Abstract: Systems and methods for automated recording of an ultrasound clip are based on quality scores of ultrasound images in a sequence of ultrasound image frames. An ultrasound imaging system includes a probe for capturing ultrasound images, an image buffer to store a sequence of image frames, a quality buffer to store a sequence of quality scores, and a computing subsystem that automatically records an ultrasound clip when the quality scores in the quality buffer corresponding to a set of contiguous image frames in the image buffer equal or exceed a first quality threshold and the set of contiguous image frames is at least a first predetermined size. Additionally, a smart capture feature automatically records an ultrasound clip including an alternate set of contiguous image frames having quality scores equaling or exceeding a second quality threshold that is less than the first quality threshold and meets a second predetermined size.

Abstract: A diagnostic facility is described. The facility accesses a set of trained machine learning models. For each of a plurality of stages of a diagnostic ultrasound protocol for blood vessels, the facility causes an ultrasound device to capture from the person an ultrasound artifact of a type specified for the stage that features a blood vessel specified for the stage; applies one of the trained machine learning models to the captured ultrasound artifact to produce a prediction; and determines a score for the stage based at least in part on the produced prediction. The facility combines the determined scores to produce a diagnosis grade for the person.

Abstract: A machine learning model is described that is trained without labels to predict a motion field between a pair of images. The trained model can be applied to a distinguished pair of images to predict a motion field between the distinguished pair of images.

Abstract: A machine learning model is described that usable to improve the quality of an ultrasound image captured from a person using a set of ultrasound machine setting values that is collectively sub-optimal. Different versions of the model predict from such a starting ultrasound image either (a) a new set of setting values that can be used to reimage the person to produce a higher-quality ultrasound image, or (b) this higher-quality ultrasound image directly.

Abstract: A facility for automatically establishing measurement location controls for Doppler ultrasound studies is described. The facility receives a first ultrasound image, and user input selecting an anatomical structure appearing in it. The facility performs localization to determine the location of the selected anatomical structure in the initial image, and determines a first placement of measurement location controls relative to the structure. On the ultrasound machine, the facility invokes one or more first Doppler ultrasound modes using the first placement. The facility receives a second ultrasound image produced by the ultrasound machine using the one or more first modes; determines a flow location and direction based on the second ultrasound image; and determines a second placement relative to the flow location. The facility invokes one or more second Doppler ultrasound modes using the second placement, and receives results from the invocation of the one or more second Doppler ultrasound modes.

Abstract: A facility for detecting a target structure is described. The facility receives an ultrasound image. It subjects the ultrasound image to a detection model to obtain, for each of one or more occurrences of a target structure appearing in the ultrasound image, a set of parameter values fitting a distinguished shape to the target structure occurrence. The facility stores the obtained one or more parameter value sets in connection with the ultrasound image.

Abstract: A facility for assessing an ultrasound image captured from a patient with a particular depth setting is described. The facility subjects the received ultrasound image to at least one neural network to produce, for each neural network, an inference. On the basis of the produced inferences, the facility determines whether the depth setting at which the ultrasound image was captured was optimal.

Abstract: A facility identifies anatomical objects visualized by a medical imaging image. The facility applies two machine learning models to the image: a first trained to predict a view probability vector that, for each of a list of views, attributes a probability that the image was captured from the view, and a second trained to predict an object probability vector that, for each of a list of anatomical objects, attributes a probability that the object is visualized by the image. For each object, the facility: (1) accesses a list of views in which the object is permitted; (2) multiplies the predicted probability that the object is visualized by the image by the sum of the predicted probabilities that the accessed image was captured from views in which the object is permitted; and (3) where the resulting probability exceeds a threshold, determines that the object is visualized by the accessed image.

Abstract: A facility for processing a medical imaging image is described. The facility applies each of a number of constituent models making up an ensemble machine learning models to the image to produce a constituent model result that predicts a value for each pixel of the image. The facility aggregates the results produced by the constituent models of the plurality to determine a result of the ensemble machine learning model. For each of the pixels of the accessed image, the facility determines a measure of variation among the values predicted for the pixel among the constituent models. Facility determines a confidence measure for the ensemble machine learning model result based at least in part on for how many of the pixels of the accessed image a variation measure is determined that exceeds a variation threshold.

Abstract: A facility identifies anatomical objects visualized by a medical imaging image. The facility applies two machine learning models to the image: a first trained to predict a view probability vector that, for each of a list of views, attributes a probability that the image was captured from the view, and a second trained to predict an object probability vector that, for each of a list of anatomical objects, attributes a probability that the object is visualized by the image. For each object, the facility: (1) accesses a list of views in which the object is permitted; (2) multiplies the predicted probability that the object is visualized by the image by the sum of the predicted probabilities that the accessed image was captured from views in which the object is permitted; and (3) where the resulting probability exceeds a threshold, determines that the object is visualized by the accessed image.

Abstract: A facility for processing a medical imaging image is described. The facility applies each of a number of constituent models making up an ensemble machine learning models to the image to produce a constituent model result that predicts a value for each pixel of the image. The facility aggregates the results produced by the constituent models of the plurality to determine a result of the ensemble machine learning model. For each of the pixels of the accessed image, the facility determines a measure of variation among the values predicted for the pixel among the constituent models. Facility determines a confidence measure for the ensemble machine learning model result based at least in part on for how many of the pixels of the accessed image a variation measure is determined that exceeds a variation threshold.

Abstract: A machine learning model is described that is trained without labels to predict a motion field between a pair of images. The trained model can be applied to a distinguished pair of images to predict a motion field between the distinguished pair of images.

Abstract: Systems and methods for automated physiological parameter estimation from ultrasound image sequences are provided. An ultrasound system includes an ultrasound imaging device configured to acquire a sequence of ultrasound images of a patient. An anatomical structure recognition module includes processing circuitry configured to receive the acquired sequence of ultrasound images from the ultrasound imaging device, and automatically recognize an anatomical structure in the received sequence of ultrasound images. A physiological parameters estimation module includes processing circuitry configured to automatically estimate one or more physiological parameters associated with the recognized anatomical structure.

Abstract: Automated ultrasound image labeling and quality grading systems and methods are provided. An ultrasound system includes an ultrasound imaging device configured to acquire ultrasound images of a patient. An anatomical structure recognition and labeling module receives the acquired ultrasound images from the ultrasound imaging device, and automatically recognizes anatomical structures in the received ultrasound images. The anatomical structure recognition and labeling module automatically labels the anatomical structures in the images with information that identifies the anatomical structures. The acquired ultrasound images and the labeled anatomical structures are displayed on a display of the ultrasound imaging device.

Abstract: A method receives a captured image depicting image content including an object, the captured image being captured by an image sensor located at a sensor position; generates, using a trained first machine learning logic, a lighting-corrected image from an imitative simulation image depicting at least a portion of the image content of the captured image in a simulation style associated with an environment simulator; generates, using a trained second machine learning logic, a depth estimation image from the lighting-corrected image, the depth estimation image indicating a relative distance between the object depicted in the captured image and the sensor position of the image sensor; and determines an object position of the object depicted in the captured image based on the depth estimation image.

Abstract: A method receives a captured image depicting image content including an object, the captured image being captured by an image sensor located at a sensor position; generates, using a trained first machine learning logic, a lighting-corrected image from an imitative simulation image depicting at least a portion of the image content of the captured image in a simulation style associated with an environment simulator; generates, using a trained second machine learning logic, a depth estimation image from the lighting-corrected image, the depth estimation image indicating a relative distance between the object depicted in the captured image and the sensor position of the image sensor; and determines an object position of the object depicted in the captured image based on the depth estimation image.