Ultrasonic viscoelasticity imaging of nonpalpable breast lesions

The prognosis of breast cancer patients improves with early and accurate diagnosis. The quality of the strain image produced by displacement estimation algorithms are heavily influenced by the RF echo data recorded by different ultrasound systems. A small clinical study was conducted with 21 women having a single nonpalpable breast lesion, each detected mammographically with later pathology confirmation. Elasticity images were acquired on each patient to test for the ability to differentiating malignant and benign lesions. The mechanical relaxation time T1 images showed a tissue-specific T1 contrast that is negative for all 11 malignant lesions and positive for all 10 benign lesions. Strain images were estimated using a regularized multi-scale optical flow (ROF) algorithm. Adjustments to the input parameters to the ROF and their subsequent effects on T1 estimation and computation time are shown to have a strong effect on diagnostic performance