Novel Approaches To Image Reconstruction

Several novel approaches to image reconstruction have been incorporated into NDRM. One of the most important is the use of Total Variation regularization, a technique that allows reconstruction of much sharper image profiles than traditional algorithms. This improves the ability to distinguish the interface between two organs. New methods have been developed involving the use of the 1-Norm on the data term of the image reconstruction, This approach results in a strong robustness to the presence of image outliers in the data.


Above, the image on the left shows an MRI image of a volunteer patient with breast cancer, indicated by a brighter spot in the lower part of the breast. The center image shows a 2D section of a 3D electrical impedance tomography image, where the red color indicates the presence of the tumor. This image is reconstructed with a traditional algorithm, which tends to soften the boundaries of the tumor. On the right, the same image was reconstructed with the algorithm proposed, which uses a Total Variation formulation. The tumor is now much more localized and the boundary is much sharper compared to a traditional approach.