MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

How Low Can CT Dose Go? Future Dose Reduction Technologies

J Maier1*, M McNitt-Gray2*, G Gang3*, F Noo4*, M Kachelriess5*, (1) German Cancer Research Center, Heidelberg, DE, (2) David Geffen School of Medicine at UCLA, Los Angeles, CA, (3) Johns Hopkins University, Baltimore, MD, (4) University Utah, Salt Lake City, UT, (5) DKFZ Heidelberg, FS05, Heidelberg, BW, DE







Presentations

(Wednesday, 7/15/2020) 10:30 AM - 12:30 PM [Eastern Time (GMT-4)]

Room: Track 1

During the last decades tremendous technical efforts to improve the dose efficiency of diagnos-tic CT systems have been seen. Among those are improved x-ray tubes, tube current modula-tion, x-ray spectrum optimization, highly integrated detectors, iterative image reconstruction, automatic exposure control and patient- or even organ-specific technologies.

Improved dose efficiency can be used to improve CT image quality or to reduce patient dose. This session discusses possible future technologies that may be used to increase the dose effi-ciency in CT. Among those are risk-specific exposure control techniques, adaptive bow tie filters and methods for improved spectral shaping. Techniques that are widely discussed elsewhere at this conference, such as photon counting detectors and deep learning-based image reconstruc-tion are not part of this session.

Options for Automatic Exposure Control (AEC):
AEC in CT is performing an automatic tube cur-rent setting as well as a modulation of the tube current during the gantry rotation (xy-direction) as well as along z. Compared to a CT scan without tube current modulation, dose reduction val-ues of up to 60% have been realized in clinical routine. Further improvements can be expected from recent developments in the field of deep learning. One promising approach is the so-called deep dose estimation (DDE) that trains a deep convolutional neural network to predict patient-specific dose estimates in real-time. This opens new options for AEC such as optimizing the tube current level and its modulation curve to minimize the the patient‘s radiation risk rather than minimizing the mAs product. We will provide the corresponding basics of AEC and DDE, and will discuss the additional dose reduction that can be expected from a DDE-based tube current modulation, which is expected to be in the order of 10%.

Optimization of X-Ray Spectra:
Some manufacturers have added patient-specific variable thick-ness x-ray bem filtration for a few reasons. One is help tune the x-ray spectra and in particular for spectral separation in the context of dual or multi-energy scanning. Another is to reduce x-ray photon fluence and dose to the patient. In this latter instance, radiation dose per tube cur-rent is reduced, which enables tube current modulation schemes to avoid truncation at the minimum tube current and, thus, maintain effective modulation. A third reason is that such fil-ters predominantly filter out the low energy photons whose contribution to patient dose is high while their contribution to the CT image is low. These strategies may lead to dose reductions of 20 to 40% if being widely implemented. This section will also discuss situations where addition-al filtration may not be optimal in terms of imaging tasks.

Dynamic Fluence Field Modulation:
Static bowties do not adequately account for the changing patient cross-sections as a function of rotation angles and may incur dose penalties when the patient is miscentered. To address these limitations, dynamic fluence field modulation (FFM) has received significant interest. This talk will review current state-of-the-art FFM using static bow-tie, recent development of dynamic beam modulators, and theoretical FFM designs for different reconstruction methods and image quality objectives. Dynamic FFM, in combination with opti-mized image reconstruction methods, may have the potential to further reduce CT dose by 50%.

Are the Dose Reduction Claims Justified?
Proper assessment of image quality is critical to justify dose reduction claims. We will review the many approaches that can be used for this assess-ment, from using simple quantities such as the structure similarity index metric to task-based concepts involving the receiver-operating characteristic curve. Implementation aspects will be discussed as well as the strengths, limitations and pitfalls of each approach. Undoubtedly, the advent of deep learning solutions for image reconstruction is further complicating the problem. We will touch on this issue and ways in which deep learning may help with image quality as-sessment.

Summary:
The estimated future dose reduction potential seen here is 10% (risk-specific AEC), 20% or more (optimized spectra), and 50% (dynamic bowtie). Another 30% and more dose re-duction can be expected from the introduction of photon counting CT detectors [Willemink et al., Radiology 289:293–312, 2018][Pourmorteza et al., Invest Radiol 53:365–372, 2018][Klein et al., Invest. Radiol. 55(2):111-119, 2020]. The introduction of machine learning algorithms may further reduce dose compared to today‘s iterative reconstruction. Since this value can hardly be quantified with current data, we cautiously assume that the reduction amounts to only 10%. Optimistically assuming all technologies would be technically realized and became widely avail-able and that the anticipated dose reduction techniques would act independently then a total dose reduction of 1-(1-0.1)(1-0.2)(1-0.5)(1-0.3)(1-0.1) = 77% may be seen in the future. The truth is likely to be lower but yet quite significant.

Learning Objectives:
1. Learn about real-time patient dose assessment using deep learning
2. Understand the basics and potential future of automatic exposure control
3. See how prefiltration is used in single and dual energy CT and if there are benefits of thicker and patient-specific prefilters
4. Find out about today's and future techniques of beam modulators and their potential for dose reduction
5. Learn how to assess image quality and how to validate dose reduction claims

Funding Support, Disclosures, and Conflict of Interest: Michael McNitt-Gray: Recipient of research support from Siemens Healthineers; Member, Scientific Advisory Board, Hura Imaging, Inc.;UCLA Department of Radiology has a Master Research Agreement with Siemens Healthineers Frederic Noo: Consultant for nView Medical Inc., Salt Lake City, Utah; Research collaborations with Siemens Healthineers through research agreement with University of Utah

Handouts

Keywords

Dose, CT

Taxonomy

IM- CT: Development (New technology and techniques)

Contact Email