Meeting Banner
Abstract #0310

Improving Perfusion Image Quality and Quantification Accuracy Using Multi-contrast MRI and Deep Convolutional Neural Networks

Jia Guo1, Enhao Gong2, Maged Goubran1, Audrey P. Fan1, Mohammad M. Khalighi3, and Greg Zaharchuk1

1Radiology, Stanford University, Stanford, CA, United States, 2Electrical Engineering, Stanford University, Stanford, CA, United States, 3Global Applied Science Lab, GE Healthcare, Menlo Park, CA, United States

We propose a novel method that uses deep convolutional neural networks (dCNNs) to combine multiple contrasts from MRI, including single- and multi-delay pseudo-continuous arterial spin labeling (PCASL) and structural scans, to synthesize perfusion maps that approach the accuracy of the PET perfusion measurements. The dCNN was trained and tested on both healthy and patient datasets, and demonstrated significant improvement on both image quality (higher structural similarity and lower normalized root mean square error) and quantification accuracy (regional CBF comparable with PET) than either ASL method alone. This method may potentially be generalized to other qualitative/quantitative applications.

This abstract and the presentation materials are available to members only; a login is required.

Join Here