Guided system and method for ultrasound acquisition of coronary blood flow

Abstract

A method and system for guided acquisition of coronary blood flow using ultrasound imaging. Real-time ultrasound data acquired from a patient at an initial cardiac imaging view is analyzed using a first trained model to determine a patient-specific coronary artery segment for Doppler assessment. Real-time guidance instructions are provided to reposition and orient the ultrasound probe toward the segment. The ultrasound data at the segment is analyzed using a second trained model to identify a patient-specific final acquisition position optimized for pulsed-wave Doppler (PWD) signal acquisition. Placement of a Doppler sampling cursor is guided to the identified position, and a PWD signal is acquired from the positioned cursor.

Description

ISUS Ref: 53516-003 PCTGUIDED SYSTEM AND METHOD FOR ULTRASOUND ACQUISITION OF CORONARY BLOOD FLOWCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Provisional Application No. 63 / 777,690, titled AUTOMATIC DETECTION, DIAGNOSTICS, EVALUATION AND MONITORING OF CARDIOVASCULAR DISEASES, filed March 26, 2025, which is hereby incorporated by reference in its entirety.TECHNICAL FIELD

[0002] The present disclosure relates to medical imaging technologies for cardiovascular diagnostics, in general, and more particularly to ALguided ultrasound acquisition and analysis systems for non-invasive assessment of coronary blood flow using transthoracic Doppler echocardiography.BACKGROUND

[0003] Ultrasound imaging is widely used as a non-invasive modality for visualizing cardiac anatomy and assessing blood flow within the heart and surrounding vasculature. In clinical practice, echocardiographic techniques may be used to evaluate cardiac structure, valve function, and hemodynamic parameters. Doppler-based ultrasound techniques, including pulsed-wave Doppler (PWD) and color Doppler imaging, are commonly employed to measure blood flow velocity and characterize flow patterns within cardiovascular structures.

[0004] In some clinical scenarios, ultrasound imaging may be used to obtain Doppler signals associated with blood flow in coronary arteries. The acquisition of Doppler flow measurements typically involves positioning an ultrasound probe on the patient's body surface, acquiring ultrasound image data, and manually adjusting probe position, orientation, and imaging parameters in order to visualize anatomical structures and detect blood flow signals. In addition, an operator may place a Doppler sampling cursor within an ultrasound image to define a sampling location from which Doppler signals are obtained.ISUS Ref: 53516-003 PCT

[0005] The quality and reliability of Doppler measurements may depend on multiple factors, including probe positioning, probe orientation relative to the direction of blood flow, placement of the Doppler sampling location, and the acoustic characteristics of the imaging window. In practice, operators may adjust probe position and imaging parameters while observing real-time ultrasound images and Doppler displays in order to attempt to obtain suitable signals.

[0006] Ultrasound examinations may therefore involve a sequence of manual adjustments to probe placement, imaging views, and Doppler sampling positions while monitoring real-time ultrasound data. In some cases, operators may rely on their training and experience to interpret image data and to determine whether the acquired signals correspond to physiological blood flow of interest.

[0007] Various ultrasound systems may provide tools for displaying imaging data, Doppler information, and related measurements during an examination. However, the acquisition and interpretation of Doppler signals may still depend on operator interaction with the ultrasound system during the imaging process.ISUS Ref: 53516-003 PCTBRIEF SUMMARY

[0008] One exemplary embodiment of the disclosed subject matter is a computer-implemented method for guided acquisition of coronary blood flow using ultrasound imaging, the method comprising: obtaining real-time ultrasound data acquired from a patient by an ultrasound probe positioned to obtain an initial cardiac imaging view; analyzing the real-time ultrasound data using a first trained model to determine a patientspecific coronary artery segment suitable for Doppler-based flow assessment; providing real-time guidance instructions to guide repositioning and orienting of the ultrasound probe toward the patient-specific coronary artery segment; analyzing the real-time ultrasound data at the patient- specific coronary artery segment using a second trained model, to identify a patient-specific final acquisition position optimized for Pulsed-Wave Doppler (PWD) signal acquisition; guiding placement of a Doppler sampling cursor on a visualization of the real-time ultrasound data to the identified patient- specific final acquisition position, by providing instructions for cursor placement within an ultrasound imaging system; acquiring a PWD signal from the positioned Doppler sampling cursor at the patient- specific final acquisition position.

[0009] Optionally, the method further comprising receiving, from one or more sensors associated with the ultrasound probe, real-time probe state data including at least one of spatial position, orientation, motion, or contact-related parameters.

[0010] Optionally, the one or more sensors comprise at least one of: an inertial measurement unit (IMU), an electromagnetic tracking sensor, optical sensor, sound or ultrasound sensors, a pressure sensor, or a contact sensor.

[0011] Optionally, the one or more sensors integrated with or coupled to the ultrasound probe and configured to generate real-time probe state data including at least position, orientation, or contact-related measurements.

[0012] Optionally, the one or more sensors are configured to generate real-time probe state data remotely without being integrated or coupled to the ultrasound probe. Optionally, some of the sensors or at least one element associated with the one or more sensors, is integrated with or coupled to the ultrasound probe and some are positioned remotely.ISUS Ref: 53516-003 PCT

[0013] Optionally, determining the real-time guidance instructions comprises applying a reinforcement learning model configured to output probe movement actions based on a current acquisition state.

[0014] Optionally, the reinforcement learning model is trained using a reward signal based on at least one of: a clinical validity score of a Doppler signal, a technical quality score, or expert-labeled probe movement preferences.

[0015] Optionally, the real-time guidance instructions are determined based on a combination of the real-time ultrasound data and the probe state data.

[0016] Optionally, dynamically adjusting the ultrasound acquisition parameters comprises selecting or modifying an imaging preset based on probe position, anatomical context, and detected flow characteristics.

[0017] Optionally, the probe state data is used to estimate probe alignment relative to a coronary blood flow direction.

[0018] Optionally, the method further comprising dynamically adjusting one or more ultrasound acquisition parameters, including at least one of gain, transmit frequency, pulse repetition frequency (PRF), pulse duration, imaging depth, or one or more Doppler parameters. Optionally, the dynamically adjusting is performed based on the analysis and the probe state data.

[0019] Optionally, the ultrasound acquisition parameters are adapted to normalize signal quality across different ultrasound devices or probe types.

[0020] Optionally, the method further comprising verifying that the acquired PWD signal corresponds to valid coronary artery flow and meets predefined technical quality requirements. Said verifying comprises: evaluating a clinical validity score and a technical quality score of the PWD signal; and determining whether clinical validity score and the technical quality score exceed predetermined clinical validity and technical quality thresholds, wherein the predetermined clinical validity and technical quality thresholds are adapted to patient- specific parameters; when one or more thresholds are not met, initiating iterative optimization that returns to one or more previous guidance stages to update probe position and / or cursor placement.ISUS Ref: 53516-003 PCT

[0021] Optionally, said verifying further comprises applying filtering to remove clinical noise sources including artifacts caused by unrelated vascular structures, valves, or transient physiological events, and computing the clinical validity and the technical quality scores after the filtering.

[0022] Optionally, the clinical validity score is based at least in part on coronary artery flow analysis; and wherein the technical quality score is based at least in part on signal-to-noise ratio, measurement stability or other noise metrics.

[0023] Optionally, verifying that the acquired PWD signal corresponds to valid coronary artery flow comprises segmenting the acquired PWD signal into systolic and diastolic phases and comparing the segmented data to clinical coronary flow benchmarks.

[0024] Optionally, the real-time ultrasound data comprises B-mode data and color Doppler flow data. Analyzing the real-time ultrasound data using the first trained model comprises analyzing the B-mode data at the initial cardiac imaging view. Analyzing the real-time ultrasound data using the second trained model comprises analyzing an overlay of the color Doppler flow data on the B-mode data at the patient- specific coronary artery segment.

[0025] Optionally, guiding placement of the Doppler sampling cursor comprises analyzing the color Doppler flow data in real time to detect diastolic-dominant flow within a cardiac cycle and automatically repositioning the cursor within the ultrasound device to the detected region.

[0026] Optionally, analyzing the real-time ultrasound data at the patient-specific coronary artery segment using the second trained model comprises automatically identifying diastolic flow regions by detecting flow occurring predominantly during a diastolic phase of a cardiac cycle.

[0027] Optionally, the real-time guidance instructions comprise visual guidance, audio guidance, or a combination thereof, to a human operator of the ultrasound probe.

[0028] Optionally, a same guidance logic is applied to control an autonomous transducer system or a robotic positioning mechanism configured to automatically move the ultrasound probe.ISUS Ref: 53516-003 PCT

[0029] Optionally, the real-time guidance instructions are predicted along six axes of probe movement.

[0030] Optionally, the real-time guidance instructions comprise at least one of lateral translation, angular adjustment, depth adjustment, and contact pressure modification.

[0031] Optionally, the first trained model is trained using labeled data further comprising transducer spatial position and orientation data. In some embodiments, the transducer spatial position and orientation data is obtained from at least one positioning sensor physically attached to the transducer and configured to provide high-accuracy measurements, optionally with sub-millimeter accuracy, validated by expert ultrasound technicians. Additionally, or alternatively, the transducer spatial position and orientation data may be obtained from at least one remote sensing system positioned external to the transducer, including, for example, a camera-based tracking system configured to determine the spatial position and orientation of the transducer from a remote location. The labeled data may further be validated, annotated, or refined based on expert ultrasound technician input.

[0032] Optionally, the first trained model is trained to correlate B-mode ultrasound image features with anatomical alignment and to determine anatomical views based on the B-mode ultrasound images.

[0033] Optionally, the second trained model is trained based on PWD data synchronized with cardiac cycle phases based on electrocardiogram (EKG) signals.

[0034] Optionally, positioning the Doppler sampling cursor comprises automatically repositioning the cursor within an ultrasound system or providing guidance to an operator to manually reposition the cursor.

[0035] Optionally, guidance, verification, and scoring, and the first and second trained models are adapted based on patient- specific anatomical and physiological parameters.

[0036] Optionally, at least one of the first and the second trained models is a deep reinforcement learning model trained using a reward model derived from human feedback on probe moves between views that enhance signal quality.ISUS Ref: 53516-003 PCT

[0037] Optionally, the initial cardiac imaging view is a preliminary ultrasound view of the heart acquired prior to Al-guided probe repositioning toward the patient-specific coronary artery segment.

[0038] Optionally, the method comprises iteratively optimizing probe positioning and the ultrasound acquisition parameters in a closed-loop manner based on continuous evaluation of signal quality.

[0039] Another exemplary embodiment of the disclosed subject matter is a system for guided acquisition of coronary blood flow using ultrasound imaging, comprising: an ultrasound probe configured to acquire real-time ultrasound data from a patient; a Doppler sampling cursor configured to define a sampling location for PWD signal acquisition; and a processing unit comprising one or more processors and memory storing instructions that, when executed, cause the system to: analyze the real-time ultrasound data using a first trained model to determine a patient- specific coronary artery segment suitable for Doppler-based flow assessment, the real-time ultrasound data being acquired while the ultrasound probe is positioned to obtain an initial cardiac imaging view of the heart; generate real-time guidance instructions to guide repositioning and orientation of the ultrasound probe toward the patient- specific coronary artery segment; analyze the real-time ultrasound data at the patient- specific coronary artery segment using a second trained model to identify a patient- specific final acquisition position optimized for PWD signal acquisition; guide placement of the Doppler sampling cursor to the identified patient-specific final acquisition position; and acquire a PWD signal from the positioned Doppler sampling cursor at the patient- specific final acquisition position.

[0040] Optionally, the processing unit is further configured to verify that the acquired PWD signal corresponds to valid coronary artery flow and meets predefined technical quality requirements by: evaluating a clinical validity score and a technical quality score of the PWD signal; determining whether the clinical validity score and the technical quality score exceed predetermined clinical validity and technical quality thresholds adapted to patient-specific parameters; and when one or more thresholds are not met, initiating iterative optimization that updates probe position and / or Doppler sampling cursor placement.

[0041] Optionally, the processing unit is further configured to apply filtering to remove clinical noise sources including artifacts caused by unrelated vascular structures, valves,ISUS Ref: 53516-003 PCTor transient physiological events prior to computing the clinical validity score and the technical quality score.

[0042] Optionally, the processing unit is further configured to receive the probe state data from the one or more sensors; determine probe adjustment actions based on the probe state data and the real-time ultrasound data; and adapt one or more ultrasound acquisition parameters in response to the probe state data and signal quality evaluation.

[0043] Optionally, the processing unit is configured to implement a closed-loop control system that jointly optimizes probe positioning and ultrasound acquisition parameters.

[0044] Optionally, the clinical validity score is based at least in part on coronary artery flow pattern analysis, and the technical quality score is based at least in part on signal-to-noise ratio, measurement stability, or other noise metrics.

[0045] Optionally, the processing unit is configured to segment the PWD signal into systolic and diastolic phases and determine whether the signal exhibits a diastolic-dominant coronary flow pattern.

[0046] Optionally, the real-time ultrasound data comprises B-mode data and color Doppler flow data; the first trained model analyzes the B-mode data at an initial cardiac imaging view; and the second trained model analyzes an overlay of the color Doppler flow data on the B-mode data at the patient- specific coronary artery segment.

[0047] Optionally, the processing unit is configured to analyze the color Doppler flow data in real time to detect diastolic-dominant flow within a cardiac cycle and automatically reposition the Doppler sampling cursor to a detected region.

[0048] Optionally, the second trained model is configured to automatically identify diastolic flow regions by detecting flow occurring predominantly during a diastolic phase of a cardiac cycle.

[0049] Optionally, the system further comprising a user interface configured to provide the real-time guidance instructions as visual guidance, audio guidance, or a combination thereof to a human operator of the ultrasound probe.

[0050] Optionally, the system further comprising a robotic or autonomous positioning mechanism configured to automatically move the ultrasound probe in accordance with the generated guidance instructions.ISUS Ref: 53516-003 PCT

[0051] Optionally, the real-time guidance instructions are predicted along six axes of probe movement.

[0052] Optionally, the real-time guidance instructions comprise at least one of lateral translation, angular adjustment, depth adjustment, and contact pressure modification.

[0053] Optionally, the first trained model is trained using labeled data comprising transducer spatial position and orientation data provided by a positioning sensor attached to the ultrasound probe during training with sub-millimetre accuracy and validated by expert ultrasound technicians.

[0054] Optionally, the first trained model is trained to correlate B-mode ultrasound image features with anatomical alignment and to determine anatomical views based on the B-mode ultrasound images.

[0055] Optionally, the Doppler sampling cursor is repositioned automatically by the processing unit or repositioned manually by an operator in response to guidance instructions.

[0056] Optionally, guidance, verification, scoring, and operation of the first and second trained models are adapted based on patient-specific anatomical and physiological parameters.

[0057] Optionally, at least one of the first trained model and the second trained model comprises a deep reinforcement learning model trained using a reward model derived from human feedback on probe movements between views that enhance signal quality.

[0058] Optionally, the ultrasound probe comprises an integrated positioning sensor configured to provide real-time spatial position and orientation data used for guidance and signal validation.

[0059] Optionally, the processing unit is further configured to process the acquired pulsed-wave Doppler (PWD) signal to extract coronary blood flow velocity features during a resting physiological state and during a stress physiological state; estimate coronary flow values corresponding to the resting physiological state and the stress physiological state based on the extracted features using a trained model; and compute a non-invasive coronary flow reserve (CFR) value based on a ratio between the estimatedISUS Ref: 53516-003 PCTcoronary flow value during the stress physiological state and the estimated coronary flow value during the resting physiological state.

[0060] Yet another exemplary embodiment of the disclosed subject matter is a non-transitory computer-readable medium storing instructions that, when executed by one or more processors of an ultrasound system, cause the ultrasound system to perform a method for guided acquisition of coronary blood flow using ultrasound imaging, the method comprising: obtaining real-time ultrasound data acquired from a patient by an ultrasound probe positioned to obtain an initial cardiac imaging view; analyzing the realtime ultrasound data using a first trained model to determine a patient- specific coronary artery segment suitable for Doppler-based flow assessment; providing real-time guidance instructions to guide repositioning and orienting of the ultrasound probe toward the patient-specific coronary artery segment; analyzing the real-time ultrasound data at the patient-specific coronary artery segment using a second trained model, to identify a patient-specific final acquisition position optimized for Pulsed-Wave Doppler (PWD) signal acquisition; guiding placement of a Doppler sampling cursor on a visualization of the real-time ultrasound data to the identified patient- specific final acquisition position, by providing instructions for cursor placement within an ultrasound imaging system; acquiring a PWD signal from the positioned Doppler sampling cursor at the patientspecific final acquisition position.

[0061] Yet another exemplary embodiment of the disclosed subject matter is a computer program product comprising a non-transitory computer readable storage medium retaining program instructions, which program instructions when read by a processor, cause the processor to perform a method for guided acquisition of coronary blood flow using ultrasound imaging, the method comprising: obtaining real-time ultrasound data acquired from a patient by an ultrasound probe positioned to obtain an initial cardiac imaging view; analyzing the real-time ultrasound data using a first trained model to determine a patient- specific coronary artery segment suitable for Doppler-based flow assessment; providing real-time guidance instructions to guide repositioning and orienting of the ultrasound probe toward the patient-specific coronary artery segment; analyzing the real-time ultrasound data at the patient- specific coronary artery segment using a second trained model, to identify a patient- specific final acquisition position optimized for Pulsed-Wave Doppler (PWD) signal acquisition; guiding placement of aISUS Ref: 53516-003 PCTDoppler sampling cursor on a visualization of the real-time ultrasound data to the identified patient-specific final acquisition position, by providing instructions for cursor placement within an ultrasound imaging system; acquiring a PWD signal from the positioned Doppler sampling cursor at the patient- specific final acquisition position.ISUS Ref: 53516-003 PCTTHE BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0062] The present disclosed subject matter will be understood and appreciated more fully from the following detailed description taken in conjunction with the drawings in which corresponding or like numerals or characters indicate corresponding or like components. Unless indicated otherwise, the drawings provide exemplary embodiments or aspects of the disclosure and do not limit the scope of the disclosure. In the drawings:

[0063] Figure 1 shows an exemplary environment in which the disclosed subject matter may be implemented, in accordance with some exemplary embodiments of the subject matter;

[0064] Figure 2 shows a system architecture for guided ultrasound data acquisition, in accordance with some exemplary embodiments of the disclosed subject matter;

[0065] Figure 3 shows a flowchart diagram of a method for guided ultrasound data acquisition, in accordance with some exemplary embodiments of the disclosed subject matter;

[0066] Figures 4A-4C show schematic illustrations of probe positioning guidance and probe movement directions during ultrasound acquisition, in accordance with some exemplary embodiments of the disclosed subject matter;

[0067] Figures 5A-5B show a realistic and a respective schematic ultrasound images illustrating identification of a flow region and placement of a Doppler sampling cursor, and a respective PWD signal, in accordance with some exemplary embodiments of the disclosed subject matter;

[0068] Figure 6 shows a flowchart of a feedback-based acquisition control process for iteratively evaluating and optimizing a pulsed-wave Doppler (PWD) signal, in accordance with some exemplary embodiments of the disclosed subject matter;

[0069] Figures 7A-7B show a multimodal training data representation, in accordance with some exemplary embodiments of the disclosed subject matter; and

[0070] Figure 8 shows a schematic diagram of a training architecture for training one or more machine-learning models configured to assist in ultrasound probe positioning and coronary Doppler signal acquisition, in accordance with some exemplary embodiments.ISUS Ref: 53516-003 PCTDETAILED DESCRIPTION

[0071] One technical problem addressed by the disclosed subject matter relates to the difficulty of reliably acquiring Doppler ultrasound signals corresponding to coronary artery blood flow using transthoracic ultrasound imaging. Coronary arteries are relatively small anatomical structures, and the spatial orientation of coronary arteries relative to an ultrasound probe may vary between patients and during different phases of the cardiac cycle. As a result, determining an appropriate probe position, probe orientation, and Doppler sampling location for acquiring a pulsed-wave Doppler signal corresponding to coronary blood flow may be challenging. In addition, even when a Doppler signal is obtained, determining whether the signal corresponds to coronary artery flow and whether the signal satisfies technical quality conditions required for reliable physiological analysis may be difficult.

[0072] Another technical problem addressed by the disclosed subject matter relates to the lack of automated mechanisms capable of guiding probe positioning, probe orientation, and Doppler sampling location selection during coronary Doppler imaging based on real-time ultrasound data. In many ultrasound workflows, probe positioning and sampling location selection may rely on manual adjustments while observing ultrasound images, which may lead to variability in acquisition results and inconsistent Doppler signal quality. Operators may adjust probe position and imaging parameters while observing real-time ultrasound images and Doppler displays in order to attempt to obtain suitable signals. However, the acquisition and interpretation of Doppler signals may depend heavily on operator training and experience to interpret image data and to determine whether the acquired signals correspond to physiological blood flow of interest. The absence of automated guidance mechanisms may therefore result in acquisition workflows that are highly operator-dependent and may produce variable signal quality across different examinations.

[0073] One technical solution is a computer- implemented ultrasound acquisition system configured to perform closed-loop control of coronary Doppler signal acquisition based on analysis of real-time ultrasound data. The system may process real-time ultrasound imaging data using one or more trained models in order to determine a patientspecific coronary artery segment suitable for Doppler-based flow assessment, generate guidance instructions for repositioning and orienting an ultrasound probe toward theISUS Ref: 53516-003 PCTdetermined coronary artery segment, determine a patient- specific acquisition position optimized for pulsed-wave Doppler signal acquisition, guide placement of a Doppler sampling cursor to the determined acquisition position, and evaluate the acquired Doppler signal in order to determine whether the signal satisfies predefined clinical validity and technical quality conditions.

[0074] In some exemplary embodiments, the closed-loop acquisition process may be implemented using multiple trained models that operate on different stages of the acquisition workflow. For example, a first trained model may analyze ultrasound imaging data acquired at an initial cardiac imaging view in order to determine a patient-specific coronary artery segment suitable for Doppler-based flow assessment and to generate probe positioning guidance. A second trained model may analyze ultrasound data obtained at the identified coronary artery segment in order to determine an acquisition position optimized for pulsed-wave Doppler signal acquisition.

[0075] In some exemplary embodiments, real-time ultrasound imaging data may be obtained while an ultrasound probe is positioned to obtain an initial cardiac imaging view. The first trained model may analyze the ultrasound imaging data in order to determine a patient-specific coronary artery segment suitable for Doppler-based flow assessment and to generate real-time guidance instructions for repositioning and orienting the ultrasound probe toward the identified coronary artery segment. The real-time guidance instructions may comprise visual guidance, audio guidance, or a combination thereof provided to a human operator of the ultrasound probe. In some embodiments, the guidance instructions may specify probe movements along multiple axes including lateral translation, angular adjustment, depth adjustment, and contact pressure modification.

[0076] Additionally, or alternatively, ultrasound data obtained at the patient- specific coronary artery segment may be analyzed using the second trained model to determine a patient-specific acquisition position optimized for pulsed-wave Doppler signal acquisition. In some embodiments, color Doppler flow information overlaid on B-mode ultrasound images may be analyzed in order to identify flow regions corresponding to coronary blood flow and to guide placement of a Doppler sampling cursor within an ultrasound imaging system to the determined acquisition position. In some embodiments, the Doppler sampling cursor may be automatically positioned within the ultrasound imaging system or repositioned based on guidance instructions provided to an operator.ISUS Ref: 53516-003 PCT

[0077] Additionally, or alternatively, the system may evaluate the acquired pulsed-wave Doppler signal to determine whether the signal corresponds to coronary artery flow and satisfies predefined technical quality conditions. In some exemplary embodiments, the evaluation may include determining a clinical validity score and a technical quality score associated with the Doppler signal. The clinical validity score may be based at least in part on coronary artery flow pattern analysis, while the technical quality score may be based at least in part on signal-to-noise ratio, measurement stability, or other noise-related metrics. The evaluation may further include analyzing the Doppler signal according to phases of a cardiac cycle and determining whether the signal exhibits flow characteristics associated with coronary artery flow, such as diastolic-dominant flow patterns. In some embodiments, filtering may be applied to remove clinical noise sources including artifacts caused by unrelated vascular structures, cardiac valves, or transient physiological events prior to computing the clinical validity score and the technical quality score. The system may further segment the acquired pulsed-wave Doppler signal into systolic and diastolic phases and compare the segmented data to clinical coronary flow benchmarks in order to verify that the acquired signal corresponds to valid coronary artery flow.

[0078] In some embodiments, when one or more of the clinical validity score or the technical quality score does not satisfy predetermined thresholds, the system may initiate an iterative acquisition optimization process that returns to one or more earlier acquisition stages and updates probe position, probe orientation, and / or Doppler sampling cursor placement based on analysis of the acquired Doppler signal. The predetermined thresholds may be adapted to patient-specific anatomical and physiological parameters. In some embodiments, the real-time guidance instructions may be applied to guide manual probe adjustments performed by an operator. Additionally, or alternatively, the same guidance logic may be applied to control an autonomous transducer system or a robotic positioning mechanism configured to automatically move the ultrasound probe. In some exemplary embodiments, at least one of the first trained model and the second trained model may comprise a deep reinforcement learning model trained using a reward model derived from human feedback on probe movements between views that enhance signal quality.ISUS Ref: 53516-003 PCT

[0079] One technical effect of utilizing the disclosed subject matter is improving the reliability of acquiring Doppler signals corresponding to coronary artery blood flow using transthoracic ultrasound imaging by performing closed-loop acquisition control based on analysis of real-time ultrasound imaging data and Doppler signals. The closed-loop acquisition control may enable the system to continuously monitor the quality and validity of acquired signals and to adjust probe positioning and Doppler cursor placement in response to signal characteristics. Such adaptive control may increase the likelihood of obtaining Doppler signals that correspond to coronary artery flow rather than signals originating from unrelated vascular structures or noise sources.

[0080] Another technical effect of utilizing the disclosed subject matter is reducing dependence on operator experience during coronary Doppler acquisition by generating real-time probe positioning guidance and automatically determining Doppler sampling locations suitable for coronary flow measurements. The real-time guidance instructions may enable operators with varying levels of experience to achieve consistent acquisition results by providing step-by-step instructions for probe repositioning and orientation toward patient- specific coronary artery segments. The automatic determination of Doppler sampling locations may further reduce variability in cursor placement that may otherwise result from manual selection by different operators.

[0081] Yet another technical effect of utilizing the disclosed subject matter is improving the technical quality and physiological validity of Doppler signals used for coronary flow analysis through automated signal evaluation and iterative acquisition optimization. In some embodiments, automated evaluation of clinical validity scores and technical quality scores may enable the system to identify signals that do not satisfy predefined quality thresholds. In response to such determinations, the system may initiate corrective actions such as updating probe position, modifying probe orientation, or adjusting Doppler cursor placement. The iterative optimization process may continue until the acquired signal satisfies the predefined thresholds, thereby improving the overall quality of signals used for subsequent analysis.

[0082] Yet another technical effect of utilizing the disclosed subject matter is enabling a more consistent and repeatable coronary Doppler acquisition workflow by automatically determining acquisition positions and validating acquired Doppler signals based on analysis of real-time ultrasound data. The use of trained models to determineISUS Ref: 53516-003 PCTpatient-specific coronary artery segments and corresponding acquisition positions may reduce inter-operator variability and may enable standardized acquisition protocols across different clinical environments. Validation of acquired signals against predefined clinical validity and technical quality thresholds may further provide a consistent quality control mechanism that may be applied regardless of operator experience level or patientspecific anatomical variations.

[0083] Referring now to Figure 1, showing an exemplary environment in which the disclosed subject matter may be implemented, in accordance with some exemplary embodiments of the subject matter.

[0084] In some exemplary embodiments, an Ultrasound Acquisition System 100 may be configured to acquire ultrasound data from a Patient 120 and to provide Al-guided optimization technology for acquiring Doppler signals in real time. Ultrasound Acquisition System 100 may be deployed in various clinical settings including emergency departments, inpatient units, outpatient cardiology clinics, imaging centers, and private diagnostic clinics. Ultrasound Acquisition System 100 may comprise an Ultrasound Imaging Device 115, an ultrasound Probe 102, a Display Device 110, and one or more processing components configured to analyze acquired ultrasound data and generate guidance information during the acquisition process.

[0085] In some exemplary embodiments, Probe 102 may be configured to acquire realtime ultrasound data from Patient 120. Probe 102 may be positioned relative to Patient 120 in order to obtain ultrasound imaging data corresponding to a region of interest of the patient, such as the heart. Probe 102 may be placed on an external body surface of Patient 120, such as on a chest region of the patient during a cardiac ultrasound examination.

[0086] In some embodiments, Probe 102 may further comprise one or more sensors configured to provide real-time probe state data. The one or more sensors may include, for example, an inertial measurement unit (IMU), an electromagnetic tracking sensor, a pressure sensor, a contact sensor, or any combination thereof. The probe state data may include spatial position, orientation, motion trajectory, and / or contact-related parameters associated with Probe 102 during acquisition. In some embodiments, such probe state data may be used in combination with ultrasound data to guide probe positioning, assessISUS Ref: 53516-003 PCTalignment relative to coronary blood flow, and support closed-loop acquisition optimization.

[0087] Probe 102 may be manipulated or positioned by an Operator 125, such as a cardiologist, sonographer, emergency physician, technician, or other healthcare professional performing an ultrasound examination. Operator 125 may adjust probe position, orientation, angle, or contact pressure relative to Patient 120 to obtain ultrasound imaging data.

[0088] In some embodiments, Ultrasound Imaging Device 115 may comprise a processing unit including one or more processors and memory storing instructions configured to analyze ultrasound data and generate guidance information. Ultrasound Imaging Device 115 may be connected to or integrated with Display Device 110, which may be configured to present ultrasound images and related information.

[0089] Display Device 110 may present ultrasound imaging data obtained from Patient 120, such as B-mode ultrasound images, Doppler flow information, color Doppler flow data, or other ultrasound-derived signals. In some exemplary embodiments, Display Device 110 may further present visual guidance information, such as graphical indicators or cursor placement suggestions, to assist Operator 125 during the ultrasound acquisition process.

[0090] In some embodiments, the real-time ultrasound data acquired by Probe 102 may include B-mode data and color Doppler flow data. In some exemplary embodiments, Probe 102 may comprise a high-frequency phased-array transducer designed with a relatively small footprint enabling improved maneuverability during cardiac imaging and enhanced Doppler sensitivity for detecting coronary blood flow.

[0091] In some exemplary embodiments, Probe 102 may initially be positioned to obtain an initial cardiac imaging view of the heart. The initial cardiac imaging view may correspond to a commonly used probe position used in cardiac ultrasound examinations for acquiring ultrasound data of the heart. For example, the initial cardiac imaging view may correspond to a probe position associated with one or more conventional echocardiographic imaging windows or orientations.

[0092] The initial cardiac imaging view may serve as a starting point for Al-guided probe navigation, in which the system analyzes the acquired ultrasound data and providesISUS Ref: 53516-003 PCTguidance instructions to progressively reposition and orient Probe 102 toward a target vessel segment suitable for Doppler signal acquisition.

[0093] In some exemplary embodiments, Ultrasound Imaging Device 115 may analyze the real-time ultrasound data using a first trained model to determine a patient- specific coronary artery segment suitable for Doppler-based flow assessment. The first trained model may analyze B-mode ultrasound data acquired at the initial cardiac imaging view.

[0094] Based on this analysis, Ultrasound Imaging Device 115 may generate real-time guidance instructions to guide repositioning and orientation of Probe 102 toward the patient-specific coronary artery segment. The guidance instructions may include predicted probe movements along multiple axes including lateral translation, angular adjustment, depth adjustment, and contact pressure modification. In some embodiments, the system may provide navigation instructions corresponding to ultrasound transducer gestures including rotation, tilt, sliding, and rocking movements.

[0095] The system may further provide additional acquisition guidance including breathing instructions to Patient 120 during signal acquisition or prompts to adjust ultrasound acquisition parameters or presets based on anatomical position, vessel orientation, or imaging depth.

[0096] In some exemplary embodiments, Ultrasound Imaging Device 115 may analyze the ultrasound data at the identified coronary artery segment using a second trained model to determine a patient- specific final acquisition position optimized for pulsed-wave Doppler signal acquisition.

[0097] In some embodiments, the system may guide placement of a Doppler Sampling Cursor 104 within the ultrasound image corresponding to the identified acquisition location. The system may provide instructions for manual cursor placement or may automatically reposition the cursor to a detected region corresponding to diastolic-dominant blood flow.

[0098] Ultrasound Acquisition System 100 may acquire a pulsed-wave Doppler (PWD) signal from the positioned Doppler Sampling Cursor 104 and may verify that the acquired signal corresponds to valid coronary artery flow. Verification may include evaluating a clinical validity score and a technical quality score, segmenting the signal into systolic and diastolic phases, and determining whether predefined quality thresholds are satisfied.ISUS Ref: 53516-003 PCT

[0099] When the acquired signal does not satisfy predefined thresholds, the system may perform iterative optimization by updating probe position and / or Doppler cursor placement and repeating the acquisition process. The system may apply filtering to remove noise sources including artifacts caused by unrelated vascular structures, valves, or transient physiological events prior to evaluating signal quality.

[0100] It may be noted that although Figure 1 illustrates probe manipulation performed by a human Operator 125, the disclosed subject matter is not limited to this configuration. In some embodiments, the same guidance logic used to generate instructions for Operator 125 may also be applied to control an autonomous transducer system or robotic positioning mechanism configured to automatically move Probe 102. It may further be noted that although the exemplary embodiments described herein relate primarily to ultrasound imaging modalities, the disclosed subject matter may also be applied to other imaging modalities capable of acquiring real-time hemodynamic information.

[0101] Referring now to Figure 2 showing a system architecture for guided ultrasound data acquisition, in accordance with some exemplary embodiments of the disclosed subject matter.

[0102] In some exemplary embodiments, a System 200 may be configured to perform guided acquisition of blood flow signals using ultrasound imaging. System 200 may correspond to, or form part of, Ultrasound Acquisition System 100 of Figure 1. System 200 may comprise an Ultrasound Imaging System 210, a User Interface / Display 220, and a Processing Unit 230 configured to analyze ultrasound data and generate guidance instructions during ultrasound acquisition. In some embodiments, Ultrasound Imaging System 210 may include a Probe 212 configured to acquire real-time ultrasound data from a patient. The real-time ultrasound data may be acquired while Probe 212 is positioned to obtain an initial cardiac imaging view of the heart. In some embodiments, the real-time ultrasound data may comprise B-mode ultrasound data and color Doppler flow data.

[0103] In some embodiments, Probe 212 may be associated with one or more sensors configured to generate real-time probe state data including position, orientation, motion, and / or contact-related measurements. The probe state data may be provided to Processing Unit 230 and used in combination with the real-time ultrasound data to determine probe adjustments, evaluate acquisition quality, and adapt ultrasound acquisition parameters.ISUS Ref: 53516-003 PCT

[0104] Ultrasound Imaging System 210 may further include a Robotic Mechanism 214 configured to automatically move Probe 212. In some embodiments, Robotic Mechanism 214 may be controlled by guidance instructions generated by Processing Unit 230 in order to adjust probe position, orientation, or contact pressure during ultrasound acquisition.

[0105] In some exemplary embodiments, Processing Unit 230 may comprise one or more processors and memory storing instructions that, when executed, cause System 200 to perform guided ultrasound acquisition and analysis. Processing Unit 230 may include a Memory 240 configured to store program instructions and Trained Models 242 used for ultrasound data analysis and guidance generation. In some embodiments, Trained Models 242 may include two or more trained machine learning models configured to perform different analysis and guidance tasks associated with ultrasound acquisition. In some embodiments, the system may alternatively utilize a single trained model configured to perform multiple tasks, or more than two trained models each configured for a respective stage of the acquisition workflow. In some exemplary embodiments, the trained models may be trained using multimodal ultrasound datasets labeled by human operators, clinicians, or physicians based on clinical and medical knowledge. The training data may include ultrasound-derived data streams acquired synchronously or near-synchronously, including B-mode imaging, color Doppler or other Doppler signals, PWD signals, electrocardiogram (ECG) signals, and spatial or motion information obtained from a three-dimensional sensor coupled to or positioned on the transducer or the probe, such as position, orientation, movement trajectory, or contact-related parameters. Additionally, or alternatively, the training data may include annotations indicating anatomical structures, vessel segments, or signal quality. In some embodiments, the training process may associate these multiple modalities across time to enable the model to learn relationships between anatomy, flow dynamics, cardiac cycle timing, probe positioning, and acquisition quality.

[0106] In some embodiments, labeled datasets may be generated using annotations provided by expert sonographers or cardiologists identifying clinically relevant features within ultrasound data. The labeled datasets may further include annotations indicating hemodynamic patterns, probe placement quality, and diagnostically relevant temporal events. In some embodiments, the models may be trained to use the combined multimodalISUS Ref: 53516-003 PCTinputs to identify clinically relevant features, assess scan adequacy, and improve interpretation robustness across different operators, patients, and acquisition conditions.

[0107] In some embodiments, the trained models may include a first trained model configured to analyze ultrasound data to determine a patient- specific vessel segment suitable for Doppler-based flow assessment. The first trained model may analyze B-mode ultrasound data obtained at the initial cardiac imaging view. In some embodiments, the first trained model may be trained using labeled data comprising transducer spatial position and orientation data provided by a positioning sensor attached to the probe during training, and may be configured to correlate B-mode ultrasound image features with anatomical alignment and to identify clinically relevant anatomical views

[0108] Additionally, or alternatively, the trained models may further include a second trained model configured to analyze ultrasound data acquired at the identified vessel segment in order to determine a patient- specific final acquisition position optimized for Doppler signal acquisition. The second trained model may analyze color Doppler flow data overlaid on B-mode ultrasound images to identify regions corresponding to blood flow. In some embodiments, the second trained model may be trained using PWD signals synchronized with cardiac cycle phases based on ECG signals, and may be configured to identify diastolic-dominant flow regions and optimize sampling locations accordingly.

[0109] In some embodiments, one or more of the trained models may comprise deep learning models or reinforcement learning models trained to predict probe movements or acquisition parameters that improve signal quality. Training of the reinforcement learning model may be based on feedback derived from human annotations evaluating probe movements between ultrasound views. In some embodiments, training may be formulated as a sequential decision-making process under partial observability, and may utilize one or more of supervised learning, imitation learning from expert demonstrations, reinforcement learning using reward signals, self- supervised representation learning, or teacher-student distillation techniques. In some embodiments, privileged information available during training, including probe pose or future outcome labels, may be utilized to improve model performance even if such information is not available during inference. In some embodiments, training and evaluation may be performed within a simulated or learned environment configured to model ultrasound imaging dynamics and probe-tissue interaction, thereby enabling iterative refinement of the trained models. The trainedISUS Ref: 53516-003 PCTmodels may be configured to operate in real time during inference using imaging data to generate guidance instructions for probe positioning and acquisition.

[0110] Processing Unit 230 may further include a Data Processing Module 250 configured to process ultrasound data acquired by Probe 212 and prepare the data for analysis by the trained models.

[0111] Processing Unit 230 may further include an Analysis Module 260 configured to analyze ultrasound signals. Analysis Module 260 may include a Segment Identification Module 262 and a Signal Evaluation Module 264.

[0112] Segment Identification Module 262 may be configured to identify vessel regions within ultrasound data that exhibit blood flow patterns suitable for Doppler signal acquisition. In some embodiments, the module may detect diastolic-dominant flow patterns occurring predominantly during a diastolic phase of a cardiac cycle.

[0113] Signal Evaluation Module 264 may evaluate acquired Doppler signals in order to determine whether the signals correspond to valid physiological blood flow. In some embodiments, the module may compute a clinical validity score and a technical quality score associated with the acquired signal. The clinical validity score may be based on blood flow pattern analysis, while the technical quality score may be based on metrics such as signal-to-noise ratio, measurement stability, or other noise metrics. Signal Evaluation Module 264 may determine whether the clinical validity score and technical quality score exceed predefined thresholds.

[0114] In some embodiments, verification of a Doppler signal may include segmenting the acquired signal into systolic and diastolic phases and comparing the segmented signal to expected physiological flow patterns. Signal Evaluation Module 264 may further apply filtering to remove clinical noise sources including artifacts caused by unrelated vascular structures, valves, or transient physiological events.

[0115] In some embodiments, Processing Unit 230 may further include a Robotic Positioning Mechanism 270 configured to control Robotic Mechanism 214 of Ultrasound Imaging System 210 in order to automatically adjust probe position during ultrasound acquisition.ISUS Ref: 53516-003 PCT

[0116] Processing Unit 230 may also include a Guidance Module 280 configured to generate guidance instructions for ultrasound acquisition. Guidance Module 280 may include a Probe Guidance Module 282 and a Cursor Placement Module 284.

[0117] Probe Guidance Module 282 may generate real-time guidance instructions configured to guide repositioning and orientation of Probe 212 toward a target vessel segment. The guidance instructions may include predicted probe movements along multiple axes including lateral translation, angular adjustment, depth adjustment, and contact pressure modification.

[0118] In some embodiments, Guidance Module 280 may further be configured to dynamically adjust one or more ultrasound acquisition parameters based on real-time analysis of ultrasound data and probe state data. The ultrasound acquisition parameters may include, for example, gain, imaging frequency, pulse repetition frequency (PRF), imaging depth, Doppler settings, or combinations thereof. In some embodiments, the system may select or modify an imaging preset based on anatomical context, probe position, and detected flow characteristics, thereby improving signal quality across different acquisition conditions.

[0119] Cursor Placement Module 284 may determine a location within the ultrasound image corresponding to an optimal sampling location for Doppler signal acquisition. In some embodiments, the module may guide placement of a Doppler sampling cursor at the identified acquisition position or may automatically reposition the cursor to a detected region corresponding to blood flow.

[0120] In some embodiments, when the acquired signal does not satisfy predefined quality thresholds, Processing Unit 230 may perform iterative optimization by updating probe position and / or cursor placement and repeating the acquisition process.

[0121] In some exemplary embodiments, User Interface / Display 220 may include a Visual Guidance Module 222 and a Signal Output Module 224. Visual Guidance Module 222 may present guidance instructions to a human operator in the form of visual cues, graphical indicators, or audio instructions assisting the operator in repositioning the ultrasound probe. Signal Output Module 224 may present acquired Doppler signals, ultrasound images, and analysis results generated by Processing Unit 230.ISUS Ref: 53516-003 PCT

[0122] In some embodiments, the guidance instructions generated by Guidance Module 280 may be provided both to User Interface / Display 220 for presentation to an operator and to Robotic Positioning Mechanism 270 for automated probe control.

[0123] Referring now to Figure 3 showing a flowchart diagram of a method in accordance with some exemplary embodiments of the disclosed subject matter. Method 300 may be a method for guided acquisition of coronary blood flow using ultrasound imaging. Method 300 may be performed by System 200 of Figure 2 or by Ultrasound Acquisition System 100 of Figure 1. In some embodiments, Method 300 may be a computer-implemented method.

[0124] Method 300 begins at step 310, which may comprise obtaining real-time ultrasound data acquired from a patient by an ultrasound probe positioned to obtain an initial cardiac imaging view. The real-time ultrasound data may comprise B-mode ultrasound data and color Doppler flow data. The initial cardiac imaging view may comprise a preliminary ultrasound view of the heart acquired prior to Al-guided probe repositioning toward a patient- specific coronary artery segment.

[0125] In some embodiments, experienced human operators, such as trained sonographers or physicians, may know how to position the ultrasound probe to achieve such an initial cardiac imaging view based on clinical experience. In some cases, the initial cardiac imaging view may correspond to a standard cardiac ultrasound view that may be obtained by placing the probe on the chest of the patient in a commonly used probe position during cardiac ultrasound examinations. Additionally or alternatively, the system may provide initial guidance instructions to an operator to assist in positioning the ultrasound probe at an initial standard location and orientation in order to obtain the initial cardiac imaging view. Such guidance instructions may be presented through visual, audio, or graphical cues provided by the ultrasound system or associated display device.

[0126] In some embodiments, the initial cardiac imaging view may be selected from a set of predefined cardiac imaging views. Selection of the initial view may be based on patient-specific characteristics, including demographic information, anatomical characteristics, clinical indications, medical history, or other relevant patient features. For example, the system may select a particular starting view based on factors such as patient body habitus, suspected cardiac condition, prior imaging data, or other clinical considerations.ISUS Ref: 53516-003 PCT

[0127] At step 320, the real-time ultrasound data may be analyzed using a first trained model to determine a patient-specific coronary artery segment suitable for Doppler-based flow assessment. In some embodiments, analyzing the real-time ultrasound data using the first trained model may comprise analyzing B-mode data acquired at the initial cardiac imaging view. The first trained model may identify anatomical structures, vessel trajectories, and spatial relationships within the ultrasound image in order to determine a coronary artery segment that is suitable for Doppler signal acquisition.

[0128] In some exemplary embodiments, the first trained model may be generated using self-supervised learning techniques applied to large datasets of unlabeled cardiac ultrasound videos to create foundational models capable of interpreting cardiac imagery without explicit annotations. Such foundational models may learn general representations of cardiac anatomy, motion patterns, and ultrasound imaging characteristics. The foundational models may subsequently be refined using clinically annotated datasets and alignment with corresponding textual Electronic Medical Record (EMR) notes, thereby creating robust embedding representations suitable for downstream tasks including vessel localization, probe navigation, and acquisition optimization.

[0129] Additionally, or alternatively, the first trained model may be trained using supervised learning techniques. In supervised learning, training datasets may include ultrasound data acquired during examinations performed by experienced human operators, such as expert sonographers or cardiologists. During data collection, movements of the ultrasound probe may be monitored and recorded while the operators manipulate the probe to reach a patient-specific coronary artery segment suitable for Doppler-based flow assessment. The recorded probe movements may include information describing probe position, orientation, tilt, rotation, and spatial placement relative to the patient’s body surface, as well as imaging parameters and resulting ultrasound frames. In some embodiments, probe movement information may be captured using position sensors, inertial sensors, electromagnetic tracking systems, or robotic positioning systems associated with the ultrasound probe.

[0130] In some embodiments, training datasets may include labels identifying probe locations, probe orientations, and ultrasound image states corresponding to successful identification of coronary artery segments suitable for Doppler signal acquisition. The labels may be generated by expert operators or physicians based on clinical knowledgeISUS Ref: 53516-003 PCTand interpretation of the ultrasound images. Using these labeled datasets, the first trained model may learn relationships between ultrasound image features, probe positioning parameters, and successful acquisition locations, thereby enabling the model to predict probe movements or target vessel segments during subsequent examinations. In some embodiments, the training process may combine self- supervised learning, supervised learning, and reinforcement learning approaches in order to improve model robustness and generalization across diverse patient anatomies and imaging conditions.

[0131] At step 330, real-time guidance instructions may be provided to guide repositioning and orientation of the ultrasound probe toward the identified coronary artery segment. In some embodiments, the guidance instructions may include visual guidance, audio guidance, haptic guidance, or any combination thereof, provided to a human operator of the ultrasound probe. In some embodiments, the guidance instructions may be determined based on a combination of the real-time ultrasound data and probe state data obtained from one or more sensors associated with the ultrasound probe.

[0132] In some exemplary embodiments, visual guidance may be presented on a display device associated with the ultrasound imaging system. The visual guidance may include graphical cues, overlays, or markings displayed on an ultrasound image or on a graphical representation of the patient’s body. For example, the system may display a model or schematic representation of the patient’s torso, indicating a recommended probe placement location, probe orientation, or target anatomical region. Visual indicators may include arrows, highlighted regions, probe icons, trajectory paths, or other graphical elements indicating how the probe should be moved or rotated in order to reach a desired imaging position.

[0133] In some exemplary embodiments, audio guidance may include audible instructions provided to the operator indicating probe movements required to reach the target vessel segment. The instructions may include commands such as instructions to move the probe in a particular direction, adjust probe angle, increase or decrease probe pressure, or reposition the probe relative to anatomical landmarks. Additionally or alternatively, haptic guidance may be provided, for example through tactile feedback mechanisms associated with the probe or probe holder. Such haptic feedback may include push or pull forces, vibration cues, or directional feedback indicating how the probe should be moved in order to improve the imaging view.ISUS Ref: 53516-003 PCT

[0134] In some embodiments, the guidance instructions may be predicted along multiple axes of probe movement. Probe motion may be represented within a three-dimensional coordinate space relative to the patient’s body surface, allowing the system to guide the probe through a combination of translational and rotational movements. For example, translational movements may include lateral or longitudinal translation along the patient’s body surface, while rotational movements may include probe rotation around its central axis, tilting of the probe relative to the body surface, or rocking movements used to change the ultrasound beam orientation. Additional probe adjustments may include depth adjustments, probe angle adjustments, and modifications to contact pressure applied between the probe and the patient’s body surface.

[0135] Additionally, or alternatively, the same guidance logic used to generate instructions for a human operator may additionally or alternatively be used to control an autonomous transducer system or robotic positioning mechanism configured to automatically move the ultrasound probe according to the predicted probe movements. In such embodiments, the system may automatically reposition the probe along the predicted movement axes in order to reach the identified coronary artery segment suitable for Doppler-based flow assessment.

[0136] At step 340, the real-time ultrasound data at the patient-specific coronary artery segment may be analyzed using a second trained model to identify a patient- specific final acquisition position optimized for PWD signal acquisition. In some embodiments, the second trained model may be trained during the same training process used to generate the first trained model, or alternatively the second trained model may be trained separately from the first trained model. In some embodiments, the second trained model may be trained using ultrasound datasets collected during Doppler examinations performed by experienced operators. Training data may include ultrasound image frames, color Doppler flow information, and associated pulsed-wave Doppler signals obtained from various probe positions and sampling locations.

[0137] In some embodiments, the training process may include manual labeling of PWD signals by expert clinicians, such as cardiologists or experienced sonographers. The labeled datasets may include annotations indicating whether a particular Doppler signal corresponds to valid coronary artery flow and may further include labels identifying the accurate cursor placement location, probe orientation, and probe position associated withISUS Ref: 53516-003 PCTsuccessful signal acquisition. In some embodiments, the training data may include associations between ultrasound image features, Doppler flow patterns, and the corresponding cursor locations within the ultrasound image that produced high-quality Doppler signals.

[0138] In some embodiments, the second trained model may learn relationships between ultrasound image features, Doppler flow characteristics, and optimal sampling positions, thereby enabling the model to predict the most suitable location for Doppler cursor placement during subsequent examinations. In some embodiments, the training process may incorporate supervised learning techniques, reinforcement learning approaches, or hybrid training frameworks combining multiple machine learning methodologies.

[0139] In some embodiments, analyzing the real-time ultrasound data using the second trained model may comprise analyzing color Doppler flow data in order to identify regions corresponding to blood flow within coronary arteries. In some embodiments, the analysis may comprise analyzing an overlay of color Doppler flow data on B-mode ultrasound images acquired at the patient- specific coronary artery segment. In some embodiments, the second trained model may automatically identify diastolic flow regions by detecting flow occurring predominantly during a diastolic phase of a cardiac cycle, which may be characteristic of coronary artery blood flow patterns. Identification of diastolic flow regions may assist in determining suitable locations for Doppler signal acquisition.

[0140] In some embodiments, the analysis performed by the second trained model may be based on one or more types of ultrasound data, including but not limited to color Doppler flow data, B-mode ultrasound images, spectral Doppler signals, power Doppler data, or combinations thereof. In some embodiments, the analysis may further incorporate temporal ultrasound information across multiple frames, enabling the model to detect dynamic blood flow patterns associated with coronary circulation.

[0141] Additionally, or alternatively, the second trained model may analyze ultrasound data together with additional physiological signals, such as electrocardiogram (EKG) signals, in order to determine cardiac cycle phases and improve detection of diastolic-dominant blood flow regions. By combining ultrasound imaging data with temporalISUS Ref: 53516-003 PCTphysiological signals, the model may more accurately determine a patient- specific final acquisition position optimized for Doppler signal acquisition.

[0142] At step 350, placement of a Doppler sampling cursor on a visualization of the real-time ultrasound data corresponding to the identified patient-specific final acquisition position may be guided. In some embodiments, cursor placement guidance may include providing visual or graphical instructions within an ultrasound imaging system indicating an appropriate sampling location for acquiring PWD data. In some embodiments, the cursor placement guidance may be performed based on the analysis of the second trained model described above. The second trained model may identify image regions corresponding to blood flow within the coronary artery segment and may determine a cursor location that is predicted to produce a high-quality Doppler signal.

[0143] Additionally or alternatively, guiding placement of the Doppler sampling cursor may comprise analyzing color Doppler flow data in real time in order to detect diastolic-dominant flow patterns characteristic of coronary artery blood flow. Based on this analysis, the system may determine a suitable sampling region corresponding to the detected flow and may indicate the location for cursor placement within the ultrasound image.

[0144] In some embodiments, the system may automatically reposition the Doppler sampling cursor within the ultrasound imaging system to the detected flow region. Automatic repositioning may occur when the system determines, based on the analyzed ultrasound data, that the cursor should be moved to a different position in order to improve Doppler signal acquisition. In other embodiments, the system may provide guidance cues to an operator indicating the direction and magnitude of cursor movement required to reach the desired sampling location. Such guidance may be presented through visual indicators, cursor trajectories, highlighting of target regions, or other graphical cues displayed on the ultrasound system interface. In some embodiments, automatic cursor repositioning or operator guidance may be repeated iteratively until the cursor reaches a location suitable for acquiring PWD data corresponding to coronary artery blood flow.

[0145] At step 360, a PWD signal may be acquired from the positioned Doppler sampling cursor at the identified acquisition position. The acquired PWD signal may subsequently be processed for evaluation of coronary circulatory function or dysfunction.ISUS Ref: 53516-003 PCT

[0146] At decision step 370, Method 300 may comprise verifying that the acquired PWD signal corresponds to valid coronary artery flow and satisfies predefined technical quality requirements. In some embodiments, verification may include computing a clinical validity score and a technical quality score associated with the acquired PWD signal. The clinical validity score may be based at least in part on coronary artery flow pattern analysis, while the technical quality score may be based on signal-to-noise ratio (SNR), signal stability, or other noise-related metrics. Additionally, or alternatively, verification may further comprise applying filtering operations to remove clinical noise sources including artifacts caused by unrelated vascular structures, valves, or transient physiological events. The clinical validity score and technical quality score may be computed after filtering.

[0147] In some embodiments, verifying that the acquired PWD signal corresponds to valid coronary artery flow may further comprise segmenting the acquired signal into systolic and diastolic phases and comparing the segmented data to expected physiological coronary flow benchmarks.

[0148] Decision step 370 may further comprise determining whether the clinical validity score and the technical quality score exceed predetermined thresholds. In some embodiments, the thresholds may be adapted to patient- specific parameters. When one or more thresholds are not satisfied, method 300 may initiate iterative optimization by returning to one or more earlier guidance stages in order to update probe positioning and / or Doppler cursor placement. As illustrated in Figure 3, the method may return to step 330 or step 340 for iterative optimization of the acquisition process. When the acquired PWD signal satisfies the predefined quality thresholds, the method may proceed to step 380, where the PWD signal may be used for evaluation of coronary circulatory dysfunction.

[0149] In some embodiments, the trained models used in the method may further process multimodal clinical data including demographic data, vital signs, medical history, Doppler signals, and electrocardiogram (EKG) signals for stenosis evaluation. The models may analyze non-linear relationships between Doppler signals, EKG signals, demographic information, and clinical data, comparing these signals to catheter-based measurements in order to refine diagnostic accuracy. In some embodiments, feature extraction may be performed based on these comparisons, enabling the models toISUS Ref: 53516-003 PCTiteratively improve prediction accuracy and sensitivity. The multimodal clinical data fusion approach may enable comprehensive evaluation of coronary artery stenosis localization and severity.

[0150] Although Figure 3 illustrates a particular sequence of steps, the disclosed subject matter is not limited to the illustrated order. In some embodiments, steps may be performed in a different order, certain steps may be omitted, or additional steps may be included without departing from the scope of the disclosed subject matter.

[0151] Referring now to Figure 4 A, showing a schematic representation of an ultrasound Probe 410 showing probe movement directions during ultrasound acquisition is illustrated. Probe 410 may include a transducer Head 420 and a probe Body 430, which may comprise a handle portion used by an operator during probe manipulation. Multiple directional arrows surrounding Probe 410 indicate various axes of movement available during probe positioning.

[0152] As described previously, Processing Unit 230 may generate real-time guidance instructions configured to guide repositioning and orientation of Probe 410 toward a patient-specific coronary artery segment. The real-time ultrasound data may be acquired while Probe 410 is positioned to obtain an initial cardiac imaging view of the heart.

[0153] In some exemplary embodiments, the real-time guidance instructions may correspond to multiple axes of probe movement, which may collectively represent six degrees of freedom of probe positioning. These movements may include lateral translation, rotational adjustment, tilting, sliding, rocking, depth adjustment, and contact pressure modification. The directional arrows illustrated in Figure 4A represent example probe positioning movements, including sliding movement along a longitudinal axis, rotational and rocking movements around transducer Head 420, angular tilting movements near probe Body 430, lateral translation, and depth adjustment movements perpendicular to the probe orientation. In some embodiments, Probe Guidance Module 282 may generate navigation instructions corresponding to ultrasound probe gestures including rotation, tilt, sliding, and rocking movements, thereby assisting an operator in repositioning Probe 410 toward a target anatomical structure or coronary artery segment.

[0154] Referring now to Figure 4B, showing a schematic illustration of probe positioning guidance and probe movement directions during ultrasound acquisition isISUS Ref: 53516-003 PCTshown. In this example, Probe 410 is illustrated positioned against a patient Body Surface 440 during an ultrasound examination. Multiple directional arrows surrounding the probe illustrate the various axes of movement available during signal acquisition.

[0155] Figure 4B further illustrates individual probe manipulation gestures. Four smaller diagrams shown below the main illustration represent example probe movement types including rotation movement 442, tilt movement 444, sliding movement 446, and rocking movement 448. The rotation movement 442 may comprise rotational adjustment of the ultrasound probe 410 about a longitudinal axis of the probe body 430. The tilt movement 444 may comprise angular adjustment of the transducer head 420 relative to the patient body surface 440. The sliding movement 446 may comprise lateral translation of the ultrasound probe 410 along the patient body surface 440. The rocking movement 448 may comprise oscillating angular movement of the transducer head 420 while maintaining contact with the patient body surface 440.

[0156] In some embodiments, the real-time guidance instructions generated by the system may be presented to a human operator as visual guidance, audio guidance, or a combination thereof. Visual guidance may include graphical cues or directional indicators presented on a display device, while audio guidance may include spoken instructions or auditory signals corresponding to probe movement directions. In some embodiments, Probe Guidance Module 280 may also provide additional procedural guidance during signal acquisition. For example, the system may provide breathing instructions to a patient in order to improve ultrasound image quality. In some embodiments, the system may prompt the operator to adjust ultrasound presets based on anatomical position, artery orientation, or depth associated with a target coronary artery segment. Probe Guidance Module 282 may dynamically update guidance instructions based on real-time analysis of the ultrasound data, thereby adapting probe positioning guidance during the acquisition process.

[0157] Referring now to Figure 4C showing another schematic illustration of probe positioning guidance during ultrasound acquisition is shown. On the left side of Figure 4C, Probe 410 is illustrated positioned against a curved surface representing a patient Body Surface 440. Probe Body 430 includes an elongated handle portion, while Transducer Head 420 forms a contact surface used for ultrasound signal acquisition.ISUS Ref: 53516-003 PCT

[0158] Directional arrows surrounding Probe 410 illustrate additional probe positioning movements. Curved arrows near an upper portion of the probe illustrate rotational or tilting movements. A double-headed horizontal arrow illustrates lateral sliding movement across the patient body surface. A double-headed vertical arrow below the probe illustrates depth adjustment or pressure modification movement that may alter probe contact with the patient surface. On the right side of Figure 4C, a series of operator hand positions 450, 452, 454, and 456 illustrate example manual probe manipulation gestures used during probe repositioning. The sequence of hand positions represents different probe handling techniques that an operator may use when adjusting probe orientation and location in response to guidance instructions generated by the system.

[0159] In some exemplary embodiments, Probe Guidance Module 280 may employ machine learning models to determine appropriate probe movement instructions. In some embodiments, the guidance model may utilize reinforcement learning algorithms, such as Proximal Policy Optimization (PPO), to predict optimal probe actions based on ultrasound data and prior training data. Training of the guidance model may involve offline reinforcement learning or behavior cloning using datasets of ultrasound examinations performed by experienced operators. A reward model may be trained to produce a score for each imaging view based on human annotator labels indicating whether a new probe state improves signal quality. In some embodiments, the reinforcement learning model may receive as input both ultrasound-derived features and probe state data obtained from one or more sensors, and may output probe movement actions and / or ultrasound acquisition parameter adjustments configured to improve Doppler signal quality.

[0160] In some embodiments, the Trained Models 242 may analyze real-time ultrasound data using a first trained model configured to determine a patient- specific coronary artery segment suitable for Doppler-based flow assessment, and a second trained model configured to identify a final acquisition position optimized for Doppler signal acquisition. At least one of these trained models may comprise a deep reinforcement learning model trained using human feedback on probe movements between imaging views that enhance signal quality.

[0161] The disclosed system may guide probe positioning in order to achieve or transition between standard cardiac ultrasound views, including parasternal long-axisISUS Ref: 53516-003 PCT(PLAX), parasternal short-axis (PSAX), and apical four-chamber (A4C) views. In some embodiments, the same guidance logic used to provide instructions to a human operator may also be used to control an autonomous transducer system or robotic positioning mechanism configured to automatically reposition the ultrasound probe.

[0162] In such embodiments, Robotic Positioning Mechanism 270 may receive guidance instructions from Probe Guidance Module 280 and translate the instructions into mechanical movements of a robotic mechanism (such as mechanism 214) configured to reposition Probe 410 toward patient- specific coronary artery segments.

[0163] Referring now to Figures 5A-5B showing ultrasound images illustrating identification of a flow region and placement of a Doppler sampling cursor, and a PWD signal, in accordance with some exemplary embodiments of the disclosed subject matter.

[0164] Figure 5 A (ultrasound imaging display 500a) illustrates a realistic ultrasound image, while Figure 5B (ultrasound imaging display 500b) illustrates a corresponding schematic representation of the image of Figure 5 A.

[0165] Ultrasound imaging Display 500a, 500b may comprise an upper imaging portion and a lower spectral Doppler portion. The upper imaging portion includes an Ultrasound Imaging Sector 510, representing a B-mode image of anatomical tissue, within which Doppler information may be overlaid. A Flow Region 520 may be identified within Ultrasound Imaging Sector 510, for example based on color Doppler information indicative of blood flow within a vessel. In some exemplary embodiments, Ultrasound Imaging Sector 510 may comprise B-mode data and color Doppler flow data, wherein the B-mode data provides anatomical context and the color Doppler flow data provides flow-related information used for identifying Flow Region 520.

[0166] In some exemplary embodiments, a Doppler sampling Cursor 530 may be positioned within the identified Flow Region 520. Cursor 530 may define a sampling volume or gate from which Doppler signals are acquired. In some embodiments, the positioning of Cursor 530 may be performed manually by a user and / or automatically by a processing unit based on analysis of Flow Region 520, including real-time analysis of color Doppler flow data to detect diastolic-dominant flow regions and to position the Cursor 530 within a region optimized for coronary flow acquisition.ISUS Ref: 53516-003 PCT

[0167] A magnified inset View 540 may be provided, illustrating an enlarged portion of the ultrasound imaging sector 510, for example to enhance visualization of Flow Region 520 and facilitate accurate placement of Cursor 530 within a target vessel. In some embodiments, the magnified inset View 540 may be used as part of a user interface configured to provide visual guidance instructions to an operator to assist in accurate placement of Cursor 530.

[0168] The lower portion of the display may include a Spectral Doppler Display 560 presenting a PWD waveform corresponding to blood flow at the sampling location defined by Cursor 530. The Doppler waveform may represent blood flow velocity over time. A Baseline 570 may be provided, representing a zero-velocity reference. A positive flow Component 572, corresponding to flow in a first direction relative to the transducer, may be displayed above Baseline 570, while a negative flow Component 574, corresponding to flow in an opposite direction, may be displayed below Baseline 570. In some exemplary embodiments, the Doppler waveform may be processed to segment the signal into systolic and diastolic phases, and to extract flow-related features including peak velocity, flow envelope characteristics, and temporal flow patterns, which may be compared to predefined coronary flow benchmarks.

[0169] In some embodiments, Spectral Doppler Display 560 may further include a physiological reference signal, such as an ECG Signal 580, which may be used to correlate the Doppler waveform with phases of the cardiac cycle and to enable phasebased analysis of coronary flow, including identification of diastolic-dominant flow patterns.

[0170] In some exemplary embodiments, the Doppler waveform displayed in Spectral Doppler Display 560 is obtained from the sampling location defined by Cursor 530 positioned within Flow Region 520 in Ultrasound Imaging Sector 510, thereby providing a linkage between spatial localization of blood flow and corresponding velocity measurements. In some embodiments, the identification of Flow Region 520, positioning of Cursor 530, and acquisition of the Doppler waveform are performed as part of a guided acquisition process, wherein real-time ultrasound data is analyzed using one or more trained models to determine a patient- specific coronary artery segment and a patientspecific final acquisition position optimized for PWD signal acquisition.ISUS Ref: 53516-003 PCT

[0171] In some embodiments, the system may further provide real-time guidance instructions to guide repositioning and / or orientation of the ultrasound probe and / or placement of Cursor 530, and may iteratively update the acquisition process based on analysis of the acquired Doppler waveform.

[0172] In some exemplary embodiments, the acquired Doppler waveform may be evaluated to determine whether it corresponds to valid coronary artery flow and meets predefined technical quality requirements. Such evaluation may comprise computing a clinical validity score based on coronary flow pattern analysis and a technical quality score based on signal characteristics including signal-to-noise ratio, measurement stability, or noise metrics.

[0173] In some embodiments, the Doppler waveform may be filtered to remove clinical noise sources including artifacts caused by unrelated vascular structures, valves, or transient physiological events prior to computing the scores. In some embodiments, when one or more of the clinical validity score or technical quality score do not exceed corresponding thresholds, the system may initiate an iterative optimization process, returning to one or more previous stages to update probe positioning and / or Cursor 530 placement.

[0174] Referring now to Figure 6 showing a flowchart of a feedback-based acquisition control process for iteratively evaluating and optimizing a PWD signal, in accordance with some exemplary embodiments of the disclosed subject matter.

[0175] The feedback-based acquisition control process 600 may be performed by Processing Unit 230 of Figure 2 or Ultrasound Imaging Device 115 of Figure 1. Process 600 provides a closed-loop verification mechanism configured to verify that an acquired PWD signal corresponds to valid coronary artery flow and meets predefined technical quality requirements.

[0176] Step 610 comprises acquiring a PWD signal at a current probe position. The PWD signal may be acquired from positioned Doppler Sampling Cursor 104 at the patient-specific final acquisition position as described previously.

[0177] At step 620, the acquired PWD signal may be evaluated. In some embodiments, evaluation of the PWD signal may be performed by Signal Evaluation Module 264 of the Analysis Module 260.ISUS Ref: 53516-003 PCT

[0178] At step 630, the system may compute a clinical validity score based on coronary artery flow pattern analysis. The clinical validity score may represent a measure of whether the acquired signal corresponds to physiologically plausible coronary artery blood flow. In some embodiments, the coronary artery flow pattern analysis may include identifying flow characteristics associated with coronary artery hemodynamics, such as diastolic-dominant flow patterns that distinguish coronary arterial flow from other cardiovascular structures.

[0179] In some embodiments, Signal Evaluation Module 264 may segment the acquired PWD signal into systolic and diastolic phases and compare the segmented data to clinical coronary flow benchmarks. Such benchmarks may include expected diastolic-to-systolic flow ratios, characteristic velocity profiles, or other flow features associated with coronary arterial circulation.

[0180] At step 640, the system may compute a technical quality score associated with the acquired PWD signal. The technical quality score may represent a measure of signal reliability and measurement quality. In some embodiments, the technical quality score may be based at least in part on signal-to-noise ratio, measurement stability, or other noise metrics. Measurement stability may be evaluated by analyzing temporal consistency of flow velocity measurements across multiple cardiac cycles. Additional noise metrics may include spectral broadening analysis, baseline stability, artifact detection, or signal discontinuity detection.

[0181] In some embodiments, prior to computing the clinical validity score and the technical quality score, Processing Unit 230 may apply filtering operations to remove clinical noise sources. Such filtering may remove artifacts caused by unrelated vascular structures, valve motion artifacts, flow signals from non-coronary vessels, or transient physiological events that may contaminate the PWD signal. Signal Evaluation Module 264 may compute the clinical validity score and the technical quality score after the filtering process.

[0182] At decision step 650, the feedback-based acquisition control process 700 may determine whether the clinical validity score and the technical quality score exceed predetermined thresholds. In some embodiments, the predetermined clinical validity and technical quality thresholds may be adapted to patient-specific parameters. Patientspecific parameters may include anatomical variations, body habitus, acoustic windowISUS Ref: 53516-003 PCTquality, demographic factors, and clinical history, which may affect expected Doppler signal characteristics. In some embodiments, the predetermined thresholds may be dynamically adjusted based on patient demographics, medical history, and real-time assessment of imaging conditions.

[0183] When one or more thresholds are not met, the feedback-based acquisition control process 600 proceeds to step 660, in which the probe position and / or cursor placement is updated. Processing Unit 230 may initiate iterative optimization that returns to one or more previous guidance stages to update probe position and / or cursor placement. The iterative optimization may include adjusting the position and orientation of Probe 212 through Probe Guidance Module 282, and / or repositioning Doppler Sampling Cursor 104 through Cursor Placement Module 284. After the adjustments are applied, the process may return to step 610 to acquire a new PWD signal at the updated position.

[0184] In some embodiments, the iterative optimization process may further include jointly optimizing probe positioning and ultrasound acquisition parameters based on both the real-time ultrasound data and probe state data obtained from one or more sensors. In such embodiments, the system may operate as a closed-loop control system that continuously adapts both physical probe configuration and imaging parameters in response to signal evaluation.

[0185] When the clinical validity score and the technical quality score exceed the predetermined thresholds, the feedback-based acquisition control process 600 proceeds to step 670, where the validated PWD signal is output for further analysis. In some embodiments, the validated PWD signal may be provided to the Signal Output Module 224 for display and diagnostic evaluation, including evaluation of coronary circulatory dysfunction.

[0186] The feedback-based acquisition control process 600 may continue iteratively until a PWD signal satisfying both clinical validity and technical quality requirements is obtained. Accordingly, process 600 provides a closed-loop acquisition control mechanism that dynamically adapts probe positioning and sampling location based on real-time signal evaluation, thereby improving the reliability and diagnostic value of the acquired Doppler signal under varying patient- specific conditions and imaging environments.ISUS Ref: 53516-003 PCT

[0187] Referring now to Figures 7A-7B, showing a multimodal training data representation, in accordance with some exemplary embodiments of the disclosed subject matter.

[0188] Figure 7A (Multimodal Training Data Display 700a) illustrates a realistic representation of multimodal data acquired during ultrasound examinations and / or training sessions. Figure 7B (Multimodal Training Data Display 700b) illustrates a corresponding schematic representation of the multimodal data of Figure 7 A.

[0189] As shown in Figures 7A-7B, the multimodal training data representation 700a, 700b may comprise multiple synchronized data modalities associated with ultrasound acquisition. The multimodal data may include: ultrasound imaging data, including B-Mode Image Data 710 and additional Anatomical Representations 720, corresponding to different imaging views of a cardiac region; Doppler-related data, including Color Doppler Data 730, indicative of blood flow within one or more vessels; and probe-related spatial data, including a transducer of Probe Representation 740 within a spatial coordinate system, indicative of probe position and orientation during acquisition.

[0190] In some exemplary embodiments, the multimodal training data may further comprise time-dependent signals associated with probe operation and signal acquisition, including: Probe Position Data 750, representing spatial movement of the probe over time; Probe Orientation Data 760, representing angular orientation of the probe across one or more axes; and Signal Quality Data 770, representing one or more quality metrics associated with acquired Doppler signals.

[0191] In some embodiments, Signal Quality Data 770 may correspond to clinical validity measures, technical quality metrics, or reward signals associated with Doppler signal acquisition.

[0192] In some embodiments, the Multimodal Training Data Representation 700a, 700b may be generated using data collected from real ultrasound examinations, simulated acquisitions, or annotated datasets. The data may be temporally synchronized such that ultrasound image data, Doppler signals, probe position and orientation data, and physiological signals correspond to a common time reference.

[0193] In some exemplary embodiments, the multimodal training data represented in Figures 7A-7B may correspond to, or form part of, the Ultrasound Training Data 802,ISUS Ref: 53516-003 PCTLabeled Probe Positions 804, and / or Expert Annotations 806 used in Training Architecture 800 of Figure 8. For example: Ultrasound Imaging Data 710, 720 and Doppler Data 730 may correspond to Ultrasound Training Data 802; Probe Position And Orientation Data 750, 760 may correspond to Labeled Probe Positions 804; and Signal Quality Data 770 and / or derived annotations may correspond to Expert Annotations 806 and / or inputs to a reward model.

[0194] In some embodiments, the multimodal data may be used to train one or more machine-learning models to: identify anatomical regions of interest, including coronary artery segments; predict probe positioning or movement actions; identify target acquisition locations optimized for Doppler signal acquisition; evaluate Doppler signal quality; and determine whether an acquired signal satisfies predefined clinical or technical criteria.

[0195] In some embodiments, the multimodal data may be used to train models that learn relationships between anatomical image features, Doppler-derived flow information, cardiac cycle timing, and probe motion dynamics, thereby enabling guided acquisition of coronary Doppler signals as described in the claims.

[0196] Referring now to Figure 8 showing a schematic diagram of a training architecture for training one or more machine-learning models configured to assist in ultrasound probe positioning and coronary Doppler signal acquisition, in accordance with some exemplary embodiments.

[0197] In some exemplary embodiments, Training Architecture 800 may be used to generate the Trained Models 242 stored in Memory 240 of Processing Unit 230 as described previously with reference to Figure 2. Training Architecture 800 may utilize multiple data sources including ultrasound Training Data 802, Labeled Probe Positions 804, and Expert Annotations 806. Such data sources may be collected from previously performed ultrasound examinations, simulated ultrasound acquisitions, or curated clinical datasets.

[0198] In some exemplary embodiments, the multimodal training data representations described previously with reference to Figures 7A-7B may correspond to, or form part of, the data utilized by Training Architecture 800. For example, the ultrasound image data and Doppler-related data represented in multimodal training data displays 700a and 700bISUS Ref: 53516-003 PCTmay correspond to portions of Ultrasound Training Data 802, probe pose and movement information may correspond to Labeled Probe Positions 804, and signal-quality-related data and / or expert-evaluated acquisition outcomes may correspond to Expert Annotations 806.

[0199] In some embodiments, Ultrasound Training Data 802 may include raw or processed ultrasound image sequences, Doppler signals, probe motion trajectories, ECG signals, color Doppler data, PWD data, B-mode image data, three-dimensional sensor data associated with the transducer, or associated metadata obtained during scanning procedures. Such sensor data may include probe position, orientation, motion trajectory, and contact-related parameters, and may be used during training to enable the models to learn relationships between probe state, ultrasound signal characteristics, and acquisition quality. Labeled Probe Positions 804 may include ground-truth probe pose information such as spatial coordinates, orientation parameters, motion vectors, or acquisition positions corresponding to successful Doppler measurements. Expert Annotations 806 may include clinician-provided labels, quality assessments, anatomical identifications, temporal event markings, hemodynamic interpretations, or guidance signals reflecting expert knowledge regarding proper probe placement and signal acquisition. In some embodiments, Ultrasound Training Data 802 may further include image data corresponding to initial cardiac imaging views acquired prior to guided repositioning toward a coronary artery segment, as well as image data corresponding to patient- specific coronary artery segments selected for Doppler-based flow assessment.

[0200] In some embodiments, the training phase described in Training Architecture 800 may include preprocessing, synchronization, and temporal alignment of the multiple input modalities so that data acquired from different sources during an examination may be correlated in time and space. For example, B-mode data, color Doppler data, PWD signals, ECG waveforms, and three-dimensional transducer sensor outputs may be aligned to enable the model to learn relationships between anatomical structures, bloodflow characteristics, cardiac-cycle timing, and probe position or movement. In some embodiments, the training phase described in Training Architecture 800 may further include normalization, filtering, segmentation, frame selection, quality-based weighting, or augmentation of one or more modalities to improve model robustness across different operators, patients, devices, and acquisition conditions.ISUS Ref: 53516-003 PCT

[0201] In some exemplary embodiments, the training process described in Model Training Pipeline 808 may be supervised (e.g., through Supervised Learning Module 812), semi-supervised, self-supervised (e.g., through Self-Supervised Learning Module 810), or a combination thereof. During training, the model may be iteratively updated based on comparisons between model outputs and corresponding labels, annotations, or target probe positions, such that the model learns to predict anatomical location, assess signal quality, identify suitable acquisition windows, estimate probe adjustment actions, or determine whether a scan satisfies predefined clinical or procedural criteria. In some embodiments, the trained models may thereby learn multimodal representations that integrate ultrasound appearance, Doppler-derived flow information, physiological timing information from ECG, and spatial information from the transducer sensor to support improved guidance, interpretation, and measurement performance.

[0202] In some embodiments, the trained models may include a first trained model configured to analyze B-mode data at an initial cardiac imaging view to determine a patient-specific coronary artery segment suitable for Doppler-based flow assessment, and a second trained model configured to analyze an overlay of color Doppler flow data on B-mode data at the patient- specific coronary artery segment to identify a patient-specific final acquisition position optimized for pulsed-wave Doppler (PWD) signal acquisition.

[0203] In some embodiments, the second trained model may be trained to identify diastolic flow regions by detecting flow occurring predominantly during a diastolic phase of a cardiac cycle, and to support placement of a Doppler sampling cursor within an identified region. In some embodiments, the system may analyze color Doppler flow data in real time to detect diastolic-dominant flow and may automatically reposition the cursor within an ultrasound system, or provide instructions to an operator for manual repositioning of the cursor, to a detected region.

[0204] In some exemplary embodiments, Ultrasound Training Data 802, Labeled Probe Positions 804, and Expert Annotations 806 may be provided as inputs to a Model Training Pipeline 808. Model Training Pipeline 808 may include several coordinated learning modules for training the system. In some embodiments, Model Training Pipeline 808 may include a Self-Supervised Learning Module 810. Self-Supervised Learning Module 810 may be configured to learn representations from large volumes of unlabeled ultrasound data by identifying intrinsic structures or patterns within the ultrasound signalsISUS Ref: 53516-003 PCTor image sequences. Such representations may improve the ability of the system to generalize across different patients, imaging conditions, and anatomical variations. In some embodiments, Self-Supervised Learning Module 810 and / or one or more downstream modules may process training data synchronized with ECG signals so as to enable learning of cardiac -phase-dependent flow characteristics, including training of a model based on PWD data synchronized with systolic and diastolic phases.

[0205] Additionally, or alternatively, Model Training Pipeline 808 may include a Supervised Learning Module 812. Supervised Learning Module 812 may utilize Labeled Probe Positions 804 and / or Expert Annotations 806 to train models to predict clinically relevant outputs. For example, Supervised Learning Module 812 may train a model to estimate optimal probe positioning, detect coronary flow patterns, assess signal quality, or identify anatomical landmarks. The supervised learning process may employ standard training procedures such as minimizing prediction error between model outputs and ground-truth labels. In some embodiments, the clinically relevant outputs may further include real-time guidance instructions, including visual guidance, audio guidance, or a combination thereof, for a human operator, and / or control outputs for an autonomous transducer system or robotic positioning mechanism configured to move the ultrasound probe.

[0206] In some embodiments, Model Training Pipeline 808 may further include a Reinforcement Learning Module 814 configured to learn probe positioning or acquisition strategies through reward-based optimization. Reinforcement Learning Module 814 may treat probe positioning actions as sequential decisions and may iteratively improve the model's strategy based on feedback signals that reflect the quality of the resulting Doppler signal or imaging outcome. In some embodiments, the sequential decisions learned by Reinforcement Learning Module 814 may correspond to probe adjustment actions including lateral translation, angular adjustment, depth adjustment, and contact pressure modification.

[0207] To provide such feedback, Model Training Pipeline 808 may include a Reward Model 816 based on human feedback. Reward Model 816 may be trained using the expert annotations 806, clinician evaluations, or user feedback indicating the quality or clinical validity of particular probe placements or Doppler signals. Reward Model 816 may generate reward signals used by Reinforcement Learning Module 814 to guide theISUS Ref: 53516-003 PCTtraining process toward clinically meaningful acquisition behaviors. As described previously with reference to Figure 4C, Reward Model 816 may be trained to produce a score per view based on human annotator labels assessing whether a new state is preferable.

[0208] In some embodiments, Reward Model 816 may be configured to generate one or more feedback values based on an acquired Doppler signal, including a clinical validity score and a technical quality score. The clinical validity score may be based at least in part on coronary artery flow analysis, and the technical quality score may be based at least in part on signal-to-noise ratio, measurement stability, or other noise metrics. In some embodiments, such scores may be compared to predetermined thresholds, optionally adapted to patient- specific parameters.

[0209] In some embodiments, prior to computing the clinical validity score and the technical quality score, the Doppler signal may be filtered to remove clinical noise sources including artifacts caused by unrelated vascular structures, valves, or transient physiological events. In some embodiments, the Doppler signal may be segmented into systolic and diastolic phases and compared to predefined clinical coronary flow benchmarks, including assessment of whether the signal exhibits a diastolic-dominant coronary flow pattern.

[0210] Through the coordinated operation of Self-Supervised Learning Module 810, Supervised Learning Module 812, and Reinforcement Learning Module 814, Model Training Pipeline 808 may produce Trained Models 818. Trained Models 818 may include, for example, probe-guidance models, signal quality evaluation models, anatomical localization models, acquisition control models, multimodal fusion models, or decision-support models configured to process and interpret multiple real-time ultrasound-related inputs. In some embodiments, the Al models may be configured to receive and process one or more of B-mode ultrasound data, color Doppler data, pulsed-wave Doppler (PWD) signals, ECG signals, and three-dimensional sensor data associated with the transducer, including transducer position, orientation, and motion over time, and to learn relationships among anatomical appearance, blood-flow information, physiological timing, and probe placement dynamics. In some embodiments, one or more of Trained Models 818 may be configured such that, when a clinical validity score and / or technical quality score fails to satisfy one or more corresponding thresholds, the systemISUS Ref: 53516-003 PCTinitiates iterative optimization that returns to one or more prior guidance stages to update probe position and / or Doppler sampling cursor placement.

[0211] In some embodiments, one or more of the Trained Models 818 may include neural network-based architectures, including convolutional neural networks, recurrent neural networks, transformer-based architectures, temporal sequence models, multimodal fusion architectures, or combinations thereof. Such models may be configured to extract spatial features from ultrasound image data, temporal features from Doppler or ECG signals, and motion- or pose-related features from transducer sensor inputs, and to combine such features to generate predictions, classifications, recommendations, or control outputs. In some embodiments, one or more such models may be trained to correlate B-mode ultrasound image features with anatomical alignment and to determine anatomical views based on the B-mode ultrasound images.

[0212] In some embodiments, the Al models may be trained to identify anatomical targets, predict whether a current insonation view is likely to yield a diagnostically acceptable signal, estimate an operator adjustment to probe position or angle, assess acquisition quality, or determine when a target coronary Doppler signal has been successfully obtained.

[0213] In some exemplary embodiments, the self- supervised learning module may enable the Al models to learn latent structure and multimodal representations from partially labeled or unlabeled ultrasound examinations, the supervised learning module may refine the models using expert annotations and known target outputs, and the reinforcement learning module may optimize sequential probe-guidance behavior or acquisition strategies based on reward criteria associated with successful signal acquisition, image quality, workflow efficiency, or adherence to clinical targets.

[0214] In some embodiments, the deployed Al models may generate real-time outputs including anatomical localization results, scan quality scores, acquisition readiness indicators, recommended probe adjustments, or alerts indicating suboptimal positioning or signal conditions. In some embodiments, the real-time outputs may further include cursor-placement recommendations, automatic cursor repositioning commands, target acquisition location indicators, or outputs indicating whether an acquired signal is suitable for subsequent feature extraction or coronary flow reserve computation.ISUS Ref: 53516-003 PCT

[0215] Trained Models 818 may subsequently be deployed within an ultrasound imaging system, processing unit, or clinical guidance platform to assist operators in acquiring diagnostically valid coronary Doppler signals or other ultrasound measurements. Trained Models 818 may correspond to Trained Models 242 described previously with reference to Figure 2, including the first trained model configured to determine a patient- specific coronary artery segment and the second trained model configured to identify a patient- specific final acquisition position optimized for PWD signal acquisition.

[0216] In some embodiments, one or more of Trained Models 818 may further be configured, during deployment, to process an acquired pulsed-wave Doppler (PWD) signal to extract coronary blood flow velocity features during a resting physiological state and during a stress physiological state, estimate coronary flow values corresponding to the resting physiological state and the stress physiological state, and compute a non-invasive coronary flow reserve (CFR) value based on a ratio between the estimated coronary flow value during the stress physiological state and the estimated coronary flow value during the resting physiological state.

[0217] In some embodiments, the trained models are not limited to abstract data classification, but are configured to process real-time physical ultrasound acquisition data, including image data, Doppler-derived flow data, and transducer motion or pose data, to generate outputs that control, guide, or modify operation of an ultrasound acquisition process, including probe repositioning, cursor placement, acquisition readiness determination, and signal quality verification. Such processing may improve the technical operation of an ultrasound imaging system by increasing the likelihood of obtaining diagnostically valid coronary Doppler signals, reducing operator dependence, reducing acquisition variability, and improving signal quality.

[0218] Although the illustrated embodiment shows specific modules and data sources, additional training components, preprocessing stages, model architectures, or feedback mechanisms may be incorporated without departing from the scope of the present disclosure. Likewise, the order of training operations and the specific learning techniques used in each module may vary depending on the particular implementation.

[0219] The present invention may be a system, a method, and / or a computer program product. The computer program product may include a computer readable storageISUS Ref: 53516-003 PCTmedium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present invention.

[0220] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0221] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network may comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0222] Computer readable program instructions for carrying out operations of the present invention may be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmwareISUS Ref: 53516-003 PCTinstructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) may execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present invention.

[0223] Aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer readable program instructions.

[0224] These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implementISUS Ref: 53516-003 PCTaspects of the function / act specified in the flowchart and / or block diagram block or blocks.

[0225] The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0226] The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustration, and combinations of blocks in the block diagrams and / or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts or carry out combinations of special purpose hardware and computer instructions.

[0227] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.ISUS Ref: 53516-003 PCT

[0228] The corresponding structures, materials, acts, and equivalents of all means or step plus function elements in the claims below are intended to include any structure, material, or act for performing the function in combination with other claimed elements as specifically claimed. The description of the present invention has been presented for purposes of illustration and description, but is not intended to be exhaustive or limited to the invention in the form disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the invention. The embodiment was chosen and described in order to best explain the principles of the invention and the practical application, and to enable others of ordinary skill in the art to understand the invention for various embodiments with various modifications as are suited to the particular use contemplated.

[0229] A number of implementations have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the disclosure. Accordingly, other implementations are within the scope of the following claims.

Claims

1. ISUS Ref: 53516-003 PCTCLAIMSWhat is claimed is:

1. A computer-implemented method for guided acquisition of coronary blood flow using ultrasound imaging, the method comprising:obtaining real-time ultrasound data acquired from a patient by an ultrasound probe positioned to obtain an initial cardiac imaging view; analyzing the real-time ultrasound data using a first trained model to determine a patient- specific coronary artery segment suitable for Doppler-based flow assessment;providing real-time guidance instructions to guide repositioning and orienting of the ultrasound probe toward the patient-specific coronary artery segment;analyzing the real-time ultrasound data at the patient- specific coronary artery segment using a second trained model, to identify a patient-specific final acquisition position optimized for Pulsed-Wave Doppler (PWD) signal acquisition;guiding placement of a Doppler sampling cursor on a visualization of the real-time ultrasound data to the identified patient-specific final acquisition position, by providing instructions for cursor placement within an ultrasound imaging system;acquiring a PWD signal from the positioned Doppler sampling cursor at the patient- specific final acquisition position.

2. The method of Claim 1 further comprising:verifying that the acquired PWD signal corresponds to valid coronary artery flow and meets predefined technical quality requirements, said verifying comprises:evaluating a clinical validity score and a technical quality score of the PWD signal; anddetermining whether clinical validity score and the technical quality score exceed predetermined clinical validity and technical quality thresholds, wherein the predetermined clinical validity andISUS Ref: 53516-003 PCTtechnical quality thresholds are adapted to patient- specific parameters;when one or more thresholds are not met, initiating iterative optimization that returns to one or more previous guidance stages to update probe position and / or cursor placement.

3. The method of Claim 2, wherein said verifying further comprises applying filtering to remove clinical noise sources including artifacts caused by unrelated vascular structures, valves, or transient physiological events, and computing the clinical validity and the technical quality scores after the filtering.

4. The method of Claim 2, wherein the clinical validity score is based at least in part on coronary artery flow analysis; and wherein the technical quality score is based at least in part on signal-to-noise ratio, measurement stability or other noise metrics.

5. The method of Claim 2, wherein verifying that the acquired PWD signal corresponds to valid coronary artery flow comprises segmenting the acquired PWD signal into systolic and diastolic phases and comparing the segmented data to clinical coronary flow benchmarks.

6. The method of Claim 1,wherein the real-time ultrasound data comprises B-mode data and color Doppler flow data;wherein analyzing the real-time ultrasound data using the first trained model comprises analyzing the B-mode data at the initial cardiac imaging view; andwherein analyzing the real-time ultrasound data using the second trained model comprises analyzing an overlay of the color Doppler flow data on the B- mode data at the patient- specific coronary artery segment.

7. The method of Claim 6, wherein guiding placement of the Doppler sampling cursor comprises analyzing the color Doppler flow data in real time to detect diastolic- dominant flow within a cardiac cycle and automatically repositioning the cursor within the ultrasound device to the detected region.ISUS Ref: 53516-003 PCT8. The method of Claim 1, wherein analyzing the real-time ultrasound data at the patient-specific coronary artery segment using the second trained model comprises automatically identifying diastolic flow regions by detecting flow occurring predominantly during a diastolic phase of a cardiac cycle.

9. The method of Claim 1, wherein the real-time guidance instructions comprise visual guidance, audio guidance, or a combination thereof, to a human operator of the ultrasound probe.

10. The method of Claim 9, wherein a same guidance logic is applied to control an autonomous transducer system or a robotic positioning mechanism configured to automatically move the ultrasound probe.

11. The method of Claim 1, wherein the real-time guidance instructions are predicted along six axes of probe movement.

12. The method of Claim 1, wherein the real-time guidance instructions comprise at least one of lateral translation, angular adjustment, depth adjustment, and contact pressure modification.

13. The method of Claim 1, wherein the first trained model is trained using labeled data with transducer spatial position and orientation data provided by a positioning sensor attached to the transducer during training with sub-millimeter accuracy, validated by expert ultrasound technicians.

14. The method of Claim 13, wherein the first trained model is trained to correlate B- mode ultrasound image features with anatomical alignment and to determine anatomical views based on the B-mode ultrasound images.

15. The method of Claim 1, wherein the second trained model is trained based on PWD data synchronized with cardiac cycle phases based on electrocardiogram (EKG) signals.

16. The method of Claim 1, wherein positioning the Doppler sampling cursor comprises automatically repositioning the cursor within an ultrasound system or providing guidance to an operator to manually reposition the cursor.ISUS Ref: 53516-003 PCT17. The method of Claim 1, wherein guidance, verification, and scoring, and the first and second trained models are adapted based on patient- specific anatomical and physiological parameters.

18. The method of Claim 1, wherein at least one of the first and the second trained models is a deep reinforcement learning model trained using a reward model derived from human feedback on probe moves between views that enhance signal quality.

19. The method of Claim 1, wherein the initial cardiac imaging view is a preliminary ultrasound view of the heart acquired prior to Al-guided probe repositioning toward the patient- specific coronary artery segment.

20. A system for guided acquisition of coronary blood flow using ultrasound imaging, comprising:an ultrasound probe configured to acquire real-time ultrasound data from a patient;a Doppler sampling cursor configured to define a sampling location for pulsed-wave Doppler (PWD) signal acquisition; anda processing unit comprising one or more processors and memory storing instructions that, when executed, cause the system to:analyze the real-time ultrasound data using a first trained model to determine a patient- specific coronary artery segment suitable for Doppler-based flow assessment, the real-time ultrasound data being acquired while the ultrasound probe is positioned to obtain an initial cardiac imaging view of the heart;generate real-time guidance instructions to guide repositioning and orientation of the ultrasound probe toward the patient- specific coronary artery segment;analyze the real-time ultrasound data at the patient- specific coronary artery segment using a second trained model to identify a patient- specific final acquisition position optimized for PWD signal acquisition;guide placement of the Doppler sampling cursor to the identified patientspecific final acquisition position; andISUS Ref: 53516-003 PCTacquire a PWD signal from the positioned Doppler sampling cursor at the patient-specific final acquisition position.

21. The system of Claim 20, wherein the processing unit is further configured to:verify that the acquired PWD signal corresponds to valid coronary artery flow and meets predefined technical quality requirements by:evaluating a clinical validity score and a technical quality score of the PWD signal;determining whether the clinical validity score and the technical quality score exceed predetermined clinical validity and technical quality thresholds adapted to patient-specific parameters; andwhen one or more thresholds are not met, initiating iterative optimization that updates probe position and / or Doppler sampling cursor placement.

22. The system of Claim 21, wherein the processing unit is further configured to apply filtering to remove clinical noise sources including artifacts caused by unrelated vascular structures, valves, or transient physiological events prior to computing the clinical validity score and the technical quality score.

23. The system of Claim 21, wherein the clinical validity score is based at least in part on coronary artery flow pattern analysis, and the technical quality score is based at least in part on signal-to-noise ratio, measurement stability, or other noise metrics.

24. The system of Claim 21, wherein the processing unit is configured to segment the PWD signal into systolic and diastolic phases and determine whether the signal exhibits a diastolic-dominant coronary flow pattern.

25. The system of Claim 20, wherein:the real-time ultrasound data comprises B-mode data and color Doppler flow data;the first trained model analyzes the B-mode data at an initial cardiac imaging view; andthe second trained model analyzes an overlay of the color Doppler flow data on the B-mode data at the patient- specific coronary artery segment.ISUS Ref: 53516-003 PCT26. The system of Claim 25, wherein the processing unit is configured to analyze the color Doppler flow data in real time to detect diastolic-dominant flow within a cardiac cycle and automatically reposition the Doppler sampling cursor to a detected region.

27. The system of Claim 20, wherein the second trained model is configured to automatically identify diastolic flow regions by detecting flow occurring predominantly during a diastolic phase of a cardiac cycle.

28. The system of Claim 20, further comprising a user interface configured to provide the real-time guidance instructions as visual guidance, audio guidance, or a combination thereof to a human operator of the ultrasound probe.

29. The system of Claim 20, further comprising a robotic or autonomous positioning mechanism configured to automatically move the ultrasound probe in accordance with the generated guidance instructions.

30. The system of Claim 20, wherein the real-time guidance instructions are predicted along six axes of probe movement.

31. The system of Claim 20, wherein the real-time guidance instructions comprise at least one of lateral translation, angular adjustment, depth adjustment, and contact pressure modification.

32. The system of Claim 20, wherein the first trained model is trained using labeled data comprising transducer spatial position and orientation data provided by a positioning sensor attached to the ultrasound probe during training with sub-millimetre accuracy and validated by expert ultrasound technicians.

33. The system of Claim 32, wherein the first trained model is trained to correlate B- mode ultrasound image features with anatomical alignment and to determine anatomical views based on the B-mode ultrasound images.

34. The system of Claim 20, wherein the Doppler sampling cursor is repositioned automatically by the processing unit or repositioned manually by an operator in response to guidance instructions.ISUS Ref: 53516-003 PCT35. The system of Claim 20, wherein guidance, verification, scoring, and operation of the first and second trained models are adapted based on patient- specific anatomical and physiological parameters.

36. The system of Claim 20, wherein at least one of the first trained model and the second trained model comprises a deep reinforcement learning model trained using a reward model derived from human feedback on probe movements between views that enhance signal quality.

37. The system of Claim 20, wherein the ultrasound probe comprises an integrated positioning sensor configured to provide real-time spatial position and orientation data used for guidance and signal validation.

38. The system of Claim 20, wherein the processing unit is further configured to process the acquired pulsed- wave Doppler (PWD) signal to extract coronary blood flow velocity features during a resting physiological state and during a stress physiological state;estimate coronary flow values corresponding to the resting physiological state and the stress physiological state based on the extracted features using a trained model; andcompute a non-invasive coronary flow reserve (CFR) value based on a ratio between the estimated coronary flow value during the stress physiological state and the estimated coronary flow value during the resting physiological state.

39. A non-transitory computer-readable medium storing instructions that, when executed by one or more processors of an ultrasound system, cause the ultrasound system to perform the method of Claim 1.

40. A computer program product comprising a non-transitory computer readable storage medium retaining program instructions, which program instructions when read by a processor, cause the processor to perform the method of Claim 1.

Images