Systems and methods for clutter artifact removal using a data-driven model trained on images formed using a lower clutter image sequence and a higher clutter image sequence generated by an artifact overlay on the lower clutter image sequence

An image processing system comprising:

• a processor, and a non-transitory memory including instructions that when executed cause the processor to:
• input a sequence of ultrasound images including an amount of clutter artifacts into a trained data-driven model, the data-driven model trained on a plurality of training pairs of image sequences, each training pair including a first, ground truth ultrasound image sequence with a lower amount of clutter artifacts acquired from a subject during an examination, and a second ultrasound image sequence with a higher amount of clutter artifacts, the second ultrasound image sequence generated by superimposing the clutter artifacts on the first, ground truth ultrasound image sequence;
• and
• display an artifact-reduced version of the sequence of ultrasound images on a display device of the image processing system and/or storing the artifact-reduced version in a memory of the image processing system, the artifact-reduced version outputted by the data-driven model;
• wherein the superimposed clutter artifacts of the second ultrasound image sequence further comprise a plurality of haze patches positioned to cover a respective ultrasound image of the first, ground truth ultrasound image sequence while minimizing a space between each haze patch, the haze patches extracted from cavities in ultrasound images of a third ultrasound image sequence including a higher amount of clutter artifacts.