During the year 2010 the experimental studies has been completed and the results were excellent in spite of the various difficulties (lack of funds and technical problems). In fact, the new instrument was assembled and characterized; the tomographic tests were carried out on dedicated phantoms and they showed how the RITOR instrumentation were able to give a spatial resolution of at least three times to medical scanners, with a decent contrast and an estimated dose to the patient up to 1/40 of the dose delivered by the standard tomographic instrumentation. These very promising results were achieved at the end of the 2010 notwithstanding the reduced budget of the RITOR experiment and the various technical difficulties that got over thanks to the researchers efforts (see the detailed report of the 2010 July).
In the year 2011 a new detector would have to be purchased to enhance the contrast of reconstructed images (the only weak point of the flat-panel system), unfortunately this point has not been funded. Moreover also the purchase of a new E4 cluster has been proposed (it could highly increase the computing performance of the HPC system of the INFN group V) but also this point has not been funded.
For the year 2011, only 3 thousand of euros (consumable materials) has been assigned to the Bologna section (afterward decreased to 2,6 thousand of euros) and to the Cosenza Group 10 thousand of euros "sub judice" bound to the purchase of the software VGStudioMax. Then the section of Bologna could only buy some consumable materials dedicated to the laboratory and modify/fix the translational system (supplied to the group). The section of Cosenza purchased the license of the software for the reconstruction and the rendering of the tomographic images. In particular, the funds were enough only to buy only one node-locked license, and, although the Cosenza group has quickly provided (by means of the network) the pc where the software was installed, the management of a very large amount of data (up to several tens of GB) across the network turned out not feasible.
Notwithstanding the 2011 funds were almost negligible, the section of Bologna has studied and developed a parallel software dedicated to the reconstruction (currently under patent application jointly by INFN and UniBo), has carried out fruitful collaboration with the Grenoble ESRF synchrotron (to study and project an innovative CT system), has presented the results at various conferences, has made contact with several companies interested both in the instrument and in the software (Gilardoni, GE, Cefla, etc... etc...) even if the contacts have not yet been formalized. The section of Cosenza has got along to the study of the maximum resolution achievable with an X-ray point source by means of a dedicated Monte Carlo with Geant4, has studied and experimented the purchased software and is come along to work on the simulation in order to reduce statistical fluctuations and to effectively interpret the results.
The collaboration with the synchrotron in Grenoble, France, ESRF (European Synchrotron Research Facility) was established in 2009 when the research team went to Grenoble for the experimental tests on a plexiglass head phantom with an insert of idrossiopatite (see about the detailed report of the years 2009 and 2010).
The results achieved were exciting both for the high resolution and for the elevated contrast of the reconstructed images. The research group of ESRF was favourably stunned by our instrument which is composed by a fan of optical fibers with scintillator and is coupled to a CCD camera.
This innovative system acts as a geometry transducer: the multislice projections are converted into rectangular images that can be easily read out by a CCD sensor. A system of this type changes a detector of 3000x3000 pixels in a multislice sensor of 20000x400 pixels. In this configuration it can perform at the same time tomographic reconstructions of 400 slices each one composed by 20 thousand per 20 thousand pixels. The resolution is obviously very high (up to 69 um of voxels on a field of view of 17 cm and up to 49 um in the local tomography configuration).
The interest generated by our expertise in the ESRF researchers has led to the formalization of a future collaboration. The synchrotron has in fact been established to allocate funds for the study and for the implementation of a new tomographic tool that is based on our optical fiber fan coupled to an innovative CCD camera partially built by the ESRF. This new CCD should have a very high dynamic range, a high read out speed and it should be coupled directly to our fan in order to obtain the highest possible efficiency. Therefore the DQE of this new system would be superior to all the instruments studied until now, and it would give a voxel around 14 um.
The project will start in 2012, we are currently studying the design of the instrument and we are evaluating different quotations. The cost will be fully supported by the ESRF. As soon as the system will be ready we will go to Grenoble for the experimental tests with synchrotron beam radiation up to 150 keV.
The parallel software
During the 2011 an innovative parallel software system was designed and developed. This software is able to run on HPC clusters with the OS Windows HPC Server 2008. Given that the method and the software are currently under patent application jointly by INFN and UniBo and then under confidentiality agreement, here the most important results obtained will be presented in detail without getting to the hearth of the invention.
The new software provides the management of multiple parallel processes, created by MPI (Message Passing Interface), by means of a code in C/C++ (compiled with VisualStudio 2010) and it is able to simultaneously work in parallel and display graphics. The method also provides the use of National Instruments libraries useful both for the graphical interface and for the management of the instruments acquisition (management of boards, frame grabbers, tools, etc. ..). The elements required to develop the new system are very simple: the Visual Studio 2010 Professional, MPI (communications standard for parallelism), National Instruments libraries (acquisition and graphics), and the algorithms for tomographic imaging.
The program processes the radiographic projections, acquired by the instrument, in several steps until to the three-dimensional reconstruction of the whole volume. The first step is the computation of the attenuated radiographies starting from the projections, then the sinograms (images that contain all the information necessary for the reconstruction of one slice) are generated, and finally the slice are reconstructed. In addition to these standard steps, we have realized dedicated algorithms for noise reduction, for metal artifacts correction and for ring artefacts removal. Moreover a set of utilities are written in order to crop the images, to collate different frames in a one section (as in the study of objects with dimensions greater than the detector), to change the size or bit depth, to perform segmentation analysis both on the basis of gray levels and of morphological properties. The reconstruction step obviously is the most complex both in term of the algorithm complexity and of the time calculation cost. The reconstruction algorithm of fan beam is well known in literature and it is used in the program.
The program performs calculations both in parallel and in sequence, it was designed to adapt itself to the computer or group of computers on which it is installed, it manages dynamically and efficiently the memory, the CPUs and the disk space. The software was designed in order to allow the user to graphically follow each step of the reconstruction and to modify any parameters of the algorithm. It is studied to graphically preview the results before to start the complete calculation.
The test of speed and efficiency are ongoing but the preliminary results are very amazing: the computation time is reduced by sixty compared to the time needed to reconstruct sequentially with a speedup factor of approximately 33 with only 32 real cores). These results are very promising because the system developed not only allows reconstructing in a very short time (essential for the medical CT) but it also permits to consider the use of much more complex algorithms that can improve the contrast and the resolution to obtain more information by the images.
Monte Carlo simulation
The software "Vgstudio" has been purchased; this program is dedicated to the analysis and the rendering of CT data. The software allows performing tomographic reconstructions (Feldkamp) quickly and accurately, while also allowing the display of 2D/3D volumes obtained from CT scans. Once the volume is reconstructed it is possible to segment and to analyze it with numerous tools.
Starting from this processing software and from the data acquired by the prototypes into the laboratory of Bologna, the study of the maximum resolution achievable with an X-ray point-source was started in 2011 in Cosenza/Catanzaro using a dedicated Monte Carlo with Geant4. In practice a simulation model, which reflects the characteristics and the acquisition mode of the prototype used in laboratory experimentation, has been studied and produced.
In particular, the head phantom has been divided into voxels so that the slice containing the inner ear has dimensions, in voxel, of 1 mm3. This solution serves the purpose to estimate the delivered dose in the head. Moreover a detector, with pixels of 50 micron, has been inserted in the position diametrically opposite to the source, and the phantom has been placed in the middle, in order to simulate the tomographic process in laboratory and to use the Vgstudio software to perform the data analysis.
At present the study is not completed and it is being organized the best way to start the simulation in directions that rotate around the head with different step of 5 or 10 degrees. Unfortunately, the simulation will require long times to reduce the statistical fluctuations, and then to be able to effectively interpret the results obtained.
Publications of the year 2011
Because of the impossibility of the ISI database to found and report our publications, following the suggestion of the INFN technical service, the papers are here:
1) "CT Imaging of the Internal Human Ear: Test of a High Resolution Scanner", M.Bettuzzi, R.Brancaccio, M.P.Morigi, A.Gallo, S.Strolin, F.Casali, E.Lamanna, A.Marilů, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, Volume 648, Supplement 1, 21 August 2011, Pages S60-S64, ISSN 0168-9002, 10.1016/j.nima.2010.12.183.
2) Dosimetric Study in the Human Head for CT Investigation of the Inner Ear Using the Geant4 Toolkit; Lamanna, A. S. Fiorillo, A. Gallo, L. Belmonte, A. Narciso; 2010 IEEE Nuclear Science Symposium and Medical Imaging Conference Record, Knoxville Tennessee, October 30 – November 6 2010; ISBN:978-1-4244-9104-9.
At the beginning of 2010 we have received the “flat panel” CMOS Hamamatsu model C10900D, about 1200x1200 pixel, useful area 12x12 cm2, 35frame/sec,. This detector (optimized for digital panoramic to be used in odontoiatry with maximum voltage 120 kV) has a scintillator done by structured CsI. As described in the advancement report of RITOR (July 2010) there was a delay in its operation due to cable connection and to the data acquisition card of the detector.
A new software was developed in May and the characterisation tests were done during June 2010.
The “flat panel” is able to work mainly in two ways: the first with binning 1 (1216x1232 of 100 um pixels and frame rate typically 17 f/s) and the second with binning 2 (608x616 200 um pixel and frame rate typically 35 f/s). Moreover it is possible to use the panel in fast acquisition mode ( for a preliminary analysis and centring of the beam) that allows an higher frame rate but with lower contrast. The detector showed a linear behaviour in each the working modalities with a dynamic ranges of 46 and 48 db and a spatial resolution of 3.5 and 2 lp/mm (MTF 5%), respectively in the two working modalities.
The tomographic tests were carried out beginning from July to December 2010. The tests included variation of binning, frame rate, projection number and comparing the contrast obtained, the spatial resolution, the dose to the patient and the time for the total acquisition. As reference value for the dose to the patient was assumed 17 mSv, which is the medical standard for a CT of the head.
The tests gave very encouraging results concerning the panel with 150 um voxel model: a acceptable contrast (0.4±0.2), with dose from 1/7 to 1/40 of the reference dose with an acquisition time from 64 to 96 s (adequate to medical applications).
In particular we did not expected dose results so low so that there is space for a good improvement of the system for what concerns space resolution and the dynamic range. The key point for the improvement should be a new detector with a 50 um pixel and a better dynamic range as a CMOS could be or one of the new detectors we intend to develop in collaboration with ESRF (Electro Synchrotron Radiation Facility) of Grenoble, (France). Unfortunately the new detector, foreseen for 2011, has not been funded.
The results have been shown to the Head of otorhinolaringology Division of San Orsola Hospital. who found them astonishing. Moreover, during 2010, we have open an interesting collaboration with people of ESRF related to the development of a new detector (of fan-beam type) on the scheme of that we carried to Grenoble in 2009. It consists of a fan of optical fibers where the gross part of the fan is connected with a high dynamic range and fast CCD camera and the thin part of the fan is in close contact with an X-ray scintillator. We were charged with the design of the fan and Grenoble people of the design and development of the fast CCD camera. This new very advanced detector should be ready in 2013. .
We have begun the compilation of a new parallel software with MPI for the distribution, to the single cluster nodes, of the calculation for the 3D tomographic images. That work should be ended on December 2011.
Moreover many contacts with different Italian firms for the testing and commercialisation of the new diagnostic system.
For what the Monte Carlo simulation is concerned, Cosenza group did calculation of the dose distribution in a head during a high spatial resolution CT for putting in evidence the pathologies related to inner ear. The simulation of the head geometry and of the interaction of the particles with the matter was carried out by the code GEANT4(GEometryANd Tracking).Two different approaches have been used for simulating the head geometry: The first mathematical phantom has been obtained combining geometrical solids from the library of GEANT4, the second has been obtained by means of subdivision in voxels.
The study and experimentation activity are carried out both in internal laboratories and in external structures. In particular we point out the test performed on a dental CT scanner under clinical validation in the hospital “S.Orsola” in Bologna. The scanner is produced by a local company (CEFLA). Other tests have been done in the company laboratories with a better resolution prototype. The relationship with CEFLA company is going on both on the scientific and industrial side in order to create a convergence of the scientific and experimentation activity toward the possible development of a commercial product (a CT canner specific for the human internal ear).
Our research group is now in charge of the experimentation program that is preliminary to the design of the prototype. To reach this goal, after an accurate study phase carried out with a research kind of CT scanner in the laboratories of the Department of Physics, it now possible to define a quality reference and start the search and evaluation of the required components to be used in a clinical version on the basis of a certain number of scanning parameters derived experimentally. The activity carried out until the end of 2009 concerns mainly this preliminary study aspect of the project.
2.Brief Technical Description
A first CT Laboratory system has been used to perform a preliminary investigation of the problem of human inner ear X-ray tomography. As a first approach a CT system has been developed using already existing material and instrumentation. The system is composed by the following parts:
1) a detector box based on an Apogee Alta U9000 high performance CCD camera (3053x3056 pixel, over 75% peak Q.E. Kodak sensor) coupled by means of a macro lens and a 45°mirror to a flat scintillating screen
2) a manipulation-rotation platform based on PI micrometric axis
3) an X-ray source (Bosello 120kVpeak-7mA industrial X-ray tube and controller)
As a first work, the scintillating screen has been changed and different scintillators have been compared in order to verify our hypothesis about the best choice of X-ray converter for the final detector. Light output and resolution in the same irradiation conditions have been measured for 3 different kind of X-ray converters. Structured Cesium Iodide resulted at the end the best choice as it is the brightest in light output. However, optimization of scintillator thickness must be carefully evaluated to improve the spatial resolution (we obtained 4.0 lp/mm at 5% MTF with 1mm thick CsI screen). A new CsI screen 0.5mm thick 12x12 cm2 area has been acquired. New measurements revealed a much improved spatial resolution of 6.5 lp/mm at 5% MTF with the new screen. A new detector box has been built around the Apogee Alta U900 camera and the new CsI screen. Other improvement are expected with shielding, large format lenses, mirror regulation, precise box design. However as this system is just a preliminary approach to the inner ear CT problem, we wont invest more time on the development of it and we will go on with new components as a CMOS panel and a new X-ray source.
A number of different scanning have been performed on a human head phantom. The phantom is a commercial kind lend us by our industrial partner CEFLA. It is made of soft tissue equivalent material externally but it contains a real human skull inside. This part maintains all the fine structure features of the inner ear and its surrounding. Thus we considered it a realistic test object. The different scanning produced results of different quality (contrast, resolution) depending on a number of parameters. Results have been submitted to medical experts to be evaluated and have a feedback. At the end of our work the feedback for what we considered the best scanning was enthusiastic. Thus we considered those data as a quality reference for further developments of the project.
4.Choice of new components
The choices operated at the end of the experimentation phase of year 2009, concerning the detector and the X-ray source to be used as components of the prototype, have been suggested by the industrial partner and discussed together on the basis of the results obtained in the laboratories and in the other tests with an eye on the requirements of a final clinical application.
5.Tests at the ESRF
The main activities in our laboratories have been carried out in spring-summer 2009. But at the end of 2009 another experimentation phase has been performed at the European Synchrotron Radiation Facility in Grenoble. The opportunity of using synchrotron radiation is important in terms of absolute quality of the results that is possible to reach in a CT analysis, as it is well known that high flux and mono-energetic radiation are key factors to have reconstructions without artifacts and with the best definition of density values of the materials. In order to do that, we prepared a special detector and a specific human head phantom. The detector is based on a variation of a former coherent fiber optics linear to square optical coupler device that we call “big fan”. The linear side of that fan matches well with the synchrotron radiation beam geometry while the square part matches with a CCD or CMOS camera like device. The coupling could be direct or indirect (by means of a lens). The human head phantom is a coarse geometry representation of a sliced head made of soft tissue equivalent material (Lucite) with an insert of idrossiapatite porous part simulating the skull and the inner ear. Results obtained with this equipment at the ESRF have been recently elaborated and show how it is possible to obtain a very high resolution on the whole human head section including fine details of the inner hear. It is an experiment far from clinical application but very important as a study of the information that one can retrieve from the specific human skull part under investigation to be used mostly as a reference for further work and shown as a case study.
6.Simulation activity (Cosenza Section)
Simulations with Geant4 have been carried out by the connected research group of Cosenza for what concern the dosimetric aspect of the inner ear CT problem. The human skull geometry has been described first using a vectorial model. This approach is indicated to speed up the simulation when a global mean dose value is required in a short time. The second approach is a voxel model, directly connected with real data retrieved from a CT analysis of the human internal ear. This second model is indicated to have information on the localized distribution of the dose. The fineness of the voxel is tunable, according with the required spatial resolution. The level of detail of the simulation, in this case, will affect strongly the calculation time.