Identifying Brain Networks at Multiple Time Scales via Deep Recurrent Neural Network

Supervised Brain Network Learning Based on Deep Recurrent Neural Networks

“…In [32], a RNN based brain state recognition model is proposed to effectively separate different brain state. In [33], we proposed a deep recurrent neural network (DRNN) model to identify group-wise task-related functional brain networks and brain response patterns in tfMRI data. Although this group-wise activation detection framework is limited in characterizing individual specific brain networks, the identified brain response patterns are diverse and meaningful [33].…”

“…In [33], we proposed a deep recurrent neural network (DRNN) model to identify group-wise task-related functional brain networks and brain response patterns in tfMRI data. Although this group-wise activation detection framework is limited in characterizing individual specific brain networks, the identified brain response patterns are diverse and meaningful [33]. These response patterns naturally account for the temporal dependencies of input stimulus and reflect the possible variations of brain response to input stimulus.…”

“…The basic idea of SUDRNN framework is combing the superior ability of RNN to characterize brain response patterns and the ability of supervised learning in brain network reconstruction which largely overcomes limitations in either model-driven or data-driven brain network models. Specifically, we utilized our recently developed deep recurrent neural network model [33] to learn adaptable regressors from real tfMRI data and employed supervised dictionary learning method to identify both task-related and spontaneous brain networks simultaneously. It should be noticed that, although preliminary work [33] demonstrates its ability in data-drivenly estimating group-wise task-related functional brain networks and brain response patterns, DRNN-derived response patterns has never been adopted as regressors for model-driven method.…”

“…Specifically, we utilized our recently developed deep recurrent neural network model [33] to learn adaptable regressors from real tfMRI data and employed supervised dictionary learning method to identify both task-related and spontaneous brain networks simultaneously. It should be noticed that, although preliminary work [33] demonstrates its ability in data-drivenly estimating group-wise task-related functional brain networks and brain response patterns, DRNN-derived response patterns has never been adopted as regressors for model-driven method. An important characteristic of these DRNN-derived regressors is that they are learned from real tfMRI data and well adapt to the specific brain response variations.…”

Upper Esophageal Sphincter Opening Segmentation With Convolutional Recurrent Neural Networks in High Resolution Cervical Auscultation

“…Recurrent neural networks (RNNs) have been extensively employed for the time series modeling in the recent years, due to their capability of carrying information from arbitrarily long contexts, selective information transfer across time steps, and affordable scalability when compared to stochastic models and feed-forward networks [41]- [44]. RNNs are seemingly efficient in modeling temporal contexts in time series data and have been recently used to perform many biomedical prediction tasks of various complexities [45]- [48], but nevertheless using RNNs on raw signals is extremely hard to optimize because of the propagation of error signals through huge number of time steps [49], [50]. To overcome this, convolutional neural networks (CNNs) have been utilized for the perception of short contexts and more abstraction before feeding into RNNs for the perception of longer temporal contexts [49].…”