| Paper: | PS-2A.60 | ||
| Session: | Poster Session 2A | ||
| Location: | H Lichthof | ||
| Session Time: | Sunday, September 15, 17:15 - 20:15 | ||
| Presentation Time: | Sunday, September 15, 17:15 - 20:15 | ||
| Presentation: | Poster | ||
| Publication: | 2019 Conference on Cognitive Computational Neuroscience, 13-16 September 2019, Berlin, Germany | ||
| Paper Title: | DeepLight: A Structured Framework For The Analysis of Neuroimaging Data Through Recurrent Deep Learning Models | ||
| Manuscript: | Click here to view manuscript | ||
| License: |  This work is licensed under a Creative Commons Attribution 3.0 Unported License. |
||
| DOI: | https://doi.org/10.32470/CCN.2019.1226-0 | ||
| Authors: | Armin Thomas, Technische Universität Berlin, Germany; Hauke R. Heekeren, Freie Universität Berlin, Germany; Klaus-Robert Müller, Technische Universität Berlin, Germany; Wojciech Samek, Fraunhofer Heinrich Hertz Institute, Germany | ||
| Abstract: | We propose a structured framework for the application of recurrent deep learning (DL) models to the analysis of fMRI data. To identify an association between cognitive state and brain activity, DeepLight utilizes the layer-wise relevance propagation method. Thereby, decomposing the DL model’s decoding decisions into the contributions of the single input voxels to these decisions. Importantly, DeepLight is able to identify this association on multiple levels of data granularity, from the level of the group down to single subjects, trials and time points. | ||