The Project
Modern medical practice crucially relies on biomedical imaging, of which Computerized Tomography (CT) is a central pillar. Traditional CT uses a series of X-ray projections, post-processed by computer algorithms, to produce cross sectional illustrations of the inside of the body. While the results are very effective, the main drawback of X-ray CT is the harmful effect of radiations for the human body. The present project focuses on a main path of innovation in CT: reduction of patient exposure to radiation. Specifically, we will investigate mathematical and computational themes arising from the emerging technologies of:
- low-dose Computed Tomography (LD-CT), based on a reduced radiation dose per projection and/or a reduced number of projections;
- zero-dose CT, based on non-ionizing investigating signals such as light in Diffuse Optical Tomography (DOT).
In both cases, thereduced ionizing exposure comes at the price of a much noisier/subsampled signal, so that the associated reconstruction problems are severely ill-posed and require advanced mathematical and numerical techniques for adequate treatment.
Methodology
The project aims to achieve significant improvements in the analysis and development of
mathematical and computational tools for effective LD and zero-dose CT imaging, and, namely, to:
- Develop novel numerical models and methods in a unifying framework combining physical models, regularization-based optimization methods, and deep learning techniques for the solution of ill-posed inverse problems in CT imaging
- Design tailored numerical tools and software for the specific LD-CT and DOT applications
- Study networks implementation and evaluation from a Green AI perspective
Expected results
- Development of new optimization schemes easily adaptable also to different tasks for general imaging processing
- Significant improvement of the accuracy of the X-rays LD-CT and DOT reconstructions by integration of physics-driven and data-driven methods, accurate selection of priors to better remove artifacts and enhance contrast, and appropriate modelling of measurement noise
- Validation of the proposed algorithms on simulation tests and, importantly, on real data
Impact
CT has opened the doors for screening, diagnosis and treatment of many diseases including cancer, strokes, heart issues. Most recently, CT has been intensively used to diagnose with high sensitivity conditions connected to the COVID19 virus. This intensive use makes radiation exposure and increasing concern for community health, also in view of further exploit of this technique for screening (e.g., breast and lung cancer screening). The longstanding collaborations of the teams with industrial partners will ensure a solid transfer of the research into industrial innovation, driven by the idea that improving the sustainability of medical CT is a main goal in health promotion.
