# US 12,109,070 B2 · Using machine learning techniques to obtain coherence functions > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,109,070 B2 - Document ID: US12109070B2 - Status: Granted - Kind: B2 - Application number: 17/763,069 - Filed: 2020-09-28 - Published: 2022-10-27 - Granted: 2024-10-08 - Earliest claimed: 2019-09-27 - Assignees: The Johns Hopkins University - Applicants: The Johns Hopkins University - Inventors: Muyinatu Bell; Alycen Wiacek - Web page: https://patentgrep.com/patent/US12109070B2 ## CPC - A61B 8/08: Clinical applications (A61B8/02, A61B8/04, A61B8/06 take precedence) - A61B 8/5207: {involving processing of raw data to produce diagnostic data, e.g. for generating an image} - A61B 5/0095: {by applying light and detecting acoustic waves, i.e. photoacoustic measurements} - A61B 8/00: Diagnosis using ultrasonic, sonic or infrasonic waves - A61B 8/06: Measuring blood flow - A61B 8/085: {for locating body or organic structures, e.g. tumours, calculi, blood vessels, nodules} - A61B 8/485: {involving measuring strain or elastic properties} - A61B 8/5223: {for extracting a diagnostic or physiological parameter from medical diagnostic data (for algorithms to analyse biomedical images G06T7/0012)} - A61B 8/5269: {involving detection or reduction of artifacts} - G01S 7/52026: {Extracting wanted echo signals (Doppler systems G01S15/50; Doppler short range imaging systems G01S15/8979)} - G01S 7/52046: {Techniques for image enhancement involving transmitter or receiver (image enhancement by image data processing G06T5/00)} - G01S 15/8915: {using a transducer array} - G06F 30/27: using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model - G06N 3/0499: Feedforward networks - G06N 3/08: Learning methods - G06N 3/09: Supervised learning - G10K 11/34: using electrical steering of transducer arrays, e.g. beam steering {(constructional aspects B06B1/0607, B06B1/085)} - A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D) - G01S: RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G10K: SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR ## Abstract A computer-implemented method for training and using a neural network to predict a coherence function includes: training a neural network by mapping a plurality of different sets of training input samples to respective coherence function truths to generate a trained neural network; receiving an operational input sample; inputting the operational input sample into the trained neural network; obtaining, from the trained neural network, a coherence function mapped to the operational input sample in response to the inputting the operational input sample into the trained neural network; and executing a computer-based instruction based on obtaining the coherence function. The coherence function may be used to differentiate solid masses from fluid-filled masses. ## Claims ### Claim 1 A computer-implemented method comprising: training a neural network by mapping a plurality of different sets of training input samples to respective spatial coherence function truths to generate a trained neural network; receiving an operational input sample; inputting the operational input sample into the trained neural network; obtaining, from the trained neural network, a spatial coherence function mapped to the operational input sample in response to the inputting the operational input sample into the trained neural network, wherein the spatial coherence function is usable to form an image according to Rsl=∫1M{circumflex over (R)}(m)dm, where Rsl represents an image pixel value, M represents a short-lag value, and {circumflex over (R)}(m) represents the spatial coherence function; and executing a computer-based instruction based on obtaining the spatial coherence function. ### Claim 2 (depends on 1) The method of claim 1, wherein the operational input sample data includes at least one selected from the group consisting of: axial kernel data from ultrasound or ultrasound-derived channel data; samples corresponding to a single axial position in space received by multiple ultrasound system channels; an entire recorded channel data set; and all axial positions for a lateral line, or multiple recordings over a period of time. ### Claim 3 (depends on 1) The method of claim 1, wherein the inputting the operational input sample data into the trained neural network comprises transforming the operational input sample data into input axial kernel data. ### Claim 4 (depends on 3) The method of claim 3, wherein the input axial kernel data is associated with ultrasound or photoacoustic or ultrasound-derived channel data. ### Claim 5 (depends on 1) The method of claim 1, wherein executing the computer-based instruction includes at least one selected from the group consisting of: an instruction to store or output the spatial coherence function; an instruction to execute an application using the spatial coherence function, wherein the application includes at least one selected from the group consisting of: motion detection from ultrasound data, motion detection from photoacoustic data, blood-flow estimation, sound speed correction, speckle tracking, elastography, minimum variance beamforming, and coherence-weighted imaging, or advanced correlation-based beamforming techniques; and an instruction to refine the trained neural network based on the output of the application that uses the spatial coherence function. ### Claim 6 (depends on 1) The method of claim 1, wherein the training the neural network further comprises mapping respective short-lag spatial coherence (SLSC) image truths to the plurality of different sets of training input samples, the method further comprising obtaining an SLSC image mapped to the operational input sample in response to the inputting the operational input sample into the trained neural network. ### Claim 7 (depends on 1) The method of claim 1, further comprising refining the trained neural network based on output spatial coherence functions predicted by the neural network and measured spatial coherence functions. ### Claim 8 A non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computing device to cause the computing device to perform operations comprising: receiving an operational input sample; inputting the operational input sample into a trained neural network; obtaining, from the trained neural network, a spatial coherence function mapped to the operational input sample in response to the inputting the operational input sample into the trained neural network, wherein the spatial coherence function is usable to form an image according to Rsl=∫1M{circumflex over (R)}(m)dm, where Rsl represents an image pixel value, M represents a short-lag value, and {circumflex over (R)}(m) represents the spatial coherence function; and executing a computer-based instruction based on obtaining the spatial coherence function. ### Claim 9 (depends on 8) The non-transitory computer readable storage medium of claim 8, wherein the operational input sample data includes at least one selected from the group consisting of: axial kernel data from ultrasound or ultrasound-derived channel data; samples corresponding to a single axial position in space received by multiple ultrasound system channels; an entire recorded channel data set; and all axial positions for a lateral line, or multiple recordings over a period of time. ### Claim 10 (depends on 8) The non-transitory computer readable storage medium of claim 8, wherein inputting the operational input sample data into the trained neural network comprised transforming the operational input sample data into input axial kernel data. ### Claim 11 (depends on 10) The non-transitory computer readable storage medium of claim 10, wherein the input axial kernel data is associated with ultrasound or photoacoustic or ultrasound-derived channel data. ### Claim 12 (depends on 8) The non-transitory computer readable storage medium of claim 8, wherein the executing the computer-based instruction includes at least one selected from the group consisting of: an instruction to store or output the spatial coherence function; an instruction to execute an application using the spatial coherence function, wherein the application includes at least one selected from the group consisting of: motion detection from ultrasound data, motion detection from photoacoustic data, blood-flow estimation, sound speed correction, speckle tracking, elastography, minimum variance beamforming, and coherence-weighted imaging, or advanced correlation-based beamforming techniques; and an instruction to refine the trained neural network based on the output of the application that uses the spatial coherence function. ### Claim 13 (depends on 8) The non-transitory computer readable storage medium of claim 8, wherein the operations further comprise at least one selected from the group consisting of: training the neural network by mapping a plurality of different sets of training input samples to respective spatial coherence function truths or mapping respective short-lag spatial coherence (SLSC) image truths to the plurality of different sets of training input samples; obtaining an SLSC image mapped to the operational input sample in response to the inputting the operational input sample into the trained neural network; refining the trained neural network based on output spatial coherence functions predicted by the trained neural network and measured spatial coherence functions. ### Claim 14 A system comprising: a processor, a computer readable memory, a non-transitory computer readable storage medium associated with a computing device, and program instructions executable by the computing device to cause the computing device to perform operations comprising: receiving an operational input sample; inputting the operational input sample into a trained neural network that maps a plurality of different sets of input samples to one or more spatial coherence functions; obtaining, from the trained neural network, a spatial coherence function mapped to the operational input sample in response to the inputting the operational input sample into the trained neural network, wherein the spatial coherence function is usable to form an image according to Rsl=∫1M{circumflex over (R)}(m)dm, where Rsl represents an image pixel value, M represents a short-lag value, and {circumflex over (R)}(m) represents the spatial coherence function; and executing a computer-based instruction based on obtaining the spatial coherence function. ### Claim 15 (depends on 14) The system of claim 14, wherein the operational input sample data includes at least one selected from the group consisting of: axial kernel data from ultrasound or ultrasound-derived channel data; samples corresponding to a single axial position in space received by multiple ultrasound system channels; an entire recorded channel data set; and all axial positions for a lateral line, or multiple recordings over a period of time. ### Claim 16 (depends on 14) The system of claim 14, wherein the inputting the operational input sample data into the trained neural network comprised transforming the operational input sample data into input axial kernel data. ### Claim 17 (depends on 16) The system of claim 16, wherein the input axial kernel data is associated with ultrasound or ultrasound-derived channel data. ### Claim 18 (depends on 14) The system of claim 14, wherein the executing the computer-based instruction includes at least one selected from the group consisting of an instruction to store or output the spatial coherence function; an instruction to execute an application using the spatial coherence function, wherein the application includes at least one selected from the group consisting of: motion detection from ultrasound data, motion detection from photoacoustic data, blood-flow estimation, sound speed correction, speckle tracking, elastography, minimum variance beamforming, and coherence-weighted imaging, or advanced correlation-based beamforming techniques; and an instruction to refine the trained neural network based on the output of the application that uses the spatial coherence function. ### Claim 19 (depends on 14) The system of claim 14, wherein the operations further comprise at least one selected from the group consisting of: training the neural network by mapping a plurality of different sets of training input samples to respective spatial coherence function truths; training the neural network by mapping the plurality of different sets of training input samples to the respective short-lag spatial coherence (SLSC) image truths; obtaining an SLSC image mapped to the operational input sample in response to the inputting the operational input sample into the trained neural network; refining the trained neural network based on output spatial coherence functions predicted by the neural network and measured spatial coherence functions. ### Claim 20 (depends on 14) The system of claim 14, wherein the operations further comprise determining the content of the source of the operational input sample.