Members

Center for Advanced Functional Neuroimaging

Greg Zaharchuk, MD, PhD

Professor of Radiology, Neuroradiology

gregz@stanford.edu

1201 Welch Road, PS04, Stanford, CA 94305

(650) 736-6172 (office)

Stanford profile

Michael Moseley, PhD

Professor of Radiology

moseley@stanford.edu

1201 Welch Road, PS060, Stanford, CA 94305

(650) 725-6077 (office)

Stanford profile

Kevin Chen

ktchen@stanford.edu

We are working on producing diagnostic-level actual ultra-low-dose amyloid PET/MRI acquisitions with CNN enhancement. Preliminary results have shown that the ultra-low-dose CNN, pre-trained from simulated ultra-low-dose data, was able to produce qualitatively similar images from the ultra-low-dose images compared to their full-dose counterparts.

 

Sanaz Nazari Farsani

sanafa@stanford.edu

The success of ischemic stroke treatment highly depends on the time interval between the stroke onset and the treatment. Accordingly, stroke patients need to be triaged fast for which accurate estimation of the final lesion volume is vital. Diffusion-perfusion mismatch measurement is a common practice for the estimation of the final infarct volume. However, PWI is time-consuming, expensive and injection of the contrast agent might cause complications for some patients. We aim to predict the final infarct volume from only baseline DWI using an attention-gated U-shaped network. Our AG U-net was able to predict the final infarct volumes with comparable performance to the models which consider both DWI and PWIs. We believe this is a valuable finding, since excluding PW imaging in stroke patients could save time and money, eliminate complications related to the contrast agent injection, and accelerate patients triage. 

 

Jiahong Ouyang

jiahongo@stanford.edu

PET imaging leads the subjects in radiation exposure and is expensive and not available in the majority of hospitals in the world. Can we generate PET from widely available and harmless MRI? We built a Transformer-UNet based model with spatial-wise and channel-wise attention modules to synthesize brain tumor patients’ FDG-PET based on their multi-contrast MR images. Our method achieved high-resolution high-quality, and pathological accurate PET images. The model can also get satisfying performance on subjects with dementia and epilepsy suggesting the model’s generalizability.

Moss Zhao

mosszhao@stanford.edu

Cerebrovascular reserve (CVR) reflects the change in cerebral blood flow in response to a vasoactive stimulus. A number of studies have demonstrated the potential of CVR as an important biomarker to detect cerebral vascular diseases. My project focuses on developing non-invasive MRI techniques to quantify CVR in vivo. Currently, I am comparing CVR measurement using simultaneous PET/MRI and investigating the reproducibility of quantifying CVR using arterial spin labeling (ASL) MRI. The impact of this research includes creating a novel imaging technique to assess the health of brain tissues.

Ramy Hussein

ramyh@stanford.edu

I am interested in developing and optimizing Artificial Intelligence solutions for the early diagnosis and prediction of cerebrovascular and neurodegenrative diseases, with more focus on Ischemic Stroke and Alzheimer's Disease. My research focuses on medical Image-to-Image translation, where deep 3D encoder-decoder neural networks are used to synthesize Positron Emission Tomography (PET) scans from structural and perfusion Magnetic Resonance Imaging (MRI). Adequate quantification of PET from MRI has a great potential for increasing the accessibility of cerebrovascular diseases assessment for underserved populations, underprivileged communities, and developing nations.

Jui Khankari

juik@stanford.edu

Digital subtraction angiography (DSA) is the gold-standard method of assessing arterial blood flow and blockages in endovascular thrombectomy—a common treatment for ischemic stroke. Precisely and quickly identifying the location of vascular occlusions on DSA scans can be difficult due to the tortuosity of the vessels and their overlap in the imaging plane.  My research aims to build convolutional neural networks (CNNs) that can automatically detect vascular occlusions and other anatomical features from DSA scans. Automating the process of identifying occlusions may be helpful for more rapid decision-making during endovascular procedures, potentially preventing death or disability in over 77 million ischemic stroke patients around the world.

 

Yannan Yu

yannanyu@stanford.edu

Salvageable tissue is an important imaging marker to make treatment decisions for acute ischemic stroke. We aim to improve the current clinical method (thresholding) of identifying salvageable tissue on baseline MRI using deep learning. The salvageable tissue is shown as the difference between Maximum and Minimum lesions. The U-net model yielded better predictions than the current clinical method.

 

Yongkai Liu

yongkliu@stanford.edu

Timely and precise outcome prediction plays an important role in guiding treatment decision making and patient selection for endovascular therapy in the emergency department. I am fusing artificial intelligence-enabled MRI model with common stroke biomarkers to predict patient recovery outcome at 90 days.

 

Sophie Ostmeier

sostm@stanford.edu

My research interests in the Heit Lab are focused on stroke imaging and developing artificial intelligence techniques for the evaluation of cerebral ischemia and hemorrhagic stroke. 

Bin Jiang

binj@stanford.edu

We are building up a pipeline for this specific dataset and use advanced neuroimaging analyzing methodology to achieve not only patient screening and complications tracking but also recovery monitoring and outcome prediction.

 

Ates Fettahoglu

atesfet@stanford.edu

I'm interested in quantitative perfusion imaging using alternative PET tracers, enabled by early-phase image acquisition. I also focus on different CBF measurement modalities and their implementation in various applications including its relation to cognitive decline in Alzheimer's Disease and longitudinal CBF changes after neuronal stem cell injection in stroke patients.

 

Amirhossein Sanaat

asanaat@stanford.edu

I am using AI and different contrasts of fast MR scans to synthesize MR angiograms which is a map of arteries that help to evaluate them for stenosis, occlusions, aneurysms, or other abnormalities.

Alumni

Name Current Position
Samantha Holdsworth Senior Lecturer, Department of Anatomy and Medical Imaging, University of Aukland, Aukland, New Zealand
Hessam Jahanian Assistant Professor, Department of Radiology, University of Washington, WA, USA 
Audrey Fan Assistant Professor, Department of Biomedical Engineering, University of California at Davis, CA, USA
Wendy Ni  
Zungho (Wesley) Zun Assistant Professor, Children's National Medical Center, Department of Pediatrics, George Washington University, DC, USA
Deqiang Qiu 
Assistant Professor, Department of Radiology and Imaging Sciences, Emory University, GA, USA 
Jalal Andre 
Assistant Professor, Department of Neuroradiology, University of Washington, WA, USA 
Chunlei Liu Assistant Professor, Department of Radiology, Duke University, NC, USA
Jason Hsu 
Assistant Professor, Department of Radiology, UCSF School of Medicine, CA, USA
Ryan Spilker Senior Scientist, HeartFlow Inc., CA, USA
Nanjie Gong
PhD Student, University of Hong Kong, Hong Kong, China
Georges Hankov 
PhD Student, ETH, Zurich, Switzerland
Caleb Folkes
Postgraduate Navy training, USA
Salil Soman
Instructor in Radiology, Beth Israel Hospital, Harvard Medical School, Boston, MA, USA
Yannan Yu Resident Physician, UCSF Radiology, CA, USA 

Old Group Photos

Can you find the members who were in conference meetings when this photo was taken?

CAFN Group 2013

 

CAFN Group 2012