PROJECT SUMMARY
Quantitative dynamic contrast enhancement (DCE) MRI metrics such as tissue perfusion rates, kinetic
parameters, extravascular volume, and plasma volume allow characterization of subtle differences in tissue
states related to ischemia, vascularity, inflammation, and fibrosis in neurological and cardiovascular diseases,
and in cancers such as pancreatic adenocardinoma (PDAC). The reproducible nature of quantitative imaging
makes it more suitable for multi-center or longitudinal studies than conventional “qualitative” imaging.
Quantitative DCE metrics have been shown to be important for risk assessment, early detection, staging,
characterization, and treatment monitoring of PDAC and other diseases.
DCE MRI performs imaging before, during, and after injection of a gadolinium (Gd)-based contrast agent.
There are several major challenges, especially in moving organs: i) cardiac motion must be dealt with for heart
scans, generally by syncing acquisition with an ECG signal, leading to difficulty in arrhythmia patients as well
as low imaging efficiency and challenges for whole-heart 3D imaging; ii) respiratory motion must be dealt with,
typically by patient breath-holding; iii) safety questions surrounding Gd contrast agents lower the benefit-to-risk
ratio in many situations.
The objective of this project is to develop low-dose, motion-resolved, quantitative dynamic contrast
enhanced (DCE) MRI for PDAC. This will be accomplished by developing and validating the MR multitasking
framework for multi-dynamic, highly time-resolved T1 mapping, and correlating MRI measurements to histology
in patients undergoing surgical resection. Multitasking designs DCE MRI around the concept of images as
functions of multiple time dimensions, each corresponding to a different dynamic process (e.g., motion, T1,
DCE). It integrates machine learning, low-rank tensor modeling, compressed sensing, and deep learning to
extract reproducible, quantitative measurements from 6D DCE images, even under free-breathing conditions.
The resulting technology would be a powerful tool for quantitative MRI tissue characterization in moving
organs, and would be promising as a tool for monitoring neoadjuvant therapies prior to pancreatic resection.