Stem I

STEM is a class taught by Dr. Crowthers. In the first half of the year (August to February), the course is part of STEM I, where students work on their independent research projects. Students can choose a topic of their interest and pursue a project in it. Students share their work at the December and February STEM fairs, and they have the opportunity to advance on to regional, state, and international science fairs.

Quad Chart

This quad chart summarizes my independent research project for STEM. My project uses electroencephalogram-based (EEG) neurofeedback to study how early changes in brainwaves can predict improvements in focus and cognitive performance. By analyzing alpha, beta, and gamma activity during training, this research aims to make neurofeedback more personalized and effective.

Investigating the Relationship Between Alpha, Beta, and Gamma Neurofeedback and Cognitive Performance to Predict Training Outcomes

EEG data is analyzed across multiple frequency bands to characterize changes in brain activity over time. Analyzing alpha, beta, and gamma bands illustrates how frequency-based EEG analysis reveals meaningful patterns and helps capture differences in how individuals respond to neurofeedback. Because individuals respond differently, and some may not respond at all, the ability to identify early whether neurofeedback is beneficial is essential for guiding effective and personalized approaches.

Abstract

Neurofeedback training (NFT) has shown promise for improving cognitive performance, yet outcomes vary widely across individuals, limiting its reliability and effectiveness. This variability highlights a critical knowledge gap, as there is a lack of early training markers that can predict who will benefit from specific neurofeedback protocols. This project investigates whether early changes in EEG activity across alpha, beta, and gamma frequency bands can predict long-term cognitive improvements. Using EEG neurofeedback, participants complete multiple training sessions while brainwave activity and cognitive performance measures are collected before, during, and after training. Preliminary observations suggest that early-session modulation and stability within the trained frequency band may reflect neurofeedback trainability. By analyzing early learning slopes and spectral power changes, this study aims to identify predictive neural signatures of successful training. These findings may support the development of personalized neurofeedback approaches, improving the efficiency and effectiveness of cognitive enhancement and therapeutic interventions.

Graphical abstract

Research Proposal

Research Question

Can early changes in alpha, beta, and gamma EEG activity during neurofeedback predict long-term changes in cognitive performance?

Hypothesis

Individuals who show early increases in the trained frequency band (upper-alpha, beta, or gamma) will demonstrate greater cognitive gains across training.

Background

visual background

The human brain communicates through rhythmic electrical oscillations known as brainwaves, which can be measured using electroencephalography (EEG). Each of the five frequency bands correspond to a different state of consciousness or cognitive state. Alpha, beta, and gamma waves are different brainwave frequency bands, and they are closely related to functions in the brain like optimized cognitive performance, thinking, focus, and mental sharpness (Marzbani et al., 2016). Neurofeedback training is a type of biofeedback which uses sensors to monitor real-time brainwave activity through EEG signals. Neurofeedback training (NFT) has been used with different protocols, or brain waves, to provide effective treatment for many diseases. Originally used to treat conditions like epilepsy, NFT has evolved to be applied in extensive areas, like enhancing cognitive function (Vernon et al., 2003; Egner et al., 2004). Across different people, however, the effects of NFT can vary, which can make it difficult to predict if certain protocols are going to be more beneficial than others for every individual. Thus, this project aims to find out whether early EEG changes can help predict long-term neurofeedback outcomes.

Neurofeedback and Brain Oscillations

EEG signals indicate the brain’s electrical activity, with different frequency bands that correspond to different cognitive and behavioral states. NFT utilizes this by providing participants with feedback on the amplitudes of their brainwaves and allows the brain to self-regulate through various stimuli (Zoefel at al., 2011). Over time, individuals can modify the amplitudes of specific frequency bands, which lead to measurable changes in cognitive outcomes.

Alpha Band Activity

The alpha frequency band (8-13 Hz) is closely associated with being a strong marker of cognitive control and inhibition (Hanslmayr et al. 2005). Within this frequency band, upper alpha (UA) activity is specifically related to memory performance and improved cognitive performance (Zoefel et al., 2011).

Hanslmayr et al. (2005) provided key evidence which suggests that increasing upper alpha power through NFT can directly boost cognitive performance. In the study, participants who were trained through NFT to increase UA amplitude demonstrated significant improvements in their scores in mental ability tests. These tests were taken across participants who did and did not receive NFT to qualify performance results. From this study, Zoefel et al. (2011) conducted further sessions of NFT and found that subjects who were able to successfully increase UA amplitude showed much greater gains in cognitive tasks than the control participants. It is important to note that UA increases were independent from those of other frequency bands like lower beta, or lower alpha.

Beta Band Activity

The beta frequency band (15–32 Hz) is associated with active cognitive processing, focused attention, and task engagement. Beta frequency increases when individuals perform tasks that require fast decision-making and sustained concentration. Particularly, lower frequency beta activity (15-20 Hz) is linked to maintaining more attentional focus, while higher frequency beta activity is tied to working memory updating and executive functioning (Engel & Fries, 2010; Spitzer & Haegens, 2017).

For the purpose of this study, beta-band neurofeedback is a potential target for improving aspects of cognitive performance that rely on sustained alertness. Beta oscillations respond quickly to task demands, and thus they may also serve as an early predictor of neurofeedback trainability. Individuals who indicate faster increases in beta power during initial training sessions can show greater cognitive gains across the training period. Observing whether beta-band responsivity predicts future treatment outcomes can help identify participants who are most likely to benefit from a beta-focused neurofeedback protocol.

Gamma Band Activity

Gamma oscillations (generally 32-100+ Hz) are among the fastest brain rhythms and have been associated with high-level cognitive integration. Gamma activity is believed to support the binding of information across neural networks, enabling processes such as complex attention, feature integration, working memory maintenance, and conscious perception (Jensen et al., 2007). Increased gamma power is often observed during tasks involving mental effort, cross-modal processing, and the formation of coherent percepts from distributed neural representations.

For this project, gamma activity represents an important comparative frequency band, particularly in understanding whether higher-frequency oscillations provide additional predictive power for identifying strong neurofeedback responders. If early EEG signals for gamma waves correlate with cognitive improvements later in training, gamma-band responsiveness may be a useful biomarker of better cognitive performance. Conversely, limited early gamma trainability may indicate that lower-frequency protocols like alpha or beta may be more effective for a participant. Evaluating gamma-band behavior along with alpha and beta rhythms can therefore help determine whether individualized neurofeedback prescriptions can be optimized based on early EEG signals.

Predicting Neurofeedback Training Outcomes

Recent evidence suggests that neurofeedback training (NFT) outcomes can be predicted after only a few sessions by examining early changes in neural activity and baseline EEG characteristics. Studies such as Hanslmayr et al. (2005) and Zoefel et al. (2011) found that participants who showed early increases in individually defined upper alpha (UA) amplitude were more likely to achieve long-term gains in both UA regulation and cognitive performance. A systematic review by Weber et al. (2020) further supported this, identifying baseline power in the trained frequency band and early session performance as key predictors of NFT success. Individuals who exhibit stronger resting UA activity or early positive trends in UA modulation tend to become “responders,” showing greater cognitive improvements over time.

While alpha-frequency predictors are the most well-established, emerging evidence suggests that beta and gamma frequency bands exhibit similar early indicators of trainability. Beta rhythms, which reflect attentional engagement and task preparation, often demonstrate rapid changes during the first few NFT sessions. Early increases in beta power have been associated with later improvements in cognitive tasks requiring focus and inhibitory control (Engel & Fries, 2010; Spitzer & Haegens, 2017). This means that participants who show initial responsivity to beta-based feedback may be more capable of long-term beta modulation and associated cognitive gains. Conversely, individuals with minimal early changes in beta activity typically demonstrate weaker overall NFT outcomes, suggesting that beta-band responsiveness may function as an early biomarker of attentional trainability.

Gamma oscillations, though more challenging to measure and train, also show potential as predictors of NFT outcomes. Gamma activity reflects high-level cognitive integration, including working memory maintenance and perceptual binding (Jensen et al., 2007). Participants who exhibit early gamma responsivity may later show enhanced performance on tasks involving complex attention or cognitive performance. For this study, tracking early EEG changes across these three frequency bands will help identify which participants are most likely to be “responders” to each protocol. More importantly, this information will help determine which frequency band is best suited to improving an individual's cognitive performance.

By using early neural indicators to predict NFT outcomes, this project aims to contribute to the development of personalized neurofeedback, where protocols are selected based on each individual’s neurophysiological profile and early training performance rather than a one-size-fits-all approach.

Procedure

The materials used in this study included an OpenBCI Gelfree Electrode Cap with electrodes, an OpenBCI CytonDaisy 16-channel biosensing board, a Faraday cage to reduce electrical noise, saline conductive solution and cotton swabs for signal optimization, EEGLAB running in MATLAB for EEG processing and analysis, and standardized cognitive tasks to assess performance. Resting-state EEG was first collected to establish baseline alpha, beta, and gamma activity. Participants then completed neurofeedback training sessions targeting a specific frequency band using real-time visual EEG feedback. Standardized cognitive tasks measuring attention and working memory were administered before and after the training period. This procedure was repeated across sessions, and EEG changes during the first one to two sessions were quantified in MATLAB using EEGLAB to measure early modulation of the trained band. Early EEG changes and baseline neural activity were statistically compared with later cognitive performance outcomes to identify predictors of neurofeedback responsiveness and frequency-specific effectiveness.

Procedure Infographic

Figures

Graph 1: This plot shows simulated alpha-band power across multiple runs for each subject. Actual measurements are solid lines, while predicted values from a linear model are dashed. The y-axis corresponds to the alpha band (8–13 Hz), and the early session is shown as the first point

Figure 1 Graph of Alpha power
Figure 2 Graph of Beta power

Graph 2: This plot shows the simulated beta-band power across multiple runs for each subject. The solid lines represent actual values, while dashed lines indicate predicted values from a linear model based on run index. The early session is included as the first point on the x-axis, and the y-axis is constrained to the beta frequency band (13–30 Hz).

Graph 3: This plot shows the simulated gamma-band power across multiple runs for each subject. Solid lines show actual values, and dashed lines show predictions from a linear model using run index. Early sessions are included on the x-axis, with the y-axis reflecting the gamma frequency band (30–100 Hz).

Figure 3 Graph of Gamma power

Analysis

Early-session EEG power can provide a rough prediction of later session trends. Linear models using run index capture trends across runs for each frequency band. Predicted values closely follow actual trends, showing the model’s usefulness. Beta (13–30 Hz) and Gamma (30–100 Hz) bands show variability across subjects and runs. Early sessions are important as baseline points for tracking changes over time. Simulations highlight that higher-frequency bands (gamma) tend to have more variability than lower-frequency bands.

Discussion & Conclusion

The results suggest that early-session EEG activity may provide meaningful insight into how individuals respond to neurofeedback training. Across alpha, beta, and gamma bands, early modulation trends were reflected in later session patterns, supporting the idea that initial responsiveness may predict long-term cognitive outcomes. Linear modeling demonstrated that early changes in spectral power can approximate overall training trajectories, particularly in the alpha band, which has been strongly associated with cognitive enhancement (Zoefel et al., 2011). Variability observed in the beta and gamma bands highlights the importance of individualized analysis. Higher-frequency oscillations showed greater fluctuation across subjects, suggesting that responsiveness depends on both baseline neural activity and the specific frequency band targeted during training. These findings reinforce the hypothesis that neurofeedback should not follow a one-size-fits-all approach. From an applied perspective, identifying responders within the first few sessions could improve the efficiency of neurofeedback programs by reducing unnecessary training for non-responders and guiding frequency-band selection early in the process. This approach may have implications for cognitive enhancement, educational performance, attention training, and potential clinical interventions such as ADHD support. By using early neural indicators to personalize protocols, neurofeedback can become more targeted, cost-effective, and scientifically grounded. Overall, this study supports the potential of early EEG markers as predictors of neurofeedback success and contributes to the development of data-driven, individualized brain training strategies.

February Fair Poster

References