FisMat2017 - Submission - View

Abstract's title: Phase contrast breast-CT: from the workflow optimization to the 3D reconstruction
Submitting author: Luca Brombal
Affiliation: University of Trieste & INFN
Affiliation Address: Via Valerio 2, 34127 - Trieste
Country: Italy
Oral presentation/Poster (Author's request): Oral presentation
Other authors and affiliations: Bruno Golosio (Università di Cagliari & INFN), Pasquale Delogu (Università di Siena & INFN), Piernicola Oliva (Università di Sassari & INFN), Vittorio Di Trapani (Università di Pisa & INFN), Fulvia Arfelli (Universita’ di Trieste & INFN), Adriano Contillo (Università di Ferrara & INFN), Sandro Donato (Universita’ di Trieste & INFN), Diego Dreossi (Sincrotrone Trieste), Giovanni Mettivier (Università di Napoli “Federico II” & INFN), Luigi Rigon (Universita’ di Trieste & INFN), Angelo Taibi (Università di Ferrara & INFN), Giuliana Tromba (Sincrotrone Trieste), Renata Longo (Universita’ di Trieste & INFN)

The SYRMA-3D (SYnchrotron Radiation Mammography - 3D) project aims to perform the world’s first clinical trial of phase-contrast breast-CT with synchrotron radiation. Since the acquired images will contribute to the final medical report, a dedicated effort has been made in the optimization of data workflow to provide both good image quality and robustness. Moreover, unlike most synchrotron applications, both the experimental setup and data handling must match strict requirements regarding delivered dose, acquisition time, patient comfort and safety.

Images are acquired using an 8 modules single-photon counting CdTe detector (PIXIRAD-8) which guarantees a high detection efficiency, large field of view, small pixel size (60 µm) and can be operated in a dead-time-free mode, avoiding the deposition of un-used dose. Data need to be dumped at the end of each slice acquisition and safely stored to prevent any accidental loss. The detector requires a specific pre-processing accounting for the outlier pixels, the presence of small gaps (3 pixels) between adjacent modules, ring artefacts and charge trapping effect. Charge trapping is common in high Z crystal and it results in a systematically reproducible polarization time dependent gain variation, which induces artefacts in the tomographic reconstruction: to overcome this issue a new flat field procedure was implemented. This method is based on a two steps correction. In the first step a time dependent gain matrix is built acquiring high statistics flat fields with a homogeneous and stable conventional x-ray source. In the second step a standard flat-fielding procedure using the synchrotron beam is performed. The whole pre-processing procedure for 1200 projections (3000x60 pixels each) requires less than 2 minutes on a 4 core Intel Xeon E5-1630 @ 3.7 GHz workstation.

Subsequently the data are reconstructed using an in-house developed software including several reconstruction algorithms, both filter back projection (FBP) and iterative ones, and the possibility to apply, in the projection’s space, various phase retrieval filters. During the exam an on-the-fly reconstruction is needed for providing a fast feedback, assessing the correct positioning, exposure and the presence of unexpected problems. For this procedure a standard absorption-based FBP is used. After the exam the reconstruction parameters are optimized: a phase retrieval filter is applied and the most appropriate reconstruction algorithm is chosen. Since the volume of interest, typically several centimetres thick, is made up by thin slices (3.5 mm), an image registration procedure is required for achieving a full 3-D view. After this final step the tomographic images can be evaluated by the radiologist.

In this communication we present the building blocks of the data workflow and we show both the pre-processing effectiveness and some of the first 3D images of a breast specimen.