A shape-based approach of image reconstruction under continues-wave mode is developed for diffuse optical tomography (DOT), which aims to simultaneously recover the smooth region boundaries and optical parameters of the biological tissue. The method is based on the spherical harmonics parameterization methodology and an assumption that different anatomical regions have their respective sets of the homogeneous optical parameter distributions. The boundary element method (BEM) is used for forward modeling, and a Levenberg-Marquardt optimization method is implemented for the inverse problem. The proposed scheme is validated using a domain with two heterogeneous inclusions, the shape parameters and the optical coefficients of the domains can be simultaneously reconstructing at different noise levels. And physical experiment on a phantom is also conducted to evaluate the performance of our method. The reconstructed results show that the methodology is very promising and of good convergence. The homogeneous optical parameters and shape parameters of each region can be reconstructed with good accuracy.