Page History

1

Radiation Dose Reduction in CT and its Effects on Quantitative Imaging and Machine Learning/Artificial Intelligence Algorithms

(Dr. Michael McNitt-Gray)

CT is widely used in clinical practice with applications ranging from early detection (screening), characterization (diagnosis) and assessment of response to therapy. There has been widespread concern over the radiation dose associated with these scans, especially on pediatric patients or patients who get frequent scans. There have been significant developments which allow the reduction of radiation dose from CT including developments in automatic exposure control, advanced image reconstruction techniques , more efficient detector technologies among others that promise significant radiation dose reductions to patients, while maintaining clinical image quality.  While these technologies are exquisite and should be investigated wherever possible in a clinical environment, their effects on quantitative measures extracted from CT images and machine learning algorithms have not been well characterized. These technologies may affect image quality in ways that may limit the generalizability of quantitative imaging and Artificial Intelligence/Machine Learning (AI/ML) methods.  For example, advanced image reconstruction methods may be able to mitigate the increase in noise that is incurred when radiation dose is reduced, but there may be some impact on image resolution. In addition, many of these techniques are non-linear and adaptive to the local image anatomy and pathology, their impacts may be difficult to predict from application to application and even patient to patient (and yes, even within a patient). This presentation will provide examples of these effects and discuss possible methods to mitigate these effects, which hopefully will enable more generalizable deployment of quantitative imaging methods and AI/ML algorithms. 

2

IDC Update

Ulrike Wagner, FNLCR

3

TCIA Update

Justin Kirby, FNLCR