Room: Track 1
Purpose: use the compositional material information of the breast as a structural prior knowledge for diffuse optical tomography (DOT), the breast image was decomposed into the water, lipid, and protein volume images by use of the dual-energy digital breast tomosynthesis (DEDBT).
Methods: digital breast phantom was constructed for Monte Carlo simulation to perform the dual-energy digital breast tomosynthesis. The DBT reconstructed phantom image was decomposed into the three basis material volumes (water, lipid, and protein) using the high energy, low energy measurement data, and the total thickness. An iterative method was implemented for DOT reconstruction (NIRFAST software based) with a Laplacian matrix for regularization based on these material volumes. From the known optical properties of the materials of the phantom, the amplitude and phase data were acquired by the forward model (random noise added) in a realistic system geometry and the reconstruction was performed. The proposed method was compared with the existing method in terms of recovery of the absorption coefficient and edge-preserving.
Results: reconstruction results were compared in each z-axis slice at the same window level under different regularization parameter conditions. The proposed method, in all regularization parameter conditions, successfully recovered the absorption coefficient of the tumor and the edges were preserved while the existing method was not successful.
Conclusion: work indicates the feasibility of more accurate image reconstruction in DOT by use of DEDBT as a structural prior without any external segmentation. The sensitivity of the tumor detection for dense breast cases is supposed to increase when the proposed method is integrated with the DBT/DOT multi-imaging modality system.