| dc.description.abstract |
Simultaneous measurements of intra-cortical electrophysiology and hemodynamic signals in primates are essential for relating human neuroimaging studies with intra-cortical electrophysiology in monkeys. Previously, technically challenging and resourcefully demanding techniques such as fMRI and intrinsic-signal optical imaging have been used for such studies. Functional near-infrared spectroscopy is a relatively less cumbersome neuroimaging method that uses near-infrared light to detect small changes in concentrations of oxy-hemoglobin (HbO), deoxy-hemoglobin (HbR) and total hemoglobin (HbT) in a volume of tissue with high specificity and temporal resolution. In a series of studies, we investigated the neurovascular correlates of fNIRS signals in primates and demonstrated its feasibility in brain-state classification for applications in brain-computer interfaces and neurofeedback in humans in four studies. To test the feasibility of using epidural fNIRS with concomitant extracellular electrophysiology, in our first study we recorded neuronal and hemodynamic activity from the primary visual cortex of two anesthetized monkeys during visual stimulation. We recorded fNIRS epidurally. We performed simultaneous cortical electrophysiology using tetrodes placed between the fNIRS sensors. We observed robust responses to the visual stimulation in both [HbO] and [HbR] signals, and quantified the signal-to-noise ratio of the epidurally measured signals. Our results show that epidural fNIRS detects single-trial responses to visual stimuli on a trial-by-trial basis, and when coupled with cortical electrophysiology, is a promising tool for studying local hemodynamic signals and neurovascular coupling. A major issue with functional neuroimaging, is that the neuroimaging signal correlates with both spiking, and various bands of the local field potential (LFP), making the inability to discriminate between them a serious limitation for interpreting hemodynamic changes. In our second study, we use the technique described above to investigate the neuro-vascular correlates of the fNIRS signals, and find that lowfrequency LFPs correlate with the hemodynamic signal's peak amplitude, whereas spiking correlates with its peak-time and initial-dip. We also find spiking to be more spatially localized than low-frequency LFPs. Our results suggest that differences in the spread of spiking and low-frequency LFPs across cortical surface influence different parameters of the hemodynamic response. These results demonstrate that the hemodynamic response-amplitude is a poor correlate of spiking activity. Instead, we demonstrate that the timing of the initial-dip and the hemodynamic response are much more reliable correlates of spiking, reflecting bursts in spike-rate and total spike-counts
respectively. In our third study, we turned our attention to the initial-dip. The initial-dip is a transient decrease frequently observed in functional neuroimaging signals, immediately after stimulus onset, and is believed to originate from a rise in deoxy-hemoglobin (HbR) caused by local neural activity. It has been shown to be more spatially specific than the hemodynamic response, and is believed to represent focal neuronal activity. However, despite being observed in various neuroimaging modalities (such as fMRI, fNIRS, etc), its origins are disputed and its neuronal correlates unknown. Here, we show that the initial-dip is dominated by a decrease in total-hemoglobin (HbT). We also find a biphasic response in HbR, with an early decrease and later rebound. However, HbT decreases were always large enough to counter spiking-induced increases in HbR. Moreover, the HbT-dip and HbR-rebound were strongly coupled to highly localized spiking activity. Our results suggest that the initial-dip helps prevent accumulation of spiking-induced HbR concentration in capillaries by flushing out HbT, probably by active venule dilation. The purpose of our fourth study was to develop and test a real-time method for subject-specific and subject-independent classification of multi-channel fNIRS signals using support-vector machines (SVM), to determine its feasibility as an online neurofeedback system. We used left versus right hand movement execution and movement imagery as study paradigms in a series of experiments. In the first two experiments, activations in the motor cortex during movement execution and movement imagery were used to develop subject-dependent models that obtained high classification accuracies thereby indicating the robustness of our classification method. In the third experiment, a generalized classifier-model was developed from the first two experimental data, which was then tested in new subjects. Application of this method in new participants showed mean classification accuracy of 63% for movement imagery tasks and 80% for movement execution tasks. These results demonstrate that SVM based real-time subject-independent classification of fNIRS signals is feasible. This method has applications in the field of hemodynamic BCIs, and neuro-rehabilitation. |
en |
