ORCID
http://orcid.org/0000-0003-2543-0851
Date of Award
Summer 8-15-2020
Degree Name
Doctor of Philosophy (PhD)
Degree Type
Dissertation
Abstract
High-density diffuse optical tomography (HD-DOT) is a relatively new neuroimaging technique that detects the changes in hemoglobin concentrations following neuronal activity through the measurement of near-infrared light intensities. Thus, it has the potential to be a surrogate for functional MRI (fMRI) as a more naturalistic, portable, and cost-effective neuroimaging system. As in other neuroimaging modalities, head motion is the most common source of noise in HD-DOT data that results in spurious effects in the functional brain images. Unlike other neuroimaging modalities, data quality assessment methods are still underdeveloped for HD-DOT. Therefore, developing robust motion detection and motion removal methods in its data processing pipeline is a crucial step for making HD-DOT a reliable neuroimaging modality.
In particular, our lab is interested in using HD-DOT to study the brain function in clinical populations with metal implants that cannot be studied using fMRI due to their contraindications. Two of these populations are patients having movement disorders (Parkinson Disease or essential tremor) with deep brain stimulation (DBS) implants and individuals with cochlear implants (CI). These two groups both receive tremendous benefit from their implants at the statistical level; however, there is significant single-subject variability. Our overarching goal is to use HD-DOT to find the relationships between the neuronal function and the behavioral measures in these populations to optimize the contact location of these implant surgeries. However, one of the challenges in analyzing the data in these subjects, especially in patients with DBS, is their high levels of motion due to tremors when their DBS implant is turned off. This further motivates the importance of the methods presented herein for separating signal from noise in HD-DOT data.
To this end, I will first assess the efficacy of state-of-the-art motion correction methods introduced in the fNIRS literature for HD-DOT. Then, I will present a novel global metric inspired by motion detection methods in fMRI called GVTD (global variance of the temporal derivatives). Our results show that GVTD-based motion detection not only outperforms other comparable motion detection methods in fNIRS, but also outperforms motion detection with accelerometers.
I will then present my work on collecting and processing HD-DOT data for two clinical populations with metal implants in their brain and the preliminary results for these studies.
Our results in PD patients show that HD-DOT can reliably map neuronal activity in this group and replicate previously published results using PET and fMRI. Our results in the CI users provide evidence for the recruitment of the prefrontal cortex in processing speech to compensate for the decreased activity in the temporal cortex. These findings support the theory of cognitive demand increase in effortful listening situations.
In summary, the presented methods for separating signal from noise enable direct comparisons of HD-DOT images with those of fMRI in clinical populations with metal implants and equip this modality to be used as a surrogate for fMRI.
Language
English (en)
Chair and Committee
Ralf Wessel Joseph Culver
Committee Members
Tamara Hershey, Shankar Mukherji, Jonathan E. Peelle, Abraham Z. Snyder,
Recommended Citation
Sherafati, Arefeh, "Separating Signal from Noise in High-Density Diffuse Optical Tomography" (2020). Arts & Sciences Electronic Theses and Dissertations. 2344.
https://openscholarship.wustl.edu/art_sci_etds/2344