Breast Thermography is one of the scanning techniques used for breast cancer detection. Looking at breast thermal image it is difficult to interpret parameters of tumor such as depth, size and location which are useful for diagnosis and treatment of breast cancer. In our previous work (ITBIC) we proposed a framework for estimation of tumor size using clever algorithms and the radiative heat transfer model. In this paper, we expand it to incorporate the more realistic Pennes bio-heat transfer model and Markov Chain Monte Carlo (MCMC) method, and analyze it's performance in terms of computational speed, accuracy, robustness against noisy inputs, ability to make use of prior information and ability to estimate multiple parameters simultaneously. We discuss the influence of various parameters used in its implementation. We apply this method on clinical data and extract reliable results for the first time using breast thermography. ©2009 IEEE.