Deep scatter estimation (DSE) for X-ray computed tomography or positron emission tomography (PET) uses convolutional neural networks (CNNs) to estimate scatter distributions. We investigate the impact of physically motivated transformations and combinations of emission and attenuation input features on PET-DSE performance. Therefore, we decompose the analytical expression of a convolutional scatter model into different feature sets as a function of measured prompts and attenuation correction factors, and propose to use individual attenuation sinograms of central slabs and peripheral regions. Data from 20 patients ( 71 bed positions, 17 892 direct views) were collected and used to train CNNs to estimate the single scatter simulation (SSS) from various feature sets. Adding redundant attenuation features improved the convergence of validation metrics. Slab-wise attenuation sinograms improved training mean absolute errors by 10% and early-epoch validation metrics, yet without improvement in later epochs. In conclusion, physically motivated transformation of input features can help improve training and estimation performance in PET-DSE.
The contribution of scattered x-rays to the acquired projection data is a severe issue in cone-beam CT (CBCT). Due to the large cone angle, scatter-to-primary ratios may easily be in the order of 1. The corresponding artifacts which appear as cupping or dark streaks in the CT reconstruction may impair the diagnostic value of the CT examination. Therefore, appropriate scatter correction is essential. The gold standard is to use a Monte Carlo photon transport code to predict the distribution of scattered x-rays which can be subtracted from the measurement subsequently. However, long processing times of Monte Carlo simulations prohibit them to be used routinely. To enable fast and accurate scatter estimation we propose the deep scatter estimation (DSE). It uses a deep convolutional neural network which is trained to reproduce the output of Monte Carlo simulations using only the acquired projection data as input. Once the network is trained, DSE performs in real-time. In the present study we demonstrate the feasibility of DSE using simulations of CBCT head scans at different tube voltages. The performance is tested on data sets that significantly differ from the training data. Thereby, the scatter estimates deviate less than 2% from the Monte Carlo ground truth. A comparison to kernel-based scatter estimation techniques, as they are used today, clearly shows superior performance of DSE while being similar in terms of processing time.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.