Feasibility of tissue differentiation with multi-energy computed tomography: a Monte Carlo breast phantom study
| dc.contributor.advisor | Du Plessis, F. C. P. | |
| dc.contributor.author | Van Eeden, Déte | |
| dc.date.accessioned | 2018-01-29T06:53:15Z | |
| dc.date.available | 2018-01-29T06:53:15Z | |
| dc.date.issued | 2017-03 | |
| dc.description.abstract | English: Dedicated breast CT is a new innovative way of imaging breast tissue without the limitations of overlapping anatomical features. It has been shown that the dose received by the patient is comparable to that of conventional mammography techniques. Further developments have led to the idea of a photon-counting detector that can be utilised in conjunction with breast CT. This will produce images with higher CNRs and will improve the detection of malignant masses. Other applications of multi-energy CT include image-based energy weighting and the differentiation of different tissues. The aim of this study was to explore the feasibility for tissue differentiation in breast tissue through the Monte Carlo simulation of a virtual multi-energy CT unit. The EGSnrc Monte Carlo code was used to simulate a virtual CT unit, similar to the Toshiba Aquillion LB 16 CT. The radiation source modelling code, BEAMnrc, was used to model the different components of the virtual CT. These components include the X-ray tube, suitable filters and beam-defining components such as collimators. A phase-space file was obtained consisting of all the particles generated by the different components. The energy spectrum of the Toshiba Aquillion LB 16 CT was approximated by the virtual CT using HVL measurements. The RMI electron density CT phantom was used to benchmark the virtual CT against the Toshiba Aquillion LB 16 CT. The phantom consists of several inserts with known electron densities that produce different CT numbers. A similar phantom was modelled with an in-house developed IDL program and used for the simulations. The reconstructed images were then used for the benchmarking of the HUs. This benchmarking ensured that the method used in this study produces a realistic model of a CT unit. Breast simulator software was used to model three breast phantoms consisting of different glandularities. The composition of the different breast tissues was taken from literature. The three phantoms were simulated at 20 keV up to 65 keV in 5 keV increments. All of the image reconstructions in this study was done with a filtered backprojection algorithm by using the OSCar reconstruction software. The CNRs of the different images obtained at different energies were assessed. Image-based energy weighting was investigated to further enhance the CNRs of the images by multiplying each energy bin with a specific weighting factor. The weighting factors were determined by a random number generator in an in-house developed IDL code. Good results were obtained with a 1.2-1.3 fold increase in the CNR. Further improvements were made by applying constraints to the weighting factors of the different energy bins. A new method was proposed to differentiate between different breast tissues by using the mass attenuation information from multiple energies. This technique showed promising results and can detect malignant tissue by using a single egs_cbct simulation. In conclusion, it is feasible to differentiate between different breast tissue types when using a multiple-energy CT unit. Better CNRs are obtained when utilising the information of the entire energy spectrum. This will lead to better tumour detection, even in dense breasts consisting of 89% glandular tissue. | en_ZA |
| dc.description.abstract | Afrikaans: Toegewyde bors RT is ‘n nuwe innoverende manier om borsweefsel te beeld sonder die beperkinge van oorvleuelende anatomiese kenmerke. Dit het getoon dat die dosis wat die pasiënt ontvang, vergelykbaar is met dié van konvensionele mammografie tegnieke. Verdere ontwikkelings het gelei tot die idee van ‘n foton-tel detektor wat gebruik kan word in samewerking met die bors RT. Dit sal beelde met ‘n hoër kontras-tot-ruis verhouding produseer en sal die opsporing van kwaadaardige massas verbeter. Ander gebruike van multi-energie RT, sluit beeld-gebaseerde energie beelding en die differensiasie van verskillende weefsels in. Die doel van hierdie studie was om die haalbaarheid vir weefsel differensiasie in borsweefsel te verken deur die Monte Carlo simulasie van ‘n virtuele multi-energie RT-eenheid. Die EGSnrc Monte Carlo-kode is gebruik om ‘n virtuele RT-eenheid, soortgelyk aan die Toshiba Aquillion LB 16 RT na te boots. Die bestraling bron modellerings kode, BEAMnrc, is gebruik om die verskillende komponente van die virtuele RT te modelleer. Hierdie komponente sluit in die X-straalbuis, geskikte filters en veldgrootte definieer komponente soos kollimators in. ‘n Fase-ruimte lêer is verkry uit al die deeltjies gegenereer deur die verskillende komponente. Die energie spektrum van die Toshiba Aquillion LB 16 RT was benader deur die virtuele RT deur middel van half-waarde laag metings. Die RMI elektrondigtheid RT fantoom is gebruik om die virtuele RT teen die Toshiba Aquillion LB 16 RT te maatstaf. Die fantoom bestaan uit verskeie insetsels met ‘n bekende elektron digthede wat verskillende RT getalle te produseer. ‘n Soortgelyke fantoom is ontwikkel met ‘n self ontwikkelde IDL kode, wat gebruik was vir die simulasies. Die gerekonstrueerde beelde is dan gebruik vir die maatstaf van die Hounsfield eenhede. Die maatstaf verseker dat die metode wat gebruik word in hierdie studie, ‘n realisitiese model van ‘n RT-eenheid kan lewer. Bors simuleerder sagteware is gebruik om drie bors fantome te modelleer wat bestaan uit verskillende digthede. Die samestelling van die verskillende bors weefsel is geneem uit die literatuur. Die drie fantome is gesimuleer met 20 keV tot 65 keV in inkremente van 5 keV. Al die beeld rekonstruksies in hierdie studie is gedoen met ‘n gefilterde terug-projeksie algoritme deur gebruik te maak van die OSCaR rekonstruksie sagteware. Die kontras-tot-ruis verhouding van die verskillende beelde, by verskillende energieë was beoordeel. Beeld-gebaseerde energie beelding was ondersoek om die kontras-tot-ruis verhouding verder te versterk deur elke beeld by ‘n spesifieke energie te vermenigvuldig met ‘n spesifieke gewig faktor. Die gewig faktore word bepaal met ‘n ewekansige getal genereerder in ‘n self ontwikkelde IDL kode. Goeie resultate is verkry met ‘n 1,2-1,3 voudige toename in die kontras-tot-ruis verhouding. Verdere verbeterings is gemaak deur die toepassing van beperkings om die gewig faktore van die verskillende energie te bepaal. ‘n Nuwe metode is voorgestel om te onderskei tussen verskillende bors weefsel deur gebruik te maak van die massa attenuasie inligting uit verskeie energie. Hierdie tegniek het belowende resultate en kan kwaadaardige weefsel op spoor deur die gebruik van ‘n enkele egs_cbRT simulasie. Ten slotte, dit is moontlik om te onderskei tussen verskillende tipes borsweefsel deur die gebruik van ‘n multi-energie RT-eenheid. Beter kontras-tot-ruis verhoudings word verkry wanneer die inligting van die hele energie spektrum gebruik word. Dit sal lei tot ‘n beter gewas opsporing, selfs in digte borste wat bestaan uit 89% klierweefsel. | af |
| dc.description.sponsorship | South African Medical Research Council | en_ZA |
| dc.identifier.uri | http://hdl.handle.net/11660/7736 | |
| dc.language.iso | en | en_ZA |
| dc.publisher | University of the Free State | en_ZA |
| dc.rights.holder | University of the Free State | en_ZA |
| dc.subject | Multiple-energy CT | en_ZA |
| dc.subject | Breast CT | en_ZA |
| dc.subject | Monte Carlo | en_ZA |
| dc.subject | Egs_cbct | en_ZA |
| dc.subject | Image-based energy weighting | en_ZA |
| dc.subject | Tissue differentiation | en_ZA |
| dc.subject | Breast -- Cancer | en_ZA |
| dc.subject | Tomography | en_ZA |
| dc.subject | Monte Carlo method | en_ZA |
| dc.subject | Thesis (Ph.D. (Medical Physics))--University of the Free State, 2017 | en_ZA |
| dc.title | Feasibility of tissue differentiation with multi-energy computed tomography: a Monte Carlo breast phantom study | en_ZA |
| dc.type | Thesis | en_ZA |