In-Depth Photon Counting Detector

In-Depth photon counting detectors (PCXD) use edge-on configuration and have multi-layer segmentations. In addition to reducing the count-rate burden, our preliminary results demonstrated its dose benefits on spectral imaging. However, due to various non-idealities, PCXD is still in its initial stage of being applied to clinical scanners. Our interest is to characterize the spectral distortion from the current detector prototypes, study and understand how this would limit the performance of the PCXD towards spectral imaging. There are multiple factors causing to the non-ideal energy response, which include the x-ray interaction with the material, charge carrier transport within the material, the pulse signal detection of the electrodes, etc. We would like to study the entire chain effect as a whole and solve relevant problems. CRLB Model for Correlated Measurements. One of the questions we are trying to solve is that the inter-layer cross talk introduces correlation to the signal collected from each layer, which makes the independent Poisson model no longer valid for estimating the Cramér-Rao lower bound (CRLB) of the material decomposition. We proposed that the multivariate Gaussian model is validated to be a good substitute to the Poisson model for PCXD CRLB estimation which can avoid the failure caused by correlated measurements.


Inverse Geometry CT

We propose an inverse-geometry volumetric CT system for acquiring a 15-cm volume in one rotation with negligible cone-beam artifacts. The system uses a large-area scanned source and a smaller detector array. Two feasibility investigations were conducted. The first examined data sufficiency in the transverse planes. The second predicted the signal-to-noise ratio (SNR) compared to a conventional scanner. Results showed sufficient sampling of the full volume in less than 0.5 sec and, when compared to a conventional scanner operating at 24 kW with a 0.5 sec voxel illumination time (e.g., 0.5 s gantry rotation and pitch of 1), predicted a relative SNR of 76%.


CT Perfusion Validation and Optimization

Physicians rely on CT perfusion images and quantitative parameters extracted from the scan data, including cerebral blood flow, cerebral blood volume, and mean transit time, to make diagnostic and therapeutic decisions for many stroke patients. However, these metrics vary greatly depending on the computational method used and have yet to be fully compared to a ground truth. This project aims determine that ground truth by developing a dynamic, anthropomorphic digital brain phantom upon which we will simulate CT perfusion scans. Ultimately, this information will validate scan data and increase physicians’ ability to prescribe a plan of care based on data.


Liver Imaging Protocol

Some radiologists report a preference for blended dual-energy (DE) CT images over single energy (SE) images for liver disease diagnosis at the same dose. We hypothesized that the broad spectrum of DE might be beneficial for a combination of tasks. We compare the CNR of SE and blended DE images for single and composite tasks, in part to see if they explain the preference. The CNR of pre-contrast single kVp image mostly increases as the energy steps up while 90 kVp or lower energy yields higher CNR for post-contrast image, depending on the differential iodine concentration of each tissue. Similar trends are seen in the DE blended CNR curves as to those from SE protocols. Results from a composited multi-CNR study demonstrate that the SE protocol has better performance.


Fixed K-Edge Filtration to Fast kVp-switching CT system

Dose efficiency of dual kVp spectral imaging can be improved if a filter is used to remove photons in the common part of their spectra, thereby increasing the  . While there are a number of advantages to rapid kVp-switching for dual energy, it may not be feasible to have two different filters for the two spectra. Therefore, we are interested in finding a fixed filtration method to improve the dose efficiency of kVp-switching dual energy x-ray systems. Via both simulation and experiments, we verified that k-edge filter can reduce the variance at fixed dose at the penalty of increasing the tube power output. The optimal filter material is task dependent.


Dynamic Exposure Control

The use of ionizing radiation is fundamental to x-ray CT. While this radiation provides imaging contrast which can be used to make valuable, clinical diagnoses and save lives, the side effects of this radiation could also be significant. The guiding principle for the use of radiation is ALARA - that the radiation exposure should be As Low As Reasonably Achievable.

We are researching a method for dynamic control of the x-ray exposure which can reduce the amount of radiation delivered to the patient, without negatively impacting image quality. Current methods of x-ray exposure control are generic, and are tuned to the average patient shape and for a typical clinical task. Some measurements in CT are therefore overdosed, with more radiation than necessary for good image statistics, and other measurements are underdosed. Our innovation is a dynamic x-ray filter which shapes the beam strength so that the exposure level for every ray is close to optimal.

This x-ray filter is a physical piece of hardware, acting as a pre-patient attenuator, and consists of several metallic wedges which slide into and out of the beam. Importantly, these wedges are shaped so that they introduce minimal artifacts into the reconstructed image.

Preliminary results suggest that this concept can reduce radiation exposure by 30%. The benefits are even higher for targeted scanning or when combined with photon-counting detectors, which suffer from several non-idealities when the measurements are overdosed.

The piecewise-linear attenuator has more flexibility than conventional filters, but its thickness profile is proportional to wedge position and sensitive to the precision of wedge control. Thus, we propose another approach to the Dynamic Exposure Control project, a fluid-filled dynamic filter, a two-dimensional array of small binary elements (filled or empty), that may be more reproducible due to digital control.  Focusing on minimizing the dynamic range, simulations how improvement in RMSE and a dynamic range and more homogeneous variance in the reconstruction when compared to the piecewise-linear attenuator.

This project is supported by the National Institutes of Health (grants U01 EB017140 and R21 EB015574) and the Coulter Foundation.


Limits of lterative Reconstruction

In recent years, iterative reconstruction has exploded in popularity. These methods offer the potential for higher-quality images at lower radiation exposure by applying more sophisticated, statistical methods to the problem of reconstruction. While the improvement in quality is real, the effect is sometimes understated. Improvements by iterative reconstruction cannot be measured by traditional image quality metrics such as image variance, point spread function, or noise power spectrum, because these metrics only apply to linear reconstruction. When these methods are used, it is sometimes implied that iterative reconstruction can lead to dose reductions of 80% or greater.

We have been conducting research on the theoretical limits of these methods. Under the assumption of no a priori knowledge, our preliminary results indicate that the maximum reduction in radiation exposure using iterative reconstruction, or any kind of reconstruction, is between 40% to 80% depending on the task.

While the benefits of iterative reconstruction can be clearly seen, we must always be aware that the use of nonlinear reconstruction algorithms should also be approached with caution.