Depth-of-interaction estimates in pixelated scintillator sensors using Monte Carlo techniques

详细信息    查看全文

• 作者：Diksha Sharma^a ; Christina Sze^b ; ¹ ; Harish Bhandari^b ; Vivek Nagarkar^b ; Aldo Badano^a ; ^{aldo.badano@fda.hhs.gov" class="auth_mail" title="E-mail the corresponding author}
• 关键词：Monte Carlo methods ; Small-animal SPECT imaging ; Depth-of-interaction estimation
• 刊名：Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment
• 出版年：2017
• 出版时间：1 January 2017
• 年：2017
• 卷：841
• 期：Complete
• 页码：117-123
• 全文大小：1446 K

文摘

Image quality in thick scintillator detectors can be improved by minimizing parallax errors through depth-of-interaction (DOI) estimation. A novel sensor for low-energy single photon imaging having a thick, transparent, crystalline pixelated micro-columnar CsI:Tl scintillator structure has been described, with possible future application in small-animal single photon emission computed tomography (SPECT) imaging when using thicker structures under development. In order to understand the fundamental limits of this new structure, we introduce cartesiandetect2, an open-source optical transport package that uses Monte Carlo methods to obtain estimates of DOI for improving spatial resolution of nuclear imaging applications. Optical photon paths are calculated as a function of varying simulation parameters such as columnar surface roughness, bulk, and top-surface absorption. We use scanning electron microscope images to estimate appropriate surface roughness coefficients. Simulation results are analyzed to model and establish patterns between DOI and photon scattering. The effect of varying starting locations of optical photons on the spatial response is studied. Bulk and top-surface absorption fractions were varied to investigate their effect on spatial response as a function of DOI. We investigated the accuracy of our DOI estimation model for a particular screen with various training and testing sets, and for all cases the percent error between the estimated and actual DOI over the majority of the detector thickness was ±5% with a maximum error of up to ±10% at deeper DOIs. In addition, we found that cartesiandetect2 is computationally five times more efficient than mantis. Findings indicate that DOI estimates can be extracted from a double-Gaussian model of the detector response. We observed that our model predicts DOI in pixelated scintillator detectors reasonably well.