Detecting breast cancer using microwave imaging and stochastic optimization

Breast cancer detection is one of the most important problems in health care as it is second most frequent cancer according to WHO. Breast cancer is among cancers which are most probably curable, only if it is diagnosed at early stages. To this purpose it has been recently proposed that microwave imaging could be used as a cheaper and safer alternative to the commonly used combination of mammography. From a physical standpoint breast cancer can be modelled as a scatterer with a significantly (tenfold) larger conductivity than a healthy tissue. In our previous work we proposed a maximum likelihood based method for detection of cancer which estimates the unknown parameters by minimizing the residual error vector assuming that the error can be modelled as a multivariate (multiple antennas) random variable. In this paper we utilize stochastic optimization technique and evaluate its applicability to the detection of cancer using numerical models. Although these models have significant limitations they are potentially useful as they provide insight in required levels of noise in order to achieve desirable detection rates.