Identifying blood vessels in ultrasound images

A method implemented by a computing device for blood vessel identification, the method comprising:

• assigning, with a neural network implemented at least partially in hardware of the computing device, one of a vein classification and an artery classification to one or more blood vessels in ultrasound images;
• determining a misclassification for one blood vessel of the one or more blood vessels, the misclassification denoting the neural network assigning the one of the vein classification and the artery classification to the one blood vessel in one ultrasound image of the ultrasound images and the other of the vein classification and the artery classification to the one blood vessel in additional ultrasound images of the ultrasound images;
• displaying, in the one ultrasound image, an indication of the other of the vein classification and the artery classification for the one blood vessel;
• receiving a user selection that indicates a time interval;
• wherein the one ultrasound image and the additional ultrasound images represent frames in an ultrasound video and the determining the misclassification is based on a number of the frames corresponding to the time interval.

A method implemented by a computing device for blood vessel identification, the method comprising:

• assigning, with a neural network implemented at least partially in hardware of the computing device, one of a vein classification and an artery classification to one or more blood vessels in ultrasound images;
• determining a misclassification for one blood vessel of the one or more blood vessels, the misclassification denoting the neural network assigning the one of the vein classification and the artery classification to the one blood vessel in one ultrasound image of the ultrasound images and the other of the vein classification and the artery classification to the one blood vessel in additional ultrasound images of the ultrasound images;
• and
• displaying, in the one ultrasound image, an indication of the other of the vein classification and the artery classification for the one blood vessel;
• determining, with the neural network, diameters of the one or more blood vessels in the ultrasound images, wherein the displaying includes displaying the indication for the one blood vessel in a color corresponding to a catheter size for the one blood vessel based on a diameter determined with the neural network for the one blood vessel.

A method implemented by a computing device for blood vessel identification, the method comprising:

• assigning, with a neural network implemented at least partially in hardware of the computing device, one of a vein classification and an artery classification to one or more blood vessels in ultrasound images;
• determining a misclassification for one blood vessel of the one or more blood vessels, the misclassification denoting the neural network assigning the one of the vein classification and the artery classification to the one blood vessel in one ultrasound image of the ultrasound images and the other of the vein classification and the artery classification to the one blood vessel in additional ultrasound images of the ultrasound images;
• displaying, in the one ultrasound image, an indication of the other of the vein classification and the artery classification for the one blood vessel;
• and adjusting at least one neural network coefficient based on the misclassification.

An ultrasound system for identifying blood vessels, the ultrasound system comprising:

• an image module to generate ultrasound images based on ultrasound echo signals;
• a neural network module to identify the blood vessels in a first image of the ultrasound images;
• a map module to generate a map image that indicates locations of the blood vessels in the first image identified by the neural network module;
• and
• the neural network module to identify the blood vessels in a second image of the ultrasound images based on the map image and the second image, wherein the neural network module is configured to adjust at least one neural network coefficient based on a loss function of the locations of the blood vessels in the first image and additional locations of the blood vessels in the second image.

A method implemented by a computing device for blood vessel identification, the method comprising:

• determining, with a neural network implemented at least partially in hardware of the computing device, bounding containers of blood vessels in ultrasound images;
• determining compression amounts of the blood vessels based on the bounding containers;
• assigning, with the neural network, classifications of a vein classification or an artery classification to the blood vessels in the ultrasound images;
• and
• determining labels for the blood vessels based on the compression amounts and the classifications, the labels indicating the blood vessels as a vein or an artery, wherein the determining the compression amounts of the blood vessels includes:
• determining first lengths of the bounding containers;
• determining second lengths of the bounding containers;
• and
• determining ratios of the first lengths to the second lengths.