An implementation of the NiftyRec medical imaging library for PIXE-tomography reconstruction

Claire Michelet (1) , Philippe Barberet (1) , Pascal Desbarats (2) , Jean-François Giovannelli (3) , C. Schou (1) , I. Chebil (1) , Marie-Hélène Delville (4) , Nuria Gordillo (1) , Daniel G. Beasley (5) , Guillaume Devès (1) , Philippe Moretto (1) , Hervé Seznec (1)

1 CENBG - Centre d'Etudes Nucléaires de Bordeaux Gradignan
2 LaBRI - Laboratoire Bordelais de Recherche en Informatique
3 IMS - Laboratoire de l'intégration, du matériau au système
4 ICMCB - Institut de Chimie de la Matière Condensée de Bordeaux
5 CMIC - Centre for Medical Image Computing

A new development of the TomoRebuild software package is presented, including “thick sample” correction for non linear X-ray production (NLXP) and X-ray absorption (XA). As in the previous versions, C++ programming with standard libraries was used for easier portability. Data reduction requires different steps which may be run either from a command line instruction or via a user friendly interface, developed as a portable Java plugin in ImageJ. All experimental and reconstruction parameters can be easily modified, either directly in the ASCII parameter files or via the ImageJ interface. A detailed user guide in English is provided. Sinograms and final reconstructed images are generated in usual binary formats that can be read by most public domain graphic softwares. New MLEM and OSEM methods are proposed, using optimized methods from the NiftyRec medical imaging library. An overview of the different medical imaging methods that have been used for ion beam microtomography applications is presented. In TomoRebuild, PIXET data reduction is performed for each chemical element independently and separately from STIMT, except for two steps where the fusion of STIMT and PIXET data is required: the calculation of the correction matrix and the normalization of PIXET data to obtain mass fraction distributions. Correction matrices for NLXP and XA are calculated using procedures extracted from the DISRA code, taking into account a large X-ray detection solid angle. For this, the 3D STIMT mass density distribution is used, considering a homogeneous global composition. A first example of PIXET experiment using two detectors is presented. Reconstruction results are compared and found in good agreement between different codes: FBP, NiftyRec MLEM and OSEM of the TomoRebuild software package, the original DISRA, its accelerated version provided in JPIXET and the accelerated MLEM version of JPIXET, with or without correction.

Mots clés: PIXE tomography; Quantitative imaging; Filtered backprojection; MLEM; OSEM; ImageJ; Caenorhabditis elegans

Domaines: Matériaux

Claire Michelet, Philippe Barberet, Pascal Desbarats, Jean-François Giovannelli, C. Schou, et al.. An implementation of the NiftyRec medical imaging library for PIXE-tomography reconstruction. Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, 2017, 404, pp.131-139. ⟨10.1016/j.nimb.2017.01.067⟩. ⟨hal-01544227⟩