Estimation of nonlinear scattering for improved detection of micro calcifications for diagnosis of breast cancer using dual band ultrasound

Abstract

Breast cancer is the second largest cause of cancer death in women worldwide. In diagnosis of breast cancer the shape and distribution of micro calcifications in the breast is an important indicator. Today, breast cancer screening is primarily performed using X-ray mammography. The quality of detection in mammography is however reduced in dense breasts. Ultrasound has shown to be able to detect cancers in dense breasts, and can be used as a supplement to mammography in screening of breasts with high density. Nevertheless, todays ultrasound technology has shown insufficient in detection of micro calcifications.

This thesis presents methods for detection of micro calcifications in the breast utilizing a dual band ultrasound technique, SURF imaging. A SURF pulse complex consists of a low frequency (LF) and a high frequency (HF) pulse with commonly a separation in frequency of 1:10. The LF pulse manipulates the nonlinear elasticity of the tissue as seen by the HF pulse. Detection can be achieved by combining between three and five measurements, where the LF pulse amplitude is varied, in a suppression method called, delay corrected subtraction (DCS).

Simulations have found micro calcifications to obtain a nonlinear response dependent on the LF pulse amplitude, which is more prominent than for scattering from common interfaces in the breast, such as between fat and tissue.In vitro measurements on a commercial breast phantom containing micro calcifications have provided detection coherent with results from X-ray mammography. The particles have been detected with a signal-to-tissue ratio of on average 20 to 30 dB. Increased detection has been found when varying the placing of the HF pulse over a top of the LF pulse, and when positioning the HF pulse to detect a possible ringing effect on the particles from the LF pulse. Further contrast resolution of the particles is obtained by applying a segmentation method based on combining the B-mode image and the DCS suppressed image.