A hybrid autoencoder framework of dimensionality reduction for brain-computer interface decoding. (September 2022)

• Record Type: Journal Article
• Authors: Ran, Xingchen; Chen, Weidong; Yvert, Blaise; Zhang, Shaomin
• Is Part Of: Computers in biology and medicine. Volume 148(2022)
• Publication Date: 2022-09
• Subjects: Dimensionality reduction -- Autoencoder -- Brain-computer interface -- Neural decoding; ANOVA analysis of variance -- BCI brain-computer interface -- CC correlation coefficient -- CLAE CNN and LSTM based stacked autoencoder -- CNN convolutional neural network -- ECoG electrocorticography -- FC fully-connected -- GRNN general regression neural network -- KF Kalman filter -- KPCA kernel principal component analysis -- LFADS latent factor analysis via dynamical systems -- LFP local field potential -- LSTM long short-term memory -- MLR multivariable linear regression -- MSE mean squared error -- PCA principal component analysis -- PLS partial least square -- PSID preferential subspace identification -- RNN recurrent neural network -- RBF radial basis function -- SDBI soft Davies-Bouldin index; Medicine -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; 610.285
• DOI: 10.1016/j.compbiomed.2022.105871
• Languages: English
• ISSNs: 0010-4825

Abstract

Abstract: Objective: As the scale of neural recording increases, Brain-computer interfaces (BCIs) are restrained by high-dimensional neural features, so dimensionality reduction is required as a preprocess of neural features. In this context, we propose a novel framework based on deep learning to reduce the dimensionality of neural features that are typically extracted from electrocorticography (ECoG) or local field potential (LFP). Approach: A high-performance autoencoder was implemented by chaining convolutional layers to deal with spatial and frequency dimensions with bottleneck long short-term memory (LSTM) layers to deal with the temporal dimension of the features. Furthermore, this autoencoder is combined with a fully connected layer to regularize the training. Main results: By applying the proposed method to two different datasets, we found that this dimensionality reduction method largely outperforms kernel principal component analysis (KPCA), partial least square (PLS), preferential subspace identification (PSID), and latent factor analysis via dynamical systems (LFADS). Besides, the new features obtained by our method can be applied to various BCI decoders, without significant differences in decoding performance. Significance: A novel method is proposed as a reliable tool for efficient dimensionality reduction of neural signals. Its high performance and robustness are promising to enhance the decoding accuracy and long-term stability of online BCI systems based on large-scale neural recordings. Highlights: Growth in scale of neural recording poses challenges to neural data analysis. Dimensionality reduction affect directly to brain-computer interface performance. Behavioral-related latent representations extract from high-dimensional neural data. Hybrid autoencoder with condition module.