Meeting Banner
Abstract #3856

Functional Connectome Fingerprinting using Recurrent Neural Networks Does Not Depend on the Temporal Structure of the Data

Gokce Sarar1, Shili Wang2, Jiawei Ren3, and Thomas T. Liu1

1UCSD Center for Functional MRI, La Jolla, CA, United States, 2Beihang University, Beijing, China, 3Canyon Crest Academy, San Diego, CA, United States

Functional connectome fingerprinting can identify individuals with high accuracy with long duration scans (> 6 min) but the performance significantly degrades with shorter (72s) duration scans. It has been argued that Recurrent Neural Networks (RNN) can achieve high identification performance with short duration data by taking advantage of temporal information in the fMRI signals. We test this claim by permuting the temporal ordering of the data. We show that the RNN can achieve high accuracy for short duration data even when the temporal structure is destroyed, suggesting that the RNN performance depends primarily on the spatial correlation of the data.

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

Join Here