Contrast and B-mode combination for strain analysis in medical ultrasound imaging
For boundary detection, such as heart wall detection for strain or another quantification, the boundary is detected using multiple types of imaging, such as B-mode and contrast. The confidence in the detected locations is used to form a more accurate boundary, such as replacing low confidence locations from B-mode imaging with higher confidence locations from the contrast imaging. Even where experts may struggle to identify boundaries, a more accurate boundary is created and used by the image processor or processing.
1 . A method for quantification of heart wall performance with a medical ultrasound system, the method comprising:
detecting, by an image processor, heart wall locations for each B-mode image in a sequence of B-mode images of a patient;
determining, by the image processor, first confidence measures for the heart wall locations from the sequence of the B-mode images;
detecting, by the image processor, the heart wall locations for each contrast image in a sequence of contrast images of the patient;
determining, by the image processor, second confidence measures in the heat wall locations from the sequence of the contrast images;
generating, by the image processor, a heart wall boundary for each of different times from the heart wall locations from the sequence of B-mode images, the heart wall locations from the sequence of contrast images, the first confidence measures, and the second confidence measures;
calculating, by the image processor, a strain from the heart wall boundary for multiple of the different times; and
displaying the strain.
2 . The method of claim 1 , wherein detecting the heart wall locations in the sequence of B-mode images comprises detecting endo-cardium boundary and epi-cardium boundary locations by a first machine-learned model; and wherein detecting the heart wall locations in the sequence of contrast images comprises detecting the endo-cardium boundary locations by a second machine-learned model.
3 . The method of claim 2 , wherein determining the first confidence measures comprises receiving the first confidence measures from the first machine-learned model; and
wherein determining the second confidence measures comprises receiving the second confidence measures from the second machine-learned model.
4 . The method of claim 1 , wherein detecting the heart wall locations in the sequence of B-mode images comprises detecting the heart wall locations in an end-systole image of the B-mode images and an end-diastole image of the B-model images and tracking the heart wall locations for images of the B-mode images between the end-systole and end-diastole images; and
wherein detecting the heart wall locations in the sequence of contrast images comprises detecting the heart wall locations in an end-systole image of the contrast images and an end-diastole image of the contrast images and tracking the heart wall locations for images of the contrast images between the end-systole and end-diastole images.
5 . The method of claim 1 , wherein determining the first and second confidence measures comprise determining the first and second confidence measures as measures of a motion trajectory for each of the heart wall locations.
6 . The method of claim 5 , wherein determining the first and second confidence measures comprises determining with the measures of the motion trajectory comprising fit to a model of motion, local continuity, and/or global continuity.
7 . The method of claim 1 , further comprising aligning the B-mode and contrast images based on the heart wall locations for the B-mode images, the heart wall locations for the contrast images, the first confidence measures, and the second confidence measures; and then performing the generating.
8 . The method of claim 1 , wherein generating comprises, for each heart wall location from the B-mode images, testing the first confidence measure to a threshold, and when the first confidence measure is below the threshold, replacing the heart wall location from the B-mode images with the heart wall location from the contrast images where the second confidence measure is above the threshold.
9 . The method of claim 1 , further comprising:
fusing tissue information from the B-mode images with flow information from the contrast images into one or more heart images, the fusing using one or more masks based on the heart wall boundary; and
displaying the one or more heart images with the strain.
10 . The method of claim 1 , wherein calculating the strain comprise calculating a local or global strain from a change in the heart wall boundary between the different times.