Implementation of Digital Tomosynthesis in a Real Radiology System

The work included in this thesis is framed on one of the lines of research carried out by the Biomedical Imaging and Instrumentation Group from the Bioengineering and Aerospace Department of Universidad Carlos III de Madrid working jointly with the Gregorio Marañón Hospital. Its goal is to design and develop a new generation of Radiology Systems, valid for clinical and veterinary applications, through the research and development of innovative technologies in advanced image processing oriented to increase image quality, to reduce dose and to incorporate tomographic capabilities. The latter will allow bringing tomography to situations in which a CT system is not allowable due to cost issues or when the patient cannot be moved (for instance, during surgery or ICU). It may also be relevant to reduce the radiation dose delivered to the patient, if we can obtain a tomographic image from fewer projections than using a CT.

In that context, this thesis deals with incorporating pseudo-tomographic capabilities, through a tomosynthesis protocol, in a radiology room originally designed for planar images: the NOVA FA digital radiography system developed by SEDECAL. The room consists of an X-ray generator, a vertical wall stand system, a mobile elevating table and an automatic ceiling suspension which allows the X-ray source to cover the whole volume of the room. Images are acquired using a flat panel detector connected through Wi-Fi to the computer station.

Having evolved from conventional tomography, tomosynthesis produces section images at any depth from projections obtained at different angles along a linear sweep through the use of a suitable reconstruction algorithm.

A workflow was established for the incorporation of tomosynthesis protocols to the NOVA FA system starting from the design of the protocol down to the reconstruction step. This required the understanding of the system and the development of several software tools.

For the design of new protocols, a tomosynthesis module was incorporated to an in-house X-ray simulation tool programmed in Matlab and CUDA. 

As the X-ray room was built specifically for research, everything is manual and all the software is open. This system is designed only for planar radiography and, as a consequence, it is very cumbersome to incorporate a protocol that involves more than one projection. Therefore, a new software tool was implemented in Matlab that allows the translation of each of the source-detector positions corresponding to the tomosynthesis design to the geometrical parameters of the NOVA FA system and their automatic addition to its database.

To obtain a tomographic image from the data acquire, a reconstruction tool was developed in Matlab with the ability to use several reconstruction algorithms including Shift-and-Add and Backprojection. 

Finally, two different evaluations were conducted: a geometric evaluation to assess the correlation between the simulation tool and the X-ray room and an evaluation of the complete workflow through the design and implementation of a simple tomosynthesis protocol using a PBU-50 body phantom developed by Kyoto Kagatu. The results of these evaluation studies showed the feasibility of the proposal.

It should be noted that the work of this thesis has a clear application in industry, since it is part of a proof of concept of the new generation of radiology systems which will be commercialised worldwide by the company SEDECAL.