Development and tomographic reconstruction of an innovative MBI device for the early breast cancer diagnosis

Breast cancer affects 12% of women and is the second leading cause of cancer death for them. Thanks to improved diagnosis, death rate from beast cancer has dropped by 40% in the last decade. The standard breast cancer screening exam is X-ray mammography, thanks to its generally good sensitivity and high specificity. However it is less sensitive and specific in dense breast, which is related to the higher probability of cancer development. To improve this relevant limitations, an innovative and compact Molecular Breast Imaging (MBI) system, with two asymmetric dedicated gamma cameras, has been designed at "Istituto Superiore di Sanità" in Rome, to increase significantly the capability of small tumours detection, acting as a second level screening device. Thanks to its configuration flexibility and the use of different optics (both pinhole and parallel hole collimators) and detector sizes, the system allows either spot compression (SC) of the breast, bringing one gamma sensor closer to the suspicious lesion and then increasing sensitivity or limited angle tomography (LAT) for a 3-D reconstruction of the breast lesion and its depth localization. The current work has been focused on the improvements of the MBI system in terms of more detailed calibration and characterization in realistic conditions, the development of realistic simulation models and related analysis procedures as well as tomographic reconstruction methods in LAT configuration. Characterization measurements have been performed in clinical environment to assess and improve system performances in SC and (preliminary) LAT modes using breast and tumours phantoms. The obtained results clearly show a detector signal-to-noise ratio (SNR) larger than 5 for lesion down to 4 mm diameter in SC mode. The preliminary measurements in LAT mode suggest the ability to image tumours, at different depths, by using a limited set of breast views, and that the reconstructed tumour depth accuracy strongly depends on the mechanical alignment. Realistic and detailed Monte Carlo simulations of the MBI system, as well as the data digitization and images analysis, have been performed in the GATE framework to study the MBI and LAT reconstruction performances minimizing the need of dedicated clinical campaigns. The results of the above characterization measurements have been also used to validate the simulations. Finally the simulated data were reconstructed using an iterative maximum likelihood expectation maximization algorithm (ML-EM) with 1 mm^3 voxels and each detector head’s response was included in the reconstruction process. Moreover an ML-EM algorithm for LAT reconstruction has been tested and optimized on a simplified, fast, simulation model and the above realistic Monte Carlo. The reconstruction performances have been estimated varying tumour size, depth and uptake; overall the identification and the depth localization of tumour suffer from the LAT artefacts but they improve with a wider angular coverage. The present thesis will describe in details the MBI system and the above mentioned research and development activities and their achievements.