Early detection of breast cancer using genetic algorithm and Neuro-Fuzzy network

In this paper, microwave tomography method has been applied for early breast cancer detection. For this purpose, optimization technique has been used for reconstruction of 2-D images of dielectric properties and finally, early detection of the tumor location. The finite difference time domain method has been used for calculating the scattered Electromagnetic fields in the forward solver. The stochastic algorithm used in the inverse solver is Genetic Algorithm. To speed up the algorithm convergence, the dielectric parameters values have been divided into four intervals and the representative value for each of these intervals has been selected to be the average value of the interval. This work incorporates a new cost function to ameliorate the deficiencies existing in the other works. Simulated results demonstrate that the proposed approach is superior to its rivals in terms of both the convergence speed and the accuracy of the solutions. Afterward, in order to eliminate some potential problems due to the averaging of the dielectric constant intervals, the dominant soft and hard tissues of the image are obtained through a Neuro-Fuzzy model. Then, these equivalent values are used in optimization process. This method, not only yields the correct result in all cases, but increases the convergence speed compared to the previous cases as well.