While analytic and iterative reconstruction techniques have been studied extensively for several decades, direct reconstruction of medical images using convolutional neural networks has only recently received attention. In direct reconstruction schemes, network architecture plays a profound role, influencing how efficiently the network learns the necessary domain transform and how efficiently it learns and stores prior information. Previous studies have, however, not rigorously tested diverse architectures against each other. In this work, Monte Carlo simulations of realistic positron emission tomography data acquisition were performed, and the data were binned into 2-D sinograms. A flexible architecture was used to generate reconstruction networks whose characteristics depended on 15 hyperparameters. A Bayesian search algorithm was employed to efficiently search the hyperparameter space for the best-performing networks (tuned networks), according to two quality metrics: mean squared error (MSE) and structural similarity index metric (SSIM). A total of 341 networks were trialed. The best-performing networks consistently outperformed reconstruction by maximum likelihood expectation maximization, both in terms of mean SSIM (0.887 vs. 0.855) and MSE (0.711 vs. 0.854), but also on an image-by- image basis, as evidenced by 2D metric histograms. Furthermore, compared to untuned networks, tuned networks used less training data (70k vs. 105k training examples) and required far fewer epochs to converge (6 vs. 150). Compared to metrically inferior ML-EM images, network-reconstructed images suffered from over-smoothing, loss of finer details, and over-regularity of high-contrast regions. Since networks performed well by the metrics, this indicates that MSE and SSIM did not adequately quantify important image features.
Optical Coherence Tomography (OCT) is an imaging technique that could image subsurface structural details of an object. The imaging depth of OCT is ultimately limited due to multiple scattering of light. We report on the development of a numerical simulation to characterize the impact of multiple-scattered light in Swept-Source Optical Coherence Tomography. (see manuscript)
We are investigating methods for computational scatter estimation for scatter correction in cone-beam computed
tomography. We have developed an analytical method for estimating single scatter. The paper discusses our analytical
method and its validation using Monte Carlo simulations. The paper extends previous results to include both Compton
and Rayleigh single scatter interactions. The paper also discusses the potential for hybrid scatter estimation, in which
empirical measurements of the total scatter signal in the collimator shadow may be used to augment computational single
scatter estimates and thus account for multiple scatter.
In this paper a numerical technique for microwave imaging based on a two-dimensional inverse scattering
method is proposed. This method is a non-destructive imaging technique that combines finite difference time
domain (FDTD) analysis with genetic algorithm (GA) optimization to find the dielectric properties of the
object under test. The applications of the proposed method can vary from medical imaging to nondestructive
testing of materials and structures.
In many parts of the world, breast cancer is the leading cause mortality among women and it is the major cause of cancer death, next only to lung cancer. In recent years, microwave imaging has shown its potential as an alternative approach for breast cancer detection. Although advances have improved the likelihood of developing an early detection system based on this technology, there are still limitations. One of these limitations is that target responses are often obscured by surface reflections. Contrary to ground penetrating radar applications, a simple reference subtraction cannot be easily applied to alleviate this problem due to differences in the breast skin composition between patients. A novel surface removal technique for the removal of these high intensity reflections is proposed in this paper. This paper presents an algorithm based on the multiplication of adjacent wavelet subbands in order to enhance target echoes while reducing skin reflections. In these multiscale products, target signatures can be effectively distinguished from surface reflections. A simple threshold is applied to the signal in the wavelet domain in order to eliminate the skin responses. This final signal is reconstructed to the spatial domain in order to obtain a focused image. The proposed algorithm yielded promising results when applied to real data obtained from a phantom which mimics the dielectric properties of breast, cancer and skin tissues.
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.