Artifacts in EEG-Based BCI Therapies: Friend or Foe?

Abstract

EEG-based brain–computer interfaces (BCI) have promising therapeutic potential beyond traditional neurofeedback training, such as enabling personalized and optimized virtual reality (VR) neurorehabilitation paradigms where the timing and parameters of the visual experience is synchronized with specific brain states. While BCI algorithms are often designed to focus on whichever portion of a signal is most informative, in these brain-state-synchronized applications, it is of critical importance that the resulting decoder is sensitive to physiological brain activity representative of various mental states, and not to artifacts, such as those arising from naturalistic movements. In this study, we compare the relative classification accuracy with which different motor tasks can be decoded from both extracted brain activity and artifacts contained in the EEG signal. EEG data were collected from 17 chronic stroke patients while performing six different head, hand, and arm movements in a realistic VR-based neurorehabilitation paradigm. Results show that the artifactual component of the EEG signal is significantly more informative than brain activity with respect to classification accuracy. This finding is consistent across different feature extraction methods and classification pipelines. While informative brain signals can be recovered with suitable cleaning procedures, we recommend that features should not be designed solely to maximize classification accuracy, as this could select for remaining artifactual components. We also propose the use of machine learning approaches that are interpretable to verify that classification is driven by physiological brain states. In summary, whereas informative artifacts are a helpful friend in BCI-based communication applications, they can be a problematic foe in the estimation of physiological brain states.

A review of combined functional neuroimaging and motion capture for motor rehabilitation

“…Six out of the 19 studies were categorized as basic research [ 39 , 42 , 48 , 54 , 56 , 57 ], 9 out of 16 studies were categorized as translational research [ 43 , 45 , 47 , 49 , 50 , 52 , 53 , 55 , 58 ] and only four of the 16 studies could be classified as clinical research [ 40 , 41 , 44 , 46 ]. No clinical studies were found for diagnostic applications.…”

“…An overview is given in Table 4 . The majority of the studies focused on upper limb movements (12 out of 19) [ 39 – 41 , 43 , 46 , 46 , 49 , 50 , 52 – 54 , 58 ], with reaching being the most commonly studied movement (7 out of 19) [ 40 , 41 , 43 , 46 , 49 , 51 , 52 ]. The remaining studies focused on whole-body movements, including a balance task [ 45 , 57 ] and gait tasks based on overground walking [ 44 , 56 ] or treadmill walking [ 42 , 48 , 55 ].…”

“…In seven studies [ 44 , 46 , 47 , 51 – 54 ], the recorded data stream was downsampled, particularly for studies using higher recording sampling rates of 500 Hz and above. Downsampling was done to 56 Hz [ 46 ], 128 Hz [ 44 , 52 , 54 ], 200 Hz [ 53 ], 1000 Hz [ 58 ], or 2048 Hz [ 47 ] during further signal preprocessing.…”

Integrating Biofeedback and Artificial Intelligence into eXtended Reality Training Scenarios: A Systematic Literature Review

“…As such, details regarding integration of biosensing technologies with synthetic environment development tools (i.e., game engines) should be a focus of development approaches. The Unity game engine was used in several studies (including De Bruyne et al, 2021; Delvigne et al, 2022; Dwivedi et al, 2020; Marín-Morales et al, 2021; Porssut et al, 2022; Rodrigues et al, 2021; Stranick & Lopez, 2022; Tsai et al, 2021) and the Unreal Engine 4 was used in only one study from the 48 studies included (McDermott et al, 2022). Hardware and software integration is important for system development and Unity is currently dominating the “lock-in”, where platforms are made dependent on and interoperable with each other to assure market viability (Foxman, 2019).…”