X-ray imaging exams are widely used procedures in medical diagnosis. Whenever an x-ray imaging procedure is performed, it is accompanied by scattered radiation. Scatter is a significant contributor to the degradation of image quality in breast CT. This work uses our understanding of the physics of Compton scattering to overcome the reduction in image quality that typically results from scattered radiation. By measuring the energy of the scattered photons at various locations about the object, an electron density (ρe) image of the object can be obtained. This work investigates a system modeled using a 2D cross-sectional view of a breast CT geometry. The ρe images can be obtained using filtered backprojection over isogonic curves. If the detector has ideal energy and spatial resolution, a single projection will enable a high quality image to be reconstructed. However, these ideal characteristics cannot be achieved in practice and as the detector size and energy resolution diverge from the ideal, the image quality degrades. To compensate for the realistic detector specifications a multi-projection Compton scatter tomography (MPCST) approach was introduced. In this approach an x-ray source and an array of energy sensitive photon counting detectors located just outside the edge of the incident fan-beam, rotate around the object while acquiring scattering data. The ρe image quality is affected by the size of the detector, the energy resolution of the detector and the number of projections. These parameters, their tradeoffs and the methods for the image quality improvement were investigated. The work has shown that increasing the energy and spatial resolution of the detector improves the spatial resolution of the reconstructed ρe image. These changes in the size and energy resolution result in an increase in the noise. Thus optimizing the image quality becomes a tradeoff between blurring and noise. We established that a suitable balance is achieved with a 500 eV energy resolution and 2×2 mm2 detector. We have also established that using a multi-projection approach can offset the increase in the noise.

Breast CT is a promising modality whose inherent scatter could be used to reconstruct electron density (rho_e) images. This has led us to investigate the benefits of reconstructing linear attenuation coefficient (mu) and (rho_e) images of the breast. First generation CT provides a cost-effective and simple approach to reconstruct (rho_e) images in a laboratory but is limited by the anisotropic probability of scatter, attenuation, noise and contaminating scatter (coherent and multiple scatter). These issues were investigated using Monte Carlo (MC) simulations of a first generation breast scatter enhanced CT (B-SECT) system. A reconstruction algorithm was developed for the B-SECT system and is based on a ring of detectors which eliminates the scatter dependence on the relative position of the scattering centre. The algorithm incorporates an attenuation correction based on the (mu) image and was tested against analytical and MC simulations. MC simulations were also used to quantify the dose per scan. The ring measures a fraction of the total single incoherent scatter which is proportional to ray integrals of (rho_e) and can be quantified even when electron binding is non negligible. The algorithm typically reconstructs accurate (rho_e) images using a single correction for attenuation but has the capability for multiple iterations if required. MC simulations show that the dose coefficients are similar to those of cone beam breast CT. Coherent and multiple scatter can not be directly related to (rho_e) and lead to capping artifacts and overestimated (rho_e) by a factor greater than 2. This issue can be addressed using empirical corrections based on the radiological path of the incident beam and result in (rho_e) images of breast soft tissue with 1% accuracy, 3% precision and a mean glandular dose of 4 mGy for a 3D scan. The reconstructed (rho_e) image was more accurate than the (rho_e) estimate derived from the (mu) image. An alternative correction based on the thickness of breast traversed by the beam provides an enhanced contrast image reflecting the breast scatter properties. These results demonstrate the feasibility of detecting small (rho_e) changes in the intact breast and shows that further experimental evaluation of this technique is warranted.

Breast CT is a promising modality whose inherent scatter could be used to reconstruct electron density (rho_e) images. This has led us to investigate the benefits of reconstructing linear attenuation coefficient (mu) and (rho_e) images of the breast. First generation CT provides a cost-effective and simple approach to reconstruct (rho_e) images in a laboratory but is limited by the anisotropic probability of scatter, attenuation, noise and contaminating scatter (coherent and multiple scatter). These issues were investigated using Monte Carlo (MC) simulations of a first generation breast scatter enhanced CT (B-SECT) system. A reconstruction algorithm was developed for the B-SECT system and is based on a ring of detectors which eliminates the scatter dependence on the relative position of the scattering centre. The algorithm incorporates an attenuation correction based on the (mu) image and was tested against analytical and MC simulations. MC simulations were also used to quantify the dose per scan. The ring measures a fraction of the total single incoherent scatter which is proportional to ray integrals of (rho_e) and can be quantified even when electron binding is non negligible. The algorithm typically reconstructs accurate (rho_e) images using a single correction for attenuation but has the capability for multiple iterations if required. MC simulations show that the dose coefficients are similar to those of cone beam breast CT. Coherent and multiple scatter can not be directly related to (rho_e) and lead to capping artifacts and overestimated (rho_e) by a factor greater than 2. This issue can be addressed using empirical corrections based on the radiological path of the incident beam and result in (rho_e) images of breast soft tissue with 1% accuracy, 3% precision and a mean glandular dose of 4 mGy for a 3D scan. The reconstructed (rho_e) image was more accurate than the (rho_e) estimate derived from the (mu) image. An alternative correction based on the thickness of breast traversed by the beam provides an enhanced contrast image reflecting the breast scatter properties. These results demonstrate the feasibility of detecting small (rho_e) changes in the intact breast and shows that further experimental evaluation of this technique is warranted.

Neste trabalho, avaliou-se a quantidade de informação diagnóstica contida na distribuição da radiação espalhada em mamografia, através de simulações Monte Carlo (MC). Para isto, este trabalho consistiu de dois objetivos: o primeiro diz respeito ao desenvolvimento de um código MC para o transporte de fótons em radiodiagnóstico, com ênfase em mamografia. O segundo diz respeito ao estudo da distribuição da radiação espalhada pela mama, e seu potencial diagnóstico. O modelo geométrico adotado nas simulações consistiu de uma mama comprimida semi-infinita, com um nódulo esférico inserido, simulando um nódulo maligno. Um receptor plano ideal foi posicionado abaixo da mama, a uma distância h. A distribuição angular da radiação espalhada pela mama, e sua distribuição espacial sobre o receptor, foram obtidas para um feixe estreito incidindo perpendicularmente sobre a superfície da mama. Estas distribuições foram utilizadas para calcular valores de contraste (CS) e razão contraste-ruído (CNRS) dos fótons espalhados, comparando as distribuições provenientes de regiões da mama sem e com o nódulo. Valores de CS e CNRS foram estudados para diferentes energias do feixe incidente, tamanho e posição do nódulo, e espessura e composição da mama. A influência de feixes polienergéticos nos valores de CS e CNRS também foi investigada. Para a distribuição espacial da radiação espalhada, valores de CS e CNRS também foram estudados como função da distância do receptor à mama. Os resultados mostram que a distribuição da radiação espalhada apresenta picos de espalhamento, que são produzidos pelos fótons elasticamente espalhados, e estão relacionados com a composição da mama. As distribuições angulares de CS e CNRS mostraram que valores máximos destas distribuições ocorrem próximos ao primeiro pico de espalhamento. Valores de CS maiores que o contrate primário foram obtidos em todas situações analisadas, embora o CNRS tenha se mostrado consideravelmente mais baixo que o CNR primário. As distribuições espaciais de CS e CNRS no receptor indicam que o uso de um receptor plano não reduz os valores de CS, comparados com os obtidos para a distribuição angular, embora o CNRS decresça a medida que h aumenta. Imagens planares, obtidas utilizando o feixe espalhado, mostram que, além de fornecer valores de contraste maiores que o contraste primário, a técnica permite realçar um determinado tipo de tecido na imagem, a partir da seleção de um determinado valor de momentum transferido x. Os resultados obtidos neste trabalho indicam que a radiação espalhada contém informação diagnóstica a respeito da presença de um nódulo na mama. Estudos futuros a respeito da otimização das condições de irradiação da mama e detecção da radiação espalhada devem ser realizados, a fim de se aumentar os valores de CNRS, sem comprometer os valores de CS. / In this work, the potential of forward x-ray scattering for contrast enhancement of malignant nodules in mammography was studied through MC simulations. This work consisted of two objectives: the ¯rst one refers to the development of a Monte Carlo (MC) code for simulation of photon transport in radiodiagnostic, focusing on mammography. The second objective refers to the study of the distribution of scattered photons by the breast, and its diagnostic potential. The geometric model adopted in the simulations consisted of a semi-in¯nite compressed breast, with a spherical nodule inserted within it, simulating a nodule. A planar ideal receptor was positioned under the breast, at a distance h from it. The angular distribution of scattered photons exiting the breast, and its spatial distribution on the receptor, were obtained for a pencil beam impinging normally on the breast surface. These distributions were used to compute values of scatter contrast (CS) and contrast-to-noise ratio (CNRS), by comparing the signal from regions of the breast without and with the nodule. Values of CS and CNRS were studied for di®erent beam energies, nodule size and position, and breast thickness and composition. In°uence of polienergetic spectra on CS and CNRS were also investigated. For the spatial distribution of scattered photons on the receptor, values of CS and CNRS were also studied as a function of the distance of the receptor to the breast. Results show that the distributions of scattered photons present scattering peaks, which are yielded by the elastic scattered photons, and are related to the breast composition. The angular distributions of CS and CNRS showed that maximum values occur close to the adipose scattering peak. Values of CS greater than primary contrast were obtained in all situations analyzed, although the CNRS was considerably lower than primary CNR. The spatial distributions of CS and CNRS indicate that the use of a planar receptor does not reduce the values of CS, compared with those from the angular distribution, although the CNRS decreases as h increases. Planar images, obtained for the scattered beam, showed that, in addiction to contrast enhancement, this technique allows the accentuation of a given tissue in image, by selecting a given value of transfer momentum x. The results obtained in this work indicate that scattered radiation contains diagnostic information about the presence of a nodule within the breast. Further studies, regarding optimization of conditions of breast irradiation and radiation detection, should be performed, in order to increase values of CNRS, without reducing values of CS.

Neste trabalho, avaliou-se a quantidade de informação diagnóstica contida na distribuição da radiação espalhada em mamografia, através de simulações Monte Carlo (MC). Para isto, este trabalho consistiu de dois objetivos: o primeiro diz respeito ao desenvolvimento de um código MC para o transporte de fótons em radiodiagnóstico, com ênfase em mamografia. O segundo diz respeito ao estudo da distribuição da radiação espalhada pela mama, e seu potencial diagnóstico. O modelo geométrico adotado nas simulações consistiu de uma mama comprimida semi-infinita, com um nódulo esférico inserido, simulando um nódulo maligno. Um receptor plano ideal foi posicionado abaixo da mama, a uma distância h. A distribuição angular da radiação espalhada pela mama, e sua distribuição espacial sobre o receptor, foram obtidas para um feixe estreito incidindo perpendicularmente sobre a superfície da mama. Estas distribuições foram utilizadas para calcular valores de contraste (CS) e razão contraste-ruído (CNRS) dos fótons espalhados, comparando as distribuições provenientes de regiões da mama sem e com o nódulo. Valores de CS e CNRS foram estudados para diferentes energias do feixe incidente, tamanho e posição do nódulo, e espessura e composição da mama. A influência de feixes polienergéticos nos valores de CS e CNRS também foi investigada. Para a distribuição espacial da radiação espalhada, valores de CS e CNRS também foram estudados como função da distância do receptor à mama. Os resultados mostram que a distribuição da radiação espalhada apresenta picos de espalhamento, que são produzidos pelos fótons elasticamente espalhados, e estão relacionados com a composição da mama. As distribuições angulares de CS e CNRS mostraram que valores máximos destas distribuições ocorrem próximos ao primeiro pico de espalhamento. Valores de CS maiores que o contrate primário foram obtidos em todas situações analisadas, embora o CNRS tenha se mostrado consideravelmente mais baixo que o CNR primário. As distribuições espaciais de CS e CNRS no receptor indicam que o uso de um receptor plano não reduz os valores de CS, comparados com os obtidos para a distribuição angular, embora o CNRS decresça a medida que h aumenta. Imagens planares, obtidas utilizando o feixe espalhado, mostram que, além de fornecer valores de contraste maiores que o contraste primário, a técnica permite realçar um determinado tipo de tecido na imagem, a partir da seleção de um determinado valor de momentum transferido x. Os resultados obtidos neste trabalho indicam que a radiação espalhada contém informação diagnóstica a respeito da presença de um nódulo na mama. Estudos futuros a respeito da otimização das condições de irradiação da mama e detecção da radiação espalhada devem ser realizados, a fim de se aumentar os valores de CNRS, sem comprometer os valores de CS. / In this work, the potential of forward x-ray scattering for contrast enhancement of malignant nodules in mammography was studied through MC simulations. This work consisted of two objectives: the ¯rst one refers to the development of a Monte Carlo (MC) code for simulation of photon transport in radiodiagnostic, focusing on mammography. The second objective refers to the study of the distribution of scattered photons by the breast, and its diagnostic potential. The geometric model adopted in the simulations consisted of a semi-in¯nite compressed breast, with a spherical nodule inserted within it, simulating a nodule. A planar ideal receptor was positioned under the breast, at a distance h from it. The angular distribution of scattered photons exiting the breast, and its spatial distribution on the receptor, were obtained for a pencil beam impinging normally on the breast surface. These distributions were used to compute values of scatter contrast (CS) and contrast-to-noise ratio (CNRS), by comparing the signal from regions of the breast without and with the nodule. Values of CS and CNRS were studied for di®erent beam energies, nodule size and position, and breast thickness and composition. In°uence of polienergetic spectra on CS and CNRS were also investigated. For the spatial distribution of scattered photons on the receptor, values of CS and CNRS were also studied as a function of the distance of the receptor to the breast. Results show that the distributions of scattered photons present scattering peaks, which are yielded by the elastic scattered photons, and are related to the breast composition. The angular distributions of CS and CNRS showed that maximum values occur close to the adipose scattering peak. Values of CS greater than primary contrast were obtained in all situations analyzed, although the CNRS was considerably lower than primary CNR. The spatial distributions of CS and CNRS indicate that the use of a planar receptor does not reduce the values of CS, compared with those from the angular distribution, although the CNRS decreases as h increases. Planar images, obtained for the scattered beam, showed that, in addiction to contrast enhancement, this technique allows the accentuation of a given tissue in image, by selecting a given value of transfer momentum x. The results obtained in this work indicate that scattered radiation contains diagnostic information about the presence of a nodule within the breast. Further studies, regarding optimization of conditions of breast irradiation and radiation detection, should be performed, in order to increase values of CNRS, without reducing values of CS.