Open Collections

Chinchani, Abhijit Mahesh

University of British Columbia

In the age of big data and large neuroscience datasets, dimensionality reduction is employed to reduce the complexity of recorded data and explain it with fewer representative variables. Unlike unsupervised methods, which rely solely on the data's inherent structure, supervised techniques leverage additional information to achieve better feature extraction and data representation. In this dissertation, I present two applications of supervised techniques: one to identify cognitive correlates of psychiatric symptoms and the other to estimate latent EEG dimensions to study the effects of transcranial alternating current stimulation (tACS). First, I applied constrained principal component analysis to identify cognitive correlates of psychiatric symptoms in two datasets. In the first dataset, I examined the overlap between cognition and hallucinatory experiences and developed a set of reliability tests to minimize the reporting of spurious results due to overfitting. Next, I improved the reliability testing algorithm and used it to find links between cognition and symptoms in schizophrenia spectrum disorder in another dataset. This approach led to a nuanced and multimodal understanding of psychiatric symptoms. Next, I applied joint decorrelation, another supervised technique, to extract latent EEG oscillations corresponding to frequency-specific oscillations and to study whether tACS affects them. I first investigated the effects of tACS on auditory steady-state responses (ASSRs). Narrowband latent oscillations for ASSRs at 40Hz were extracted, but no effects of tACS on these evoked oscillations were observed. Subsequently, I examined the effects of tACS on broadband alpha oscillations. By extracting latent dimensions corresponding to alpha frequencies, I observed that tACS affected alpha oscillations in a task-specific manner. This adds a critical variable to be considered for applications of tACS, whereby increasing the electric field by placing the anode further away from the cathode has a cost of producing a less localized effect. However, my findings show that the effects of tACS can be localized by using specific tasks that engage the desired brain areas, enhancing the precision and effectiveness of the stimulation. These applications demonstrate the utility of supervised dimensionality reduction in neuropsychiatric research, providing deeper insights into cognitive correlates of psychiatric symptoms and the neural mechanisms underlying tACS.

Text

2024-09-09

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/89207

Bioinformatics

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/