Real‐time Neurofeedback Rewards Memory‐related Potentials in Older Brains

Abstract

AbstractBackgroundElectroencephalogram (EEG)‐based neurofeedback (NF) allows reward of desired brain activity in milliseconds. Compared to frequency power‐based approach during resting state EEG, event‐related potentials (ERPs) based NF during a task is technically challenging because of the cost of computation over a time window. We have previously established that left‐frontal memory‐related potentials discriminate normal cognitive aging from age‐ and education‐ matched older adults with risk of mild cognitive impairment (MCI). To modulate brainwaves during working memory retrieval towards healthier patterns, we develop an ERP‐based neurofeedback protocol to reward better memory‐related potentials.MethodThe training protocol is developed based on a modified delayed match‐to‐sample task also known as Bluegrass working memory paradigm. During each memory trial, a participant holds the sample target image in his/her working memory and responds whether each subsequent image is a Match or Nonmatch to the memory target by pressing the button. Real‐time classification of extracted ERP features is analyzed after stimulus onset, the dynamic machine‐learning program then modifies the stimulus by turning it into vivid color (reward) or leaving it with black‐and‐white (unrewarded). The processing pipeline has been developed using ERPs of older adults with and without MCI. The left frontal sites (F3 and F7) were preprocessed by re‐referencing and filtering with a 0.05 – 40 Hz band‐pass filter. For each participant, the EEG of 5 offline base trials in a 300‐600ms time window was used as a baseline. The ERP difference of the baseline versus the subsequent single trial is calculated to determine whether the image will be colored or left at black‐and‐white.ResultOur newly developed ERP‐neurofeedback protocol has succeeded in providing visual feedback based on fluctuation of single trial ERP in real time. The program was also designed to fit in each individual due to varied sensibility to NF. Currently we are to examine the feasibility and effectiveness of the proposed protocol by measuring memory indicators before and after NF training sessions.ConclusionWe have developed a new protocol using ERP‐based neurofeedback to improve personalized memory retrieval. The program is the first step to intervention trials for enhancing successful memory by rewarding desired memory‐related potentials in older brains.