Increasing clinical utilization of computed tomography (CT) has created concerns about radiation exposure to the general population and risks to individual patients, and multiple approaches (hardware, algorithms, and patient protocols) have been proposed to potentially lower CT radiation dose The ultimate limit in dose reduction is determined by the loss of diagnostic information due to the attendant increase in noise. Currently this tradeoff between risk (radiation exposure) and benefit (diagnostic utility) is poorl understood, especially on the image quality side. One would like to simply plot image quality as a function of noise, but in practice neither parameter is available. Unfortunately, one cannot evaluate new technologies or optimize protocols to be as low as reasonably achievable (ALARA) without an objective metric. We propose to develop a methodology to overcome this critical barrier by (1) measuring in situ noise in a CT image and (2) efficiently measuring the concordance between an observer's diagnosis of a conventional dose image and one acquired at a reduced dose. The enabling technology for our approach is a CT dose reduction simulator, which creates realistic reduced dose images by adding synthetic noise to sinogram data and performing a reconstruction. With this tool, we can rapidly reconstruct an ensemble of images and estimate noise statistics at any point in the image. Secondly, we can prepare an image for a specific diagnostic task at an arbitrary noise level and compare the diagnosis of that image to one made with the corresponding conventional dose image. Agreement in diagnosis between the two images will decrease with increasing noise- we wish to find the noise level that corresponds to a designated decrease in agreement. Performing measurements over the continuous range of all possible noise levels is prohibitively slow- instead, we will use an adaptive algorithm to interactively select noise levels that will efficiently parameterize observer agreement with a minimum number of samples. We will apply this method to three stress case clinical tasks and characterize its capabilities. Finally, we will validate the predictions of the metric by performing a traditional observer performance study at three dose levels. The innovation in this proposal is the novel approach to obtain the two necessary components needed for ALARA image quality (local noise and observer agreement) by extending CT dose reduction simulators to create ensemble statistics and to adaptively select sampling points along the noise continuum. This concordance metric will quantify the impact of noise on observers, allowing the objective evaluation of new technologies and the selection of optimum ALARA protocols for clinical practice, thereby improving patient healthcare.

Public Health Relevance

With present techniques it is not possible to objectively measure the impact of noise on the diagnostic accuracy of computed tomography (CT) images, which critically limits the ability to determine the possible radiation dose reduction that can be achieved from new technologies or clinical protocols. This proposal will develop a practical, efficient means to quantify image quality metrics, allowing the development of CT procedures that provide diagnostic information with doses that are as low as reasonably achievable (ALARA).

Agency
National Institute of Health (NIH)
Institute
National Institute of Biomedical Imaging and Bioengineering (NIBIB)
Type
Research Project (R01)
Project #
5R01EB019135-04
Application #
9486923
Study Section
Biomedical Imaging Technology Study Section (BMIT)
Program Officer
Shabestari, Behrouz
Project Start
2015-09-20
Project End
2019-05-31
Budget Start
2018-06-01
Budget End
2019-05-31
Support Year
4
Fiscal Year
2018
Total Cost
Indirect Cost
Name
University of Pittsburgh
Department
Radiation-Diagnostic/Oncology
Type
Schools of Medicine
DUNS #
004514360
City
Pittsburgh
State
PA
Country
United States
Zip Code
15213
Mitra, Ayan; Politte, David G; Whiting, Bruce R et al. (2017) Multi-GPU Acceleration of Branchless Distance Driven Projection and Backprojection for Clinical Helical CT. J Imaging Sci Technol 61: