Meeting Banner
Abstract #0802

Estimating perfusion and permeability using neural network with training data generated from vessel construction and transport simulation

Qihao Zhang1, Dominick Romano2, Thanh Nguyen2, Pascal Spincemaille3, and Yi Wang3
1Cornell University, Ithaca, NY, NY, United States, 2Cornell University, New York, NY, United States, 3Weill Cornell Medical College, New York, NY, United States


We propose to estimate perfusion parameters (perfusion $$$F$$$, permeability $$$K^{trans}$$$, vascular space volume $$$V_p$$$ and extravascular extracellular volume $$$V_e$$$) from contrast enhanced MRI using Quantitative Transport and Exchange network (QTEnet), a deep learning method that does not require an arterial input function. Training data were generated by solving the transport equation in simulated high-resolution vasculature and computing the corresponding 4D tracer propagation. A 3D U-net was trained to reconstruct perfusion parameters from the tracer propagation images. Tracer propagation simulated in experimentally obtained tumor vasculature was used for valiation, and the method was then applied to glioma DCE MRI data.

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

Join Here