MEG Sensor Selection for Neural Speech Decoding

Abstract

Direct decoding of speech from the brain is a faster alternative to current electroencephalography (EEG) speller-based brain-computer interfaces (BCI) in providing communication assistance to locked-in patients. Magnetoencephalography (MEG) has recently shown great potential as a non-invasive neuroimaging modality for neural speech decoding, owing in part to its spatial selectivity over other high-temporal resolution devices. Standard MEG systems have a large number of cryogenically cooled channels/sensors (200 − 300) encapsulated within a fixed liquid helium dewar, precluding their use as wearable BCI devices. Fortunately, recently developed optically pumped magnetometers (OPM) do not require cryogens, and have the potential to be wearable and movable making them more suitable for BCI applications. This design is also modular allowing for customized montages to include only the sensors necessary for a particular task. As the number of sensors bears a heavy influence on the cost, size, and weight of MEG systems, minimizing the number of sensors is critical for designing practical MEG-based BCIs in the future. In this study, we sought to identify an optimal set of MEG channels to decode imagined and spoken phrases from the MEG signals. Using a forward selection algorithm with a support vector machine classifier we found that nine optimally located MEG gradiometers provided higher decoding accuracy compared to using all channels. Additionally, the forward selection algorithm achieved similar performance to dimensionality reduction using a stacked-sparse-autoencoder. Analysis of spatial dynamics of speech decoding suggested that both left and right hemisphere sensors contribute to speech decoding. Sensors approximately located near Broca's area were found to be commonly contributing among the higher-ranked sensors across all subjects. INDEX TERMS autoencoder, brain-computer interface, forward selection algorithm, magnetoencephalography, neural speech decoding, OPM, SVM

Imagined speech can be decoded from low- and cross-frequency intracranial EEG features

“…We also found that the beta-band stood out as a prominent feature in the neural encoding of imagined speech, both in terms of power and CFC (low-beta/gamma and low-beta/BHA), with slightly better performance when using power. This finding aligns well with the notion that the beta band plays an important role in endogenous processes, notably in relation with top-down control in the language domain 16 , 17 , 40 , 43 , 48 , 49 . Although repeating a heard or written word engages automatic, almost reflex, neural routines, imagined speech is a more voluntary action requiring enhanced endogenous control from action planning frontal regions 50 – 52 .…”

“…Imagined speech decoding with non-invasive techniques, i.e. surface electroencephalography (EEG) or magnetoencephalography (MEG), has so far not led to convincing results, despite recent encouraging developments (vowels and words decoded with up to ~70% accuracy for a three-class imagined speech task) [12][13][14][15][16][17] . The most effective approach so far to advance toward a real "imagined speech" decoding system is based on electrocorticographic (ECoG) signals, which are currently only recorded in patients with refractory epilepsy undergoing presurgical evaluation.…”

Decoding Imagined Speech Based on Deep Metric Learning for Intuitive BCI Communication

“…In Dash et al [29], they proposed the forward selection algorithm using spatial selectivity of MEG signals. The aim of this study is to minimize the number of sensors for neural speech decoding.…”

“…Furthermore, Dash et al [32] this is limited to practical use in real BCIs due to their high price, large size and weight [29]. Cooney et al [33] used independent component analysis with Hessian approximation preconditioning to eliminate electrooculography signals.…”

Decoding Covert Speech From EEG-A Comprehensive Review

“…The distribution of the modalities used for decoding imagined speech in these papers is given in Figure 1 . These modalities include EEG, ECoG (Herff et al, 2015 , 2016 ), fMRI (Yoo et al, 2004 ; Abe et al, 2011 ), fNIRS (Herff et al, 2012 ; Kamavuako et al, 2018 ; Sereshkeh et al, 2018 ), MEG (Destoky et al, 2019 ; Dash et al, 2020 ), ICE (Brumberg et al, 2011 ; Kennedy et al, 2017 ; Wilson et al, 2020 ) etc. Clearly, EEG is the most popular modality used for decoding imagined speech with 18 articles using it for capturing the neural changes during imagined speech.…”