Image Reconstruction Based on Compressive Sensing Using Total Variation Spatial Regulation for Microwave Imaging

Medical Imaging is an essential tool in supporting medical diagnosis as well as early detection of a number of diseases, e.g., tumor and cancer. Computed Tomography (CT) scan and Magnetic Resonance Imaging (MRI) are the modalities that are widely used for these purposes. Recently, Microwave Imaging has emerged as a modality that promotes tomography techniques with lower economical cost and smaller size than the previous technologies. For the sake of the image quality, however, this imaging system requires a large amount of data measurements in the reconstruction process. To overcome the drawback, this research proposes an algorithm to reconstruct the microwave images with lower number of measurements using Compressive Sensing (CS) approach. CS enables reconstructing a signal from a smaller number of measurements than which is required in the conventional sampling method. To meet this framework, in our proposed formulation, the acquisition scheme of scanning process is represented by a projection matrix for which a weight matrix of Discrete Radon Transform is used. In addition, in the data acquisition process, a number of translation and rotation positions are provided and varied in combinations to confirm the fewer measurement concept of CS. As the basic sparse reconstruction had been successfully proven for this task, this research contributes by adding spatial information using total variation (TV) and solving the proposed optimization problem using Alternating Direction Method of Multipliers (ADMM). As for the sparse dictionary matrix, Discrete Cosine Transform (DCT) is selected. The experiment shows that the proposed algorithm successfully outperforms the reconstruction which applied Filtered Back Projection (FBP), Simultaneous Algebraic Reconstruction Technique (SART) algorithm, and Basis Pursuit (BP) in terms of image quality and quantitative parameters.