abscess), or differentiation among splenic infarct, lymphomatous involvement, and metastatic disease.

[0135] In some embodiments, the anatomical structure comprises an adrenal gland, and the clinical output comprises one or more of differentiation among adrenal adenoma, metastasis, pheochromocytoma, and adrenocortical carcinoma, characterization of adrenal hyperplasia patterns, or adrenal hemorrhage characterization.

[0136] In some embodiments, the anatomical structure comprises a prostate, a testicle, or an epididymis, and the clinical output comprises one or more of prostate lesion characterization (e.g., PI-RADS-equivalent assessment), differentiation between transition zone and peripheral zone tissue characteristics, prostatitis characterization (e.g., acute, chronic, or granulomatous), benign prostatic hyperplasia tissue composition assessment (e.g., stromal versus glandular), differentiation between germ cell and non-germ cell testicular tumors, testicular microlithiasis significance assessment, postvasectomy parenchymal assessment, varicocele-associated parenchymal change assessment affecting spermatogenesis, sperm quality assessment, or differentiation between epididymal inflammation and neoplasm.

[0137] In some embodiments, the anatomical structure comprises a thyroid gland or a parathyroid gland, and the clinical output comprises one or more of thyroid nodule characterization (e.g., TI-RADS -equivalent malignancy probability assessment), differentiation among thyroiditis subtypes (e.g., Hashimoto's, De Quervain's, Riedel's, or postpartum thyroiditis), differentiation between Graves' disease and toxic multinodular goiter, or parathyroid adenoma localization and characterization.

[0138] In some embodiments, the anatomical structure comprises a breast or a portion thereof, and the clinical output comprises one or more of breast lesion characterization (e.g., BI-RADS-equivalent assessment), differentiation among invasive ductal carcinoma, invasive lobular carcinoma, mucinous carcinoma, and tubular carcinoma based on scattering microstructure signatures, ductal carcinoma in situ detection, neoadjuvant chemotherapy treatment response monitoring, lactational tissue assessment, differentiation between mastitis and inflammatory breast cancer, or breast implant integrity and capsular contracture grading.

[0139] In some embodiments, the anatomical structure comprises a lymph node or a chain of lymph nodes, and the clinical output comprises one or more of differentiation among reactive, metastatic, and lymphomatous lymph node involvement, sentinel lymph node assessment in oncologic staging, or characterization of tuberculous or granulomatous lymphadenitis.

[0140] In some embodiments, the anatomical structure comprises a portion of a gastrointestinal tract, and the clinical output comprises one or more of bowel wall characterization for inflammatory bowel disease (e.g., differentiation between Crohn's disease and ulcerative colitis), inflammatory bowel disease activity grading, differentiation between fibrostenotic and inflammatory strictures, appendiceal inflammation grading, or mesenteric tissue assessment.

[0141] Systemic, Vascular, and Specialized Tissues

[0142] In some embodiments, the system provides vascular evaluation of atherosclerotic plaques. In the lungs, the system may detect pleural effusion and differentiate between transudate and exudate. Neurological applications include neonatal transcranialultrasound for brain malperfusion and fetal / adult assessments of myelination and connectivity. The system may further characterize blood fluid properties, including viscosity, aggregation, and clot formation, or evaluate muscles for myopathies.

[0143] Universal Tissue Characterization

[0144] In some embodiments, the systems and methods disclosed herein are configured to perform universal tissue characterization that is substantially independent of the specific anatomical structure under evaluation, wherein the at least one clinical output comprises one or more biophysical, structural, compositional, or pathophysiological assessments derived from the raw ultrasound data. By leveraging vector representations and / or multi-dimensional embedding vectors that encode acoustic signatures preserved in channel data, pre-beamformed data, beamformed data, and / or radiofrequency (RF) data, the system is capable of resolving tissue properties that are not visually apparent in conventional scan-converted ultrasound images and that may be common across multiple organ systems.

[0145] In some embodiments, the clinical output comprises an inflammation assessment, including without limitation differentiation between acute and chronic inflammation, grading of inflammatory severity, identification of granulomatous versus nongranulomatous inflammatory patterns, characterization of infectious versus sterile inflammation, detection of subclinical or low-grade inflammation, and monitoring of inflammatory response to therapy. In further embodiments, the clinical output comprises a fibrosis-related assessment, including without limitation fibrosis staging, differentiation between active fibrogenesis and established fibrosis, reversibility assessment, identification of fibrotic patterns (e.g., periportal, bridging, perisinusoidal, interstitial, or replacement fibrosis), and detection of early or pre-clinical fibrotic changes preceding gross structural alteration.

[0146] In some embodiments, the clinical output comprises a tissue stiffness or mechanical property assessment, including without limitation stiffness mapping, elasticity quantification, viscoelasticity characterization, anisotropy assessment, and heterogeneity mapping across a region of interest, wherein such assessment may be performed without reliance on dedicated elastography hardware, acoustic radiation force impulses, or specialized pulse sequences. In further embodiments, the clinical output comprises a tissue composition assessment, including without limitation quantification ofcollagen content, collagen cross-linking, collagen fiber orientation, elastin content, proteoglycan content, fat or lipid concentration, intracellular versus extracellular fat distribution, protein concentration, cell density or cellularity, nuclear-to-cytoplasmic ratio surrogates, calcification burden, microcalcification detection, iron deposition, copper deposition, amyloid deposition, glycogen content, melanin content, or hemosiderin content.

[0147] In some embodiments, the clinical output comprises a fluid-related or hydration-related assessment, including without limitation tissue water content quantification, intracellular versus extracellular water distribution, edema grading and characterization (e.g., interstitial, intracellular, or lymphatic edema), congestion assessment, osmolarity estimation, ion concentration surrogates, and differentiation between transudative and exudative fluid collections. In further embodiments, the clinical output comprises a perfusion-related or microvascular assessment, including without limitation tissue perfusion estimation, microvascular density assessment, capillary integrity evaluation, ischemia detection, hypoxia surrogate assessment, or characterization of neovascularization associated with neoplastic or inflammatory processes.

[0148] In some embodiments, the clinical output comprises a neoplastic tissue characterization, including without limitation differentiation between benign and malignant lesions, tumor grade estimation, tumor heterogeneity mapping, identification of tumor microenvironment characteristics, assessment of desmoplastic reaction, detection of necrotic regions within a lesion, characterization of cystic versus solid components, and monitoring of treatment response (e.g., to chemotherapy, radiotherapy, immunotherapy, or targeted therapy). In further embodiments, the clinical output comprises a degenerative or aging-related tissue assessment, including without limitation senescence-related changes, tissue atrophy characterization, sar cop enia-r elated changes, age-related collagen and elastin alterations, and degenerative microstructural changes preceding clinically detectable disease.

[0149] In some embodiments, the clinical output comprises an imaging-pathologic correlation, wherein the vector representation, multi-dimensional embedding vector, or unified multi-dimensional vector representation is mapped to one or more histopathological reference standards, including without limitation biopsy -derived histologic grading, immunohistochemical marker expression patterns, molecular subtype classification, or genomic and transcriptomic surrogate assessment, thereby enabling anon-invasive virtual biopsy or virtual histology output. In such embodiments, the system may be configured to provide a probabilistic mapping between acoustic biophysical signatures and corresponding histopathological findings, optionally accompanied by a confidence score, an uncertainty estimate, or a similarity metric to a reference cohort.

[0150] In some embodiments, the clinical output comprises a temporal or longitudinal tissue assessment, including without limitation tracking of tissue change over time, detection of disease progression or regression, monitoring of therapeutic response, prediction of future disease trajectory, and identification of early biophysical changes preceding clinically apparent disease. In such embodiments, the system may be configured to compare a current vector representation or embedding vector against one or more prior vector representations or embedding vectors associated with the same subject, against a population-level reference distribution, or against a subject-specific baseline, to derive a quantitative measure of tissue change.

[0151] It is to be noted that the aforementioned clinical outputs, pathologies, and anatomical structures serve as mere examples to illustrate the utility of the system and are not intended to limit the scope of the presently disclosed subject matter. Other diagnostic outputs, physiological measurements, or structural assessments related to any biological tissue or fluid are similarly applicable and within the spirit of this disclosure.

[0152] Further to the foregoing, and as noted above in connection with the system for providing a clinical output for a subject based on raw ultrasound data, the presently disclosed subject matter additionally relates to an ultrasound image enhancement system. In particular, while the aforementioned system is directed to extracting clinically relevant insights from raw acoustic data, the system described herein is configured to utilize the same information-rich raw acoustic signals to generate an enhanced ultrasound image. Such enhancement may be achieved by exploiting signal features that are not fully preserved in conventional beamformed or post-processed imaging pipelines, thereby enabling the generation of an image that is visually clearer and / or more amenable to downstream analytical processing compared to standard ultrasound imaging outputs.

[0153] In some embodiments, the image enhancement functionality may be implemented as a complementary or integrated module within the diagnostic system described above, such that both enhanced visualization and clinical inference are derived from a common raw data source. This approach may facilitate improved interpretability, validation, and correlation between visual features and derived clinical findings.In some embodiments, the processing circuitry may be further configured to provide an enhanced ultrasound representation of the at least one anatomical region or compartment of the subject. In such embodiments, said representation is associated with at least one enhancement embedding vector, and the processing circuitry is configured to integrate said enhancement embedding vector with said vector representation to provide said at least one clinical output. For example, during the reconstruction or enhancement process, the image enhancement machine learning model generates at least one enhancement embedding vector representing latent semantic and structural features of the tissue. The processing circuitry may be configured to integrate (e.g., via concatenation, weighted summation, or attention mechanisms) said enhancement embedding vector with the vector representation generated by the first machine learning model. By integrating these vectors, the system enriches the clinical diagnostic pathway with high-level structural semantics learned specifically during the image enhancement task, thereby providing the at least one clinical output with higher diagnostic precision.

[0154] In some embodiments, the enhanced ultrasound representation provided by system 500 may be utilized as a visual reference or a guidance tool to corroborate and / or validate the obj ective clinical findings or diagnosis outputted by system 300. For instance, the system may be configured to present the high-fidelity reconstructed image alongside the inferred clinical finding, thereby providing a clinician with a unified decision-support interface that correlates deterministic anatomical visualization with data-driven biophysical signatures. Such a synergistic framework allows the user to visually interpret the structural basis for the derived clinical risk scores or tissue characterizations, enhancing the transparency and reliability of the diagnostic process.

[0155] Bearing this in mind, attention is now drawn to Fig. 5, depicting a block diagram schematically illustrating one example of the system for ultrasound image enhancement 500, in accordance with the presently disclosed subject matter.

[0156] In accordance with the presently disclosed subject matter, the system for ultrasound image enhancement 500 (also interchangeably referred to herein as “system 500”) may comprise a network interface 506. The network interface 506 (e.g., a network card, a Wi-Fi client, 3G / 4G client, or any other component), enables system 500 to communicate over a network with external systems and handles inbound and outbound communications from such systems. For example, system 500 may receive, through network interface 506, one or more identifications of at least one anatomical Region ofInterest (ROI) associated with a portion of an initial interpretable ultrasound image depicting at least one anatomical structure of a subject.

[0157] System 500 may further comprise or be otherwise associated with a data repository 504 (e.g., a database, a storage system, a memory including Read Only Memory - ROM, Random Access Memory - RAM, or any other type of memory, etc.) configured to store data. Some examples of data that may be stored in the data repository 504 include:

[0158] • Raw ultrasound data associated with at least one subject;

[0159] • One or more machine learning models capable of receiving given raw ultrasound data and providing at least one multi-dimensional embedding vector;

[0160] • One or more multi-dimensional embedding vectors;

[0161] • One or more enhanced ultrasound representations of raw ultrasound data;

[0162] etc.

[0163] Data repository 504 may be further configured to enable retrieval and / or update and / or deletion of the stored data. It is to be noted that in some cases, data repository 504 may be distributed, while the system 500 has access to the information stored thereon, e.g., via a wired or wireless network to which system 500 is able to connect (utilizing its network interface 506).

[0164] System 500 further comprises processing circuitry 502. Processing circuitry 502 may be one or more processing units (e.g., central processing units), microprocessors, microcontrollers (e.g., microcontroller units (MCUs)) or any other computing devices or modules, including multiple and / or parallel and / or distributed processing units, which are adapted to independently or cooperatively process data for controlling relevant system 500 resources and for enabling operations related to system’s 500 resources.

[0165] The processing circuitry 502 comprises a enhancement module 508, configured to perform an enhancement process, as further detailed herein, inter alia with reference to Fig. 6.

[0166] Turning to Fig.6 there is shown a flowchart illustrating one example of operations carried out by the system for ultrasound image enhancement 500, in accordance with the presently disclosed subject matter.Accordingly, the system for ultrasound image enhancement 500 (also interchangeably referred to hereafter as “system 500”) may be configured to perform a enhancement process 600, e.g., using enhancement module 506.

[0167] For this purpose, system 500 obtains raw ultrasound data associated with a subject (block 602) In some embodiments, the raw ultrasound data may include channel data, pre-beamformed data, beamformed data, radiofrequency (RF) data, or any combination thereof.

[0168] In some embodiments, the raw ultrasound data strictly consists of channel data and the enhanced ultrasound representation, provided by system 500, is reconstructed channel data. In another embodiments, the raw ultrasound data strictly consists of channel data, and wherein the enhanced ultrasound representation comprises reconstructed pre-beamformed data.

[0169] Once obtained, system 500 provides the raw ultrasound data to an image enhancement machine learning model capable of receiving given raw ultrasound data and providing at least one multi-dimensional embedding vector representing physical, structural, and / or clinical attributes associated with the raw ultrasound data (block 604) In some embodiments, the image enhancement machine learning model may be trained utilizing a non-generative and / or generative Self-Supervised Learning (SSL) paradigm. Unlike conventional image reconstruction approaches that focus on reconstructing pixel-level intensities, a non-generative predictive approach forces the model to predict latent abstract space and / or feature embedding, an abstract representation, rather than reconstructing exact pixel-level and / or signal intensities. This allows the model to eliminate unnecessary pixel-level details and focus on high-level semantic and biophysical structures, overcoming the challenge of scarce expert-annotated medical datasets.

[0170] Based on the at least one multi-dimensional embedding vector provided by the image enhancement machine learning model, system 500 provides an enhanced ultrasound representation of the raw ultrasound data, wherein the enhanced ultrasound representation comprises reconstructed channel data or reconstructed pre-beamformed data (block 606).

[0171] By basing the image reconstruction directly on the multi-dimensional embedding vector, the enhancement process is inherently guided by the derived physical, structural, and / or clinical attributes of the tissue. This fundamentally differentiates the presentsystem from standard image processing techniques; the reconstructed acoustic data is infused with actual biophysical tissue characteristics (e.g., speed of sound, attenuation), yielding an enhanced representation of the true biological state rather than merely an aesthetic visual improvement.

[0172] Furthermore, because the reconstruction tasks involve channel-data to channeldata reconstruction, or channel-data to pre-beamformed data reconstruction, the system operates entirely upstream of conventional beamforming and scan-conversion processes. By doing so, system 500 avoids the inherent data loss and non-linear compression typical of standard imaging pipelines, producing representations that are of significantly higher diagnostic power.

[0173] In some embodiments, providing the enhanced ultrasound representation comprises improving at least one of: a signal-to-noise ratio (SNR), a spatial resolution, a temporal resolution, a speckle reduction while preserving structural edges, or a recovery of subtle tissue boundaries.

[0174] In some embodiments, the enhancement may be tailored for specific anatomical targets, where the reconstruction prioritizes the high-fidelity visualization of physical structures. By synthesizing the image based on latent biophysical embeddings derived from raw data, the system produces an output that provides a more accurate representation of the physical anatomy than is possible through conventional image-processing techniques. Moreover, rather than applying a global enhancement filter across the entire field of view, the system may perform a targeted, ROI-specific image enhancement driven by an inverse ROI mapping process.

[0175] By focusing the latent representation and subsequent signal reconstruction strictly on a specific anatomical regime (e.g., the endocervical canal or a specific lesion), the image enhancement machine learning model allocates all computational resources to resolving the specific biophysical and anatomical nuances of that distinct target.

[0176] It is to be noted, with reference to Figs. 4 and 6, that some of the blocks can be integrated into a consolidated block or can be broken down to a few blocks and / or other blocks may be added. It is to be further noted that some of the blocks are optional. It should be also noted that whilst the flow diagram is described also with reference to the system elements that realizes them, this is by no means binding, and the blocks can be performed by elements other than those described herein.It is to be understood that the presently disclosed subject matter is not limited in its application to the details set forth in the description contained herein or illustrated in the drawings. The presently disclosed subject matter is capable of other embodiments and of being practiced and carried out in various ways. Hence, it is to be understood that the phraseology and terminology employed herein are for the purpose of description and should not be regarded as limiting. As such, those skilled in the art will appreciate that the conception upon which this disclosure is based may readily be utilized as a basis for designing other structures, methods, and systems for carrying out the several purposes of the present presently disclosed subject matter.

[0177] It will also be understood that the system according to the presently disclosed subject matter can be implemented, at least partly, as a suitably programmed computer. Likewise, the presently disclosed subject matter contemplates a computer program being readable by a computer for executing the disclosed method. The presently disclosed subject matter further contemplates a machine-readable memory tangibly embodying a program of instructions executable by the machine for executing the disclosed method.

Claims

CLAIMS:

1. A system for providing clinical output for a subj ect based on raw ultrasound data, the system comprising a processing circuitry configured to:obtain raw ultrasound data associated with at least one anatomical region or compartment of said subject;provide said raw ultrasound data to a first machine learning model capable of receiving given raw ultrasound data and providing at least one given vector representation of said raw ultrasound data; and,based on a vector representation of the obtained raw ultrasound data, provided by said first machine learning model, provide at least one clinical output for said subject.

2. The system of claim 1, wherein said at least one clinical output includes one or more of: clinical finding, clinical diagnosis, precision treatment insight, prediction regarding a condition development, or combination thereof.

3. The system of claim 1, wherein said raw ultrasound data is associated with an anatomical Region of Interest (ROI) defined on an interpretable ultrasound image depicting said anatomical region or compartment, and wherein prior to providing said raw ultrasound data to said machine learning model, said processing circuitry is further configured to:perform an inverse ROI mapping on said interpretable ultrasound image to identify a specific portion of said raw ultrasound data corresponding to said anatomical ROI.

4. The system of claim 1, wherein said vector representation represents physical and / or structural attributes associated with said at least one anatomical region or compartment, and wherein said providing of at least one clinical output is based on said attributes.

5. The system of claim 1, wherein the processing circuitry is further configured to provide said raw ultrasound data to a second machine learning model capable ofgenerating at least one multi-dimensional embedding vector representing at least one biophysical attribute of said at least one anatomical region or compartment, based on given raw ultrasound data, and wherein said providing of said at least one clinical output is based on said biophysical attribute.

6. The system of claim 1, wherein the processing circuitry is further configured to obtain clinical data associated with said subject, and wherein said providing of said at least one clinical output is further based on said clinical data.

7. The system of any one of claims 5 to 6, wherein the processing circuitry is further configured to concatenate said at least one vector representation, said at least one multi-dimensional embedding vector, and / or said clinical data, to form a unified representation, and provide said unified vector representation to a designated machine learning model capable of receiving a given unified vector representation and providing a given clinical output for said subject.

8. The system of claim 1, wherein said raw ultrasound data includes at least one of:channel data, pre-beamformed data, beamformed data, radiofrequency (RF) data, or combinations thereof.

9. The system of claim 1, wherein said raw ultrasound data includes at least one of:channel data, pre-beamformed data or any combination thereof.

10. The system of claim 1, wherein said first machine learning model is trained utilizing a non-generative and / or generative Self-Supervised Learning (SSL) paradigm.

11. The system of claim 10, wherein said non-generative SSL paradigm forces said machine learning model to predict latent abstract space and / or feature embedding rather than reconstructing exact pixel-level and / or signal intensities.

12. The system of claim 1, wherein said at least one anatomical region or compartment comprises a uterine cervix or an endocervical canal, and wherein said objective clinical output comprises a Spontaneous Preterm Birth (sPTB) risk assessment.

13. The system of claim 4, wherein said physical and / or structural attributes include at least one of: an attenuation coefficient, one or more Nakagami parameters, a speed of sound, a backscattering coefficient, scattering microstructure statistics, tissue homogeneity, tissue anisotropy, echogenicity, acoustic impedance, density, compressibility, conductivity, elasticity, viscosity, and stiffness.

14. The system of claim 1, wherein said anatomical region or compartment includes a tissue, and wherein said clinical output comprises identifying a tissue condition selected from a group consisting of: fibrosis, sclerosis, edema, congestion, collagen concentration, fat concentration, protein concentration, cell concentration, water content, inflammation, and osmolarity.

15. The system of claim 5, wherein said any machine learning model's performance is evaluated using a scoring mechanism that measures its predictive accuracy against clinical ground truth data.

16. The system of claim 1, wherein said at least one clinical output is not preserved in a scan-converted ultrasound image derived from the raw ultrasound data.

17. The system of claim 1, wherein said processing circuitry is further configured to provide an enhanced ultrasound representation of said at least one anatomical region or compartment of said subject, wherein said representation is associated with at least one enhancement embedding vector, and wherein said processing circuitry is configured to integrate said enhancement embedding vector with said vector representation to provide said at least one clinical output.

18. A system for providing clinical output for a subject based on raw ultrasound data, the system comprising a processing circuitry configured to:obtain raw ultrasound data associated with at least one anatomical region or compartment of said subject;analyze said raw ultrasound data to acquire one or more biophysical outputs associated with said at least one anatomical region or compartment of said subject; and,based on said acquired biophysical outputs, provide at least one clinical output for said subject.

19. The system of claim 18, wherein the processing circuitry is further configured to obtain clinical data associated with said subject, and wherein said providing of said at least one clinical output is further based on said clinical data.

20. A system for ultrasound image enhancement, the system comprising a processing circuitry configured to:obtain raw ultrasound data associated with a subject;provide said raw ultrasound data to an image enhancement machine learning model capable of receiving given raw ultrasound data and providing at least one multi-dimensional embedding vector representing physical, structural, and / or clinical attributes associated with said raw ultrasound data, wherein said model is trained utilizing a generative Self-Supervised Learning (SSL) paradigm; and,based on at least one multi-dimensional embedding vector provided by said machine learning model, provide an enhanced ultrasound representation of said raw ultrasound data, wherein said enhanced ultrasound representation comprises reconstructed channel data or reconstructed pre-beamformed data.

21. The system of claim 20, wherein said raw ultrasound data strictly consists of channel data, and wherein said enhanced ultrasound representation comprises reconstructed channel data.

22. The system of claim 20, wherein said raw ultrasound data strictly consists of channel data, and wherein said enhanced ultrasound representation comprises reconstructed pre-beamformed data.

23. The system of claim 20, wherein providing said enhanced ultrasound representation comprises improving at least one of: a signal -to-noise ratio (SNR), a spatialresolution, a temporal resolution, a speckle reduction while preserving structural edges, or a recovery of subtle tissue boundaries.

24. A method for providing clinical output for a subject based on raw ultrasound data comprising:obtaining raw ultrasound data associated with at least one anatomical region or compartment of said subject;providing said raw ultrasound data to a first machine learning model capable of receiving given raw ultrasound data and providing at least one given vector representation of said raw ultrasound data; and,based on a vector representation of the obtained raw ultrasound data, provided by said first machine learning model, providing at least one clinical output for said subject.

25. The method of claim 24, wherein said at least one clinical output includes one or more of: clinical finding, clinical diagnosis, precision treatment insight, prediction regarding a condition development, or combination thereof.

26. The method of claim 24, wherein said raw ultrasound data is associated with an anatomical Region of Interest (ROI) defined on an interpretable ultrasound image depicting said anatomical region or compartment, and wherein prior to providing said raw ultrasound data to said machine learning model, said method further comprises:performing an inverse ROI mapping on said interpretable ultrasound image to identify a specific portion of said raw ultrasound data corresponding to said anatomical ROI.

27. The method of claim 24, wherein said vector representation represents physical and / or structural attributes associated with said at least one anatomical region or compartment, and wherein said providing of at least one clinical output is based on said attributes.

28. The method of claim 24, wherein said method further comprises providing said raw ultrasound data to a second machine learning model capable of generating at leastone multi-dimensional embedding vector representing at least one biophysical attribute of said at least one anatomical region or compartment, based on given raw ultrasound data, and wherein said providing of said at least one clinical output is based on said biophysical attribute.

29. The method of claim 24, wherein said method further comprises obtaining clinical data associated with said subject, and wherein said providing of said at least one clinical output is further based on said clinical data.

30. The method of any one of claims 28 to 29, wherein the method further comprises concatenating said at least one vector representation, said at least one multidimensional embedding vector, and / or said clinical data, to form a unified representation, and providing said unified vector representation to a designated machine learning model capable of receiving a given unified vector representation and providing a given clinical output for said subject.

31. The method of claim 24, wherein said raw ultrasound data includes at least one of:channel data, pre-beamformed data, beamformed data, radiofrequency (RF) data, or combinations thereof.

32. The method of claim 24, wherein said raw ultrasound includes at least one of: channel data, pre-beamformed data or any combination thereof.

33. The method of claim 24, wherein said first machine learning model is trained utilizing a non-generative and / or generative Self-Supervised Learning (SSL) paradigm.

34. The method of claim 33, wherein said non-generative SSL paradigm forces said machine learning model to predict latent abstract space and / or feature embedding rather than reconstructing exact pixel-level and / or signal intensities.

35. The method of claim 24, wherein said at least one anatomical region or compartment comprises a uterine cervix or an endocervical canal, and wherein said objective clinical output comprises a Spontaneous Preterm Birth (sPTB) risk assessment.

36. The method of claim 27, wherein said physical and / or structural attributes include at least one of: an attenuation coefficient, one or more Nakagami parameters, a speed of sound, a backscattering coefficient, scattering microstructure statistics, tissue homogeneity, tissue anisotropy, echogenicity, acoustic impedance, density, compressibility, conductivity, elasticity, viscosity, and stiffness.

37. The method of claim 24, wherein said anatomical region or compartment includes a tissue, and wherein said clinical output comprises identifying a tissue condition selected from a group consisting of: fibrosis, sclerosis, edema, congestion, collagen concentration, fat concentration, protein concentration, cell concentration, water content, inflammation, and osmolarity.

38. The method of claim 28, wherein said any machine learning model's performance is evaluated using a scoring mechanism that measures its predictive accuracy against clinical ground truth data.

39. The method of claim 24, wherein said at least one clinical output is not preserved in a scan-converted ultrasound image derived from the raw ultrasound data.

40. The method of claim 24, wherein said method further comprises providing an enhanced ultrasound representation of said at least one anatomical region or compartment of said subject, wherein said representation is associated with at least one enhancement embedding vector, and wherein said method further comprises integrating said enhancement embedding vector with said vector representation to provide said at least one clinical output.

41. A method for providing clinical output for a subject based on raw ultrasound data comprising:obtaining raw ultrasound data associated with at least one anatomical region or compartment of said subject;analyzing said raw ultrasound data to acquire one or more biophysical outputs associated with said at least one anatomical region or compartment of said subject; and,based on said acquired biophysical outputs, providing at least one clinical output for said subject.

42. The method of claim 41, wherein the processing circuitry is further configured to obtain clinical data associated with said subject, and wherein said providing of said at least one clinical output is further based on said clinical data.

43. A method for ultrasound image enhancement comprising:obtaining raw ultrasound data associated with a subject;providing said raw ultrasound data to an image enhancement machine learning model capable of receiving given raw ultrasound data and providing at least one multi-dimensional embedding vector representing physical, structural, and / or clinical attributes associated with said raw ultrasound data, wherein said model is trained utilizing a generative Self-Supervised Learning (SSL) paradigm; and, based on at least one multi-dimensional embedding vector provided by said machine learning model, providing an enhanced ultrasound representation of said raw ultrasound data, wherein said enhanced ultrasound representation comprises reconstructed channel data or reconstructed pre-beamformed data.

44. The method of claim 43, wherein said raw ultrasound data strictly consists of channel data, and wherein said enhanced ultrasound representation comprises reconstructed channel data.

45. The method of claim 43, wherein said raw ultrasound data strictly consists of channel data, and wherein said enhanced ultrasound representation comprises reconstructed pre-beamformed data.

46. The method of claim 43, wherein providing said enhanced ultrasound representation comprises improving at least one of: a signal -to-noise ratio (SNR), a spatial resolution, a temporal resolution, a speckle reduction while preserving structural edges, or a recovery of subtle tissue boundaries.

47. A non-transitory computer readable storage medium having computer readable program code embodied therewith, the computer readable program code, executable by at least one processing circuitry of a computer to perform a method for providing clinical output for a subject based on raw ultrasound data comprising:obtaining raw ultrasound data associated with at least one anatomical region or compartment of said subject;providing said raw ultrasound data to a first machine learning model capable of receiving given raw ultrasound data and providing at least one given vector representation of said raw ultrasound data; and,based on a vector representation of the obtained raw ultrasound data, provided by said first machine learning model, providing at least one clinical output for said subject.

48. A non-transitory computer readable storage medium having computer readable program code embodied therewith, the computer readable program code, executable by at least one processing circuitry of a computer to perform a method for providing clinical output for a subject based on raw ultrasound data comprising:obtaining raw ultrasound data associated with at least one anatomical region or compartment of said subject;analyzing said raw ultrasound data to acquire one or more biophysical outputs associated with said at least one anatomical region or compartment of said subject; and,based on said acquired biophysical outputs, providing at least one clinical output for said subject.

49. A non-transitory computer readable storage medium having computer readable program code embodied therewith, the computer readable program code, executableby at least one processing circuitry of a computer to perform a method for ultrasound image enhancement comprising:obtaining raw ultrasound data associated with a subject;providing said raw ultrasound data to an image enhancement machine learning model capable of receiving given raw ultrasound data and providing at least one multi-dimensional embedding vector representing physical, structural, and / or clinical attributes associated with said raw ultrasound data, wherein said model is trained utilizing a generative Self-Supervised Learning (SSL) paradigm; and, based on at least one multi-dimensional embedding vector provided by said machine learning model, providing an enhanced ultrasound representation of said raw ultrasound data, wherein said enhanced ultrasound representation comprises reconstructed channel data or reconstructed pre-beamformed data.