Cortical Visual Motion Response And Its Application In Minimally Invasive Brain-Computer Interface

Electrocorticography (ECoG) has been widely used in neurosurgery clinicaldiagnosis, such as the surgical treatment of epilepsy and brain tumor. In the past decade,ECoG has also attracted increasing interest for research of brain activity. Compared toelectroencephalography (EEG), ECoG has major advantages for implementingadvanced brain-computer interfaces (BCIs), such as higher spatial resolution, broaderfrequency band and higher amplitude and signal-to-nosie ratio. However, ECoG basedBCIs have difficulties toward practical use due to the risk of the open skull surgery.Therefore, the biggest challenge to transfer the ECoG BCI from concept to realapplication is how to minimize both the size of the involved brain regions and thenumber of implanted electrodes so as to reduce risks.Moving objects would elicite visual motion evoked response on EEG, but fewersuch studies were conducted on cortical neural response. The ECoG BCI experimentplatform was set up in this study. Here, we employed the attentional modulation ofvisual motion response and collected the ECoG and related medical imaging data fromfive epilepsy patients. Using the paradigm of N200speller, cortical event-relatedpotentials (ERPs) were elicited as a negative peak around200ms--N200component.Moreover, power increase of the high gamma (60-140Hz) frequency range was foundaround100-500ms to be associated with the overtly attended moving visual stimuli.Based on the characteristics of visual motion response, the classification algorithm formotion-onset visual evoked potential recorded by ECoG was proposed, combining thehigh gamma as a new feature for pattern classification. A two-layer classifier wasdesigned and ECoG signals in both the low frequency range (i.e., the ERPs) and highgamma range were extracted as features for classification. To achieve minimalinvasiveness, one optimal electrode per patient was selected for BCI control. Weachieved higher BCI classification accuracy by including high gamma feature.Classification accuracy by employing high gamma response only was higher than byusing ERP responses only (81.24±6.23%vs.75.48±4.18%), and by employing bothhigh gamma and ERP responses (84.22±5.54%) was significantly higher than either one(p<0.005). These results also demonstrated the feasibility of implementing an ECoGBCI by single channel. In the study, the significant visual motion responses of high gamma were located inthe parietal-temporal-occipital junction and the posterior occipital cortex, which hasgood task-related specificity. More importantly, the high gamma responses were locatedwithin brain regions specialized in visual motion processing as mapped by fMRI. Basedon the above results, the first minimally invasive ECoG BCI guided by non-invasivemedical imaging was proposed. The single subdural channel for BCI control can beinserted through a burr hole into the brain in a minimally invasive way employingpre-operative fMRI scaning for target region localization. This approach well solved thedifficulties of the minimal invasiveness of ECoG BCI techniques. Together with thesingle channel design, this study demonstrated that a minimally invasive BCI can beachieved by employing the visual motion response.