Breast tumour identification based on inverse scattering approach

This study introduces an iterative evolutionary method in detection of breast tumour, identification of its size and location. At primary measurement step, a set of antennas capable of transmitting wide-band Gaussian pulses, generated through the chirping technique, is placed around the breast and total wave received by each directional antenna is recorded in a bistatic manner for every cross section of the breast corresponding to a specific height for a full three-dimensional scan. Each set of data for each cross section is then analysed separately. Measured fields are used in an inverse-scattering problem, where the unknown is the tumour. To solve this problem, the finite-difference time-domain technique produces the total field resulted from the same incident wave as in measurement while illuminating an electromagnetic model of the breast, each time having a new tumour specification as a part of an evolutionary algorithm scheme. A comparison between various generated and measured reference data is repeated to reach a minimum cost function that corresponds to a minimum error in identification. Particle swarm optimisation and differential evolution are the two optimisations, each using all field elements of both TE_z and TM_z modes to reach convergence. The final results show an accurate identification of the tumour by this method.