Test–retest reliability of spatial patterns from resting-state functional MRI using the restricted Boltzmann machine and hierarchically organized spatial patterns from the deep belief network

Autoencoder and restricted Boltzmann machine for transfer learning in functional magnetic resonance imaging task classification

Deep neural networks (DNNs) have been adopted widely as classifiers for functional magnetic resonance imaging (fMRI) data, advancing beyond traditional machine learning models. Consequently, transfer learning of the pre-trained DNN becomes crucial to enhance DNN classification performance, specifically by alleviating an overfitting issue that occurs when a substantial number of DNN parameters are fitted to a relatively small number of fMRI samples. In this study, we first systematically compared the two most popularly used, unsupervised pretraining models for resting-state fMRI (rfMRI) volume data to pre-train the DNNs, namely autoencoder (AE) and restricted Boltzmann machine (RBM). The group in-brain mask used when training AE and RBM displayed a sizable overlap ratio with Yeo's seven functional brain networks (FNs). The parcellated FNs obtained from the RBM were fine-grained compared to those from the AE. The pre-trained AE and RBM served as the weight parameters of the first of the two hidden DNN layers, and the DNN fulfilled the task classifier role for fMRI (tfMRI) data in the Human Connectome Project (HCP). We tested two transfer learning schemes: (1) fixing and (2) fine-tuning the DNN's pre-trained AE or RBM weights. The DNN with transfer learning was compared to a baseline DNN, trained using random initial weights. Overall, DNN classification performance from the transfer learning proved superior when the pre-trained RBM weights were fixed and when the pre-trained AE weights were fine-tuned (average error rates: 14.8% for fixed RBM, 15.1% fine-tuned AE, and 15.5% for the baseline model) compared to the alternative scenarios of DNN transfer learning schemes. Moreover, the optimal transfer learning scheme between the fixed RBM and fine-tuned AE varied according to seven task conditions in the HCP. Nonetheless, the computational load reduced substantially for the fixed-weight-based transfer learning compared to the fine-tuning-based transfer learning (e.g., the number of weight parameters for the fixed-weight-based DNN model reduced to 1.9% compared with a baseline/fine-tuned DNN model). Our findings suggest that weight initialization at the DNN's first layer using RBM-based pre-trained weights provides the most promising approach when the whole-brain fMRI volume supports associated task classification. We believe that our proposed scheme could be applied to a variety of task conditions to improve their classification performance and to utilize computational resources efficiently using our AE/RBM-based pre-trained weights compared to random initial weights for DNN training.

Deep Learning-based Classification of Resting-state fMRI Independent-component Analysis

Functional connectivity analyses of fMRI data have shown that the activity of the brain at rest is spatially organized into resting-state networks (RSNs). RSNs appear as groups of anatomically distant but functionally tightly connected brain regions. Inter-RSN intrinsic connectivity analyses may provide an optimal spatial level of integration to analyze the variability of the functional connectome. Here we propose a deep learning approach to enable the automated classification of individual independent-component (IC) decompositions into a set of predefined RSNs. Two databases were used in this work, BIL&GIN and MRi-Share, with 427 and 1811 participants, respectively. We trained a multilayer perceptron (MLP) to classify each IC as one of 45 RSNs, using the IC classification of 282 participants in BIL&GIN for training and a 5-dimensional parameter grid search for hyperparameter optimization. It reached an accuracy of 92 %. Predictions for the remaining individuals in BIL&GIN were tested against the original classification and demonstrated good spatial overlap between the cortical RSNs. As a first application, we created an RSN atlas based on MRi-Share. This atlas defined a brain parcellation in 29 RSNs covering 96 % of the gray matter. Second, we proposed an individual-based analysis of the subdivision of the default-mode network into 4 networks. Minimal overlap between RSNs was found except in the angular gyrus and potentially in the precuneus. We thus provide the community with an individual IC classifier that can be used to analyze one dataset or to statistically compare different datasets for RSN spatial definitions.