Image processing and registration of opposed view 3d breast ultrasound

Authors:
Sumedha Sinha;Fong-Ming Hooi;Renee Pinsky;Oliver Kripfgans;Paul Carson
Affiliations:
Department of Radiology, University of Michigan, Ann Arbor, USA,Department of Biomedical Engineering, University of Michigan, Ann Arbor;Department of Radiology, University of Michigan, Ann Arbor, USA,Department of Biomedical Engineering, University of Michigan, Ann Arbor;Department of Radiology, University of Michigan, Ann Arbor;Department of Radiology, University of Michigan, Ann Arbor;Department of Radiology, University of Michigan, Ann Arbor, USA,Department of Biomedical Engineering, University of Michigan, Ann Arbor
Venue:
IWDM'12 Proceedings of the 11th international conference on Breast Imaging
Year:
2012

Citing 2
Cited 0

The nature of statistical learning theory

The nature of statistical learning theory
Segmentation of ultrasonic images using support vector machines

Pattern Recognition Letters - Speciqal issue: Ultrasonic image processing and analysis

Quantified Score

Hi-index	0.00

Visualization

Abstract

We are studying opposed view ultrasonic imaging (OVI) of the breast in the mammographic geometry, with probable future automation and alignment with X-ray tomosynthesis. OVI through a filament mesh paddle results in improved spatial resolution, contrast, and signal-to-noise ratio. We expect these images will be of a quality that justifies their use for screening purposes, especially for subjects with dense breasts. A previous study assessed machine learning for isolating image artifacts, which included posterior acoustic shadowing from cancers and enhancement arising from cysts. The image volumes were acquired on a custom breast-mimicking phantom containing multiple cysts and solid masses. This paper reports that 3D non-linear registration of opposed view image volumes was robust for the segmented image volumes with noisy areas excluded.