Date of Award

2019

Degree Type

Dissertation

Degree Name

Doctor of Philosophy in Electrical Engineering

Specialization

Biomedical Engineering

Department

Electrical, Computer, and Biomedical Engineering

First Advisor

Kunal Mankodiya

Abstract

Decoding the human's brain functional architecture is the most profound and far-reaching scientific challenges of our time. Currently existing noninvasive brain imaging technologies are constrained by costly, bulky and fixed hardware that preclude imaging of the functioning brain in a wide range of temporal and naturalistic environments.

This doctoral dissertation focused on designing and developing of a new generation portable, wearable, configurable, and wireless functional near-infrared spectroscopy (fNIRS) neuroimaging system that allows us to monitor and study brain function in a naturalistic environment. The developed optical system maps changes in chromophore concentration, oxy hemoglobin (HbO2) and deoxyhemoglobin (Hb) of the cortical surface of the brain in response to the human brain functions noninvasively.

The fNIRS hardware system was designed based on the Internet-of-Things (IoT) platform using Intel Edison for onboard intelligence, configurability and data transmission. The analog and digital circuits were designed and developed. The fNIRS controlling unit consists of three printed circuit boards (PCBs) sandwiched together: (1) embedded system PCB, (2) analog circuit PCB and (3) digitization PCB. We programmed the system to perform highly complex operations such as montage configuration, sequential NIR light injection and low-intensity back-reflected diffused light measurement from the cortical area of the brain at two wavelengths, data conversion, computation, and wireless data transmission, etc. The portable fNIRS system was capable of transferring multi-channel fNIRS data to a computer in real-time. The fNIRS channels are combinations of light sources and detectors.

The light-emitting diode (LED) and silicon photodiode (Si-PD) detector based fNIRS optodes (source and detector) were developed. We used modern design tools

such as 3D printing and laser cutting to fabricate human-centered fNIRS optodes.

Feedback was taken from participants of different groups throughout the iterative

design process. Two types of fNIRS optodes were designed; one was based on

forehead patch and the other was integrable into an electrode head cap that can

be placed along with electroencephalography (EEG) electrodes.

Our software architecture wirelessly connects the fNIRS system with a computer or Android tablet through WiFi and interacts to send configuration settings and also to receive fNIRS data in real-time. A host computer connects to the fNIRS control unit via authentication and performs bi-directional communication in real-time to instruct the fNIRS controller to operate in a synchronized manner. The host computer also simultaneously collects and processes the fNIRS signal, and displays hemodynamics responses. A MatLab-based graphical user interface (GUI) software was developed to control the fNIRS system, perform advanced fNIRS data processing, and to visualize multi-channel HbO2 and Hb in real-time.

The fNIRS system was also capable of synchronizing with the EEG systems to simultaneously collect fNIRS and EEG signal for the multimodal brain imaging. The hardware and software of the fNIRS system were evaluated for its safety, performance and working. After quantitively comparing our fNIRS system with the other portable fNIRS systems proposed by other researchers, we found that our system has several advantages in terms of portability, connectivity, wearability, configurability, comfortability, frame rate, and resolution, etc. for the naturalistic brain imaging.

Experimental studies on human participants in naturalistic environments with different experimental protocols were performed to validate the working of our fNIRS system. These experiments not only verified the working of the fNIRS system but also, we studied how the brain's hemodynamic activities are modulated by different tasks, such as meditative breathing exercise, mental arithmetic task, and working memory task, etc.

The first human subject experimental study with our fNIRS system was performed to measure the HbO2 and Hb concentration changes at the regional muscle (forearm) due to the arterial occlusion. The results of the forearm muscle response during arterial occlusion experiment on the 8 participants proved working of the hardware and software of our fNIRS system.

The second experiment was to image prefrontal cortex of the brain using our fNIRS system on 8 participants while participants performed meditative breathing exercise. The results showed that the system was able to image hemodynamic activities of the brain in this experiment. Also, we derived heart rate variability from the fNIRS data.

In the third experiment, we performed brain imaging while participants were performing mental arithmetic tasks with the difficulty levels of the task in the increasing order to increase the cognitive load. We found that hemodynamic response to the cortical activity rises in response to the mental arithmetic over time. The results also showed that the three stimulus-evoked hemodynamic responses due to the three difficulty levels of the tasks cause three peaks of HbO2 of three different amplitudes.

In the fourth experiment, we performed experimental studies using our fNIRS system to image prefrontal cortex (PFC) areas of the brain of 13 participants while they performed n-back working memory (WM) tasks. We created an n-back experimental protocol that had 16 trials (task blocks) in one run to access four different levels of WM loads and rest periods. Our results showed an incremental brain's hemodynamic activity with the incremental WM load. The results proved that our fNIRS system is instrumental in assessing and distinguishing different WM loads. We observed consistent hemodynamic activities over both left and right hemispheres of the brain during all the trials. However, we found slightly higher left PFC activation than the right one.

We also implemented machine learning (ML) method to classify working memory load from the fNIRS signal. We have extracted eight important features from the hemodynamic brain signals and trained six different k-nearest neighbor (k-NN) ML classifiers. The performance of the six k-NN classifiers was evaluated with the experimental data sets. The results showed that Weighted and Fine k-NN performed best (95 %) in classifying 5 different WM loads.

We also performed multimodal EEG and fNIRS experiments placing tripolar EEG (tEEG) electrodes and fNIRS optodes together at the same time using a common montage cap. The fNIRS measured regional cerebral hemodynamic response complimenting EEG, that directly measures the neural activity of the brain. The optodes' and EEG electrode's actual locations on the scalp during experiments were translated to a standard 10-20 electrode placement system to present fNIRS and EEG channels on a standard brain atlas. The fNIRS and EEG channels were mapped accordingly and topographically presented the brain activities in the multimodal brain studies.

The resulting novel fNIRS system provided an essential tool that is accessible to the broad research and medical community for brain and behavioral research in the naturalistic environments, with extensive applicability in neuroscience, cognitive science, and cognitive psychology.

Available for download on Sunday, August 09, 2020

Share

COinS