Infrared thermography has been shown to be a useful adjunctive tool for breast cancer detection. Previous thermography
modeling techniques generally dealt with the "forward problem", i.e., to estimate the breast thermogram from known
properties of breast tissues. The present study aims to deal with the so-called "inverse problem", namely to estimate the
thermal properties of the breast tissues from the observed surface temperature distribution. By comparison, the inverse
problem is a more direct way of interpreting a breast thermogram for specific physiological and/or pathological
information. In tumor detection, for example, it is particularly important to estimate the tumor-induced thermal contrast,
even though the corresponding non-tumor thermal background usually is unknown due to the difficulty of measuring the
individual thermal properties. Inverse problem solving is technically challenging due to its ill-posed nature, which is
evident primarily by its sensitivity to imaging noise. Taking advantage of our previously developed forward-problemsolving
techniques with comprehensive thermal-elastic modeling, we examine here the feasibility of solving the inverse
problem of the breast thermography. The approach is based on a presumed spatial constraint applied to three major
thermal properties, i.e., thermal conductivity, blood perfusion, and metabolic heat generation, for each breast tissue type.
Our results indicate that the proposed inverse-problem-solving scheme can be numerically stable under imaging noise of
SNR ranging 32 ~ 40 dB, and that the proposed techniques can be effectively used to improve the estimation to the
tumor-induced thermal contrast, especially for smaller and deeper tumors.