Segmentation of breast cancer masses in ultrasound using radio-frequency signal derived parameters and strain estimates

In order to address the tendency of ultrasound B-Mode images to show a too small sizing of tumour masses in breast cancer diagnosis, a novel segmentation method has been introduced. In this paper it has been explored if this problem can be solved by incorporating strain parameters from ultrasound elastography into a segmentation framework. By incorporating a local power estimate from an autoregressive model on the RF-data with strain parameters into a fast algorithm using Graph-cuts, very difficult to interpret cancer data could be segmented successfully with more accurate sizing than with B-Mode data alone. However, more research is necessary to correlate histology findings with elasticity parameters to find the best model to interpret strain data for breast mass segmentation.