An improved algorithm for UWB based imaging of breast tumors

In this paper we propose a novel approach for breast tumor diagnosis (detection and volume estimation) using ultrawideband (UWB) radio frequency (RF) sensors in an application specific instrumentation (ASIN) framework, which is also proposed in this paper. In this, the application specific analysis is performed without giving any intermediate output. For the case of UWB based breast tumour diagnosis we try to perform the diagnosis without the need of constructing an radio-frequency image of the breast tissue. By avoiding the intermediate step of image construction we avoid many complexities, viz impractical assumptions of RF imaging, implementation of the complex imaging algorithms and requirement for a large number of transmitter-receiver positions. We have validated the conceptual framework by applying it on extensive amount of FDTD simulated database. Decision making regarding tumor presence and tumor size is done using neural network based machine learning algorithms. We validate the scheme for both detection of the tumor and for the estimation of its size. For a rigorous analysis we have also validated the scheme for the cases involving the presence of single or multiple false-alarms, i.e. anomalies with difference in fat tissues dielectric constant which is not as high as for the case of an actual tumor. Next, the proposed scheme is tested to check whether the total volume of tumor can be estimated when the total volume consists of volumes of multiple tumors.