Homotopy-based Image Reconstruction in Breast Ultrasound Tomography

摘要

To diagnose breast cancer, an alternative approach to X-ray mammography and B-mode ultrasound is the Breast Ultrasound Computed Tomography (UCT). Concerning breast cancer detection, mammographic techniques have been considered as gold standard modality but those methods suffer through high rate of false-positive results and radiations. In addition, the use of low resolution images in the conventional ultrasound techniques is another problem that makes it difficult to differentiate between breast tissues and tumor existence. In this particular study, we have used a full wave-based tomographic technique to diagnose the breast cancer. While taking into account the physical characteristics, full wave-based approach constructs non-linear operator that transforms it as a nonlinear inverse problem or a minimization problem. To solve the non-linear inverse problem, we have used Homotopoy Continuation Methods (HCM), which is globally convergent method and provides multiple solutions. In the subsequent steps, to reduce the computational complexity and the high usage of memory for finding out the Jacobian of the cost function, we have applied discrete adjoint method. In the last, the effect of the noise to the reconstructed image has been tested. In this preliminary analysis with the use of adjoint methods, obtained results have shown the robustness and visibility of our proposed approach.