Cerebrovascular disease remains a common cause of death and major disability in the United States, and identifying and preventing strokes should be a high priority. Direct measurement of regional cerebral blood flow (CBF) is challeng- ing in these patients, since we do not have a non-invasive, radiation-free imaging method that has been appropriately validated against gold standard techniques. This is important, because there is compelling evidence that measuring the CBF change before and after a stress test meant to increase CBF (a measurement known as cerebrovascular reserve [CVR]) can identify patients at increased stroke risk. Stress tests have been a mainstay of the diagnostic workup of cardiology patients for many years, and we believe strongly that their use will benefit cerebrovascular disease patients as well. The goal of this project is to improve the quality of arterial spin label (ASL) MRI using deep learning, a powerful form of machine learning, that is currently undergoing tremendous progress. We will then to apply this in a prospective, adaptive validation trial against oxygen-15 water PET CBF, using simultaneous PET/MRI to minimize biological variability. Finally, we will apply this improv- ed tool to study the effects of gender on CVR and its reproducibility. Successful completion of this study will result in a validated methodology to assess CVR in cerebrovascular disease patients without the use of radiation or contrast. As such, it will provide solid, evidence-based recommendations for clinicians developing new paradigms and interventions in patients with impaired CVR.

Public Health Relevance

There is strong evidence that imaging of cerebrovascular reserve (CVR), the ability to increase cerebral blood flow (CBF) in response to a challenge, can identify patients at increased risk of stroke. Therefore, measuring CVR would be extremely useful for designing clinical trials of interventions to mitigate this risk. However, current methods to measure CBF and CVR are suboptimal, and do not work well in patients with cerebrovascular disease. The goals of this project are ? to improve non-contrast, radiation-free arterial spin label MRI methods using deep learning, a powerful form of artificial intelligence that has shown tremendous progress for computer vision ? to validate these methods against a CBF gold-standard, oxygen-15 water PET, using simultaneous PET/MRI, using an adaptive ?play-the-winner? strategy ? to apply them to assess gender differences in CVR and test their reproducibility, with the goal of establishing age and gender normative ranges to better identify outliers.

National Institute of Health (NIH)
National Institute of Biomedical Imaging and Bioengineering (NIBIB)
Research Project (R01)
Project #
Application #
Study Section
Biomedical Imaging Technology Study Section (BMIT)
Program Officer
Liu, Guoying
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Stanford University
Schools of Medicine
United States
Zip Code