Neurophysiological Biomarkers of Amygdala Stimulation on Medial Temporal Lobe Networks Supporting Declarative Memory in Patients With Epilepsy
Abstract number :
2.074
Submission category :
3. Neurophysiology / 3E. Brain Stimulation
Year :
2018
Submission ID :
502685
Source :
www.aesnet.org
Presentation date :
12/2/2018 4:04:48 PM
Published date :
Nov 5, 2018, 18:00 PM
Authors :
Mohammad Sendi, Georgia Institute of Technology; Cory Inman, Emory University; Joseph Manns, Emory University; Kelly Bijanki, Emory University; Robert E. Gross, Emory University School of Medicine; Jon T. Willie, Emory University; and Babak Mahmoudi, Emor
Rationale: Recent studies show that activating the amygdala using direct electrical stimulation enhances declarative memory for specific events in humans, in part by modulating the neural states in the hippocampus and perirhinal cortex. A question that remains to be answered is what underlying neurophysiological changes are induced in the hippocampus by amygdala stimulation. In this study, we utilize a machine learning approach to investigate how amygdala stimulation modulates network activities in the hippocampus and the perirhinal cortex in epilepsy. Methods: Fourteen patients with drug-resistant epilepsy and depth electrodes placed in the amygdala performed a yes/no recognition memory task for neutral images of objects. During encoding, 160 images were shown to patients. Half of the images, as a stimulated trial, were followed by 1-second of low-amplitude biphasic rectangular pulses at 0.5 mA in 8 trains of 4 pulses at 50 Hz to the amygdala. Three seconds of local field potential (LFP) signals before and after each stimulation were recorded from all memory related channels, specifically those in the basolateral amygdala, hippocampus, and perirhinal cortex. We used the average power in different electrophysical bands including d (1-4Hz), ? (4-9Hz), a (9-12Hz), ß (12-30Hz) and ? (30-90Hz, divided into 15 Hz frequency bins) as a potential biomarker to train a logistic regression classifier using an elastic net regularization (ENR) method to discriminate between pre- and post-stimulation states. Results: The average area under the receiver operating characteristic (AUROC) of the classifier with 10-fold cross-validation across all patients was 0.75±0.08 for the amygdala stimulation trials and 0.68±0.1 for the sham stimulation trials. The AUROC is representative of the predictive accuracy of the classification model where a perfect classifier will have an AUROC of 1. In addition, feature selection using ENR found that modulation of alpha (9-12 Hz), and slow gamma (30-60 Hz) frequency ranges were most informative in discriminating between stimulated and sham trials. Conclusions: These findings show amygdala stimulation induces changes in specific frequency bands in the human hippocampal subregions during the encoding phase. The predominant changes in the alpha and slow gamma bands are in agreement with previous findings on the biomarkers of memory encoding in rodent studies. These biomarkers can inform developing neuromodulation therapies for epilepsy that will have minimal or no side-effects on memory. Funding: None