The brain processes information through the constant activity of electrical signals. These signals can be measured through Brain-Computer Interface (BCI) devices, such as an electroencephalographic (EEG) headset. We investigated whether EEG signals can predict memory retention, and used these predictive features to optimize a flashcard study session. The first aim was to identify features that predict memory retention. Using 10 subjects (aged 18-21), we recorded their EEG signals while they studied a flashcard set. Approximately 24 hours later, each subject took a quiz about the flashcard set, indicating retention level. We then used the participants' EEG and score data to train a Random Forest Classifier to determine the best EEG signal features for predicting memory retention. While results remain preliminary, this approach shows promise for identifying neurological markers of learning. Secondly, we aimed to create a web app demonstrating future model use cases. The app allows users to study more effectively, using a BCI cap. EEG signals are used in real-time to predict whether the user is likely to retain the information on the card they are viewing. If the signals indicate lower retention likelihood, the interval before that card reappears is shortened, focusing study time where needed most. This project represents the possibility of more personalized learning tools that adapt to the brain in real time, with potential applications across different age groups, education levels and looking to the future, across species.

Introduction Withdrawal is one of the main reasons those with opioid use disorder continue their drug use. Noradrenaline release contributes to opioid withdrawal, where drugs such as morphine bind to μ-opioid receptors (MOR) directly on noradrenergic neurons. During withdrawal, noradrenergic neuronal activity is increased and there is increase in noradrenaline release that contributes to both physical and protracted withdrawal states such as insomnia and anxiety. We hypothesize that MOR deletion on noradrenergic neurons will reduce aversion, withdrawal behaviors, and anxiety-like behaviors in animal models. Methods We will be using adult male and female C57 mice and a conditional knockout (cKO) of MORs from all noradrenergic neurons in the whole animal by breedingDbh-cre mice withflMOR/flMOR mice. Cage mate cre- mice will be used as controls. Mice will be made opioid dependent by oral morphine administration by replacing drinking water with morphine solution (0.25-0.75 mg/mL) over 14 days. Control mice will receive normal drinking water. Affective-like behavioral tests (sucrose preference, elevated plus, social interaction, light dark test and open field) will be performed at baseline, during peak withdrawal (2–5 days after morphine cessation), and after 4 weeks of abstinence. Significance Our experiments may provide the groundwork for withdrawal therapies targeting MORs on noradrenaline neurons.

While depression affects an estimated 5.7% of adults globally, treatment remains ineffective due to limited understanding of depression’s heterogeneity across patients and its variable symptomology. Therefore in order to increase patient specificity, it is important to understand depression’s various phenotypes, such as facial expression. Facial feature dynamics have been found to correlate to internal cognition, serving as an external readout for internal emotional states. We aim to elucidate and quantify distinct patterns in facial feature dynamics specifically related to internal depressive states. Data was obtained from recordings of a Pavlovian behavioral experiment in which mice were tasked with discriminating between “go” and “no go” audio cues before and after experiencing a murine learned helplessness paradigm (LH) made to mimic depressive helplessness. To obtain data on facial feature position across experimental sessions, we utilized SLEAP (Social LEAP Estimates Animal Poses), a feature tracking and pose estimation software. We found that overall movement across different facial features stalled more in mice exposed to depressive helplessness in comparison to pre-stress sessions, suggesting a rigidity in expressiveness correlated to depressive helplessness. Future research objectives include identifying neural circuitry changes underlying this external facial rigidity and mapping these facial trends on a larger sample size of mice.

Background Varied spinal surgery complexity causes unpredictable operative times and blood loss, which complicates scheduling and resource management. Therefore, a preoperative diagnosis-based model is essential to anticipate intraoperative complexity. This study aims to develop a weighted, data-driven scoring system using preoperative diagnostic factors to predict surgical complexity, proxied by estimated blood loss (EBL). Methods Data from 20 spinal surgery cases were analyzed using a Multivariable Ordinary Least Squares (OLS) regression model. Predictors included surgical level count (Cervical, Thoracic, Lumbar, Sacral) and pathological categories (e.g., Stenosis, Scoliosis), extracted via Python-based keyword algorithms. The model’s predictive power was evaluated using R-squared values and validated against a separate Spring 2025 dataset. Results The model using EBL as the primary evaluator demonstrated high predictive accuracy (R-squared = 0.878). Sacral involvement was identified as the most significant independent predictor (p = 0.017). External validation on the Spring 2025 dataset yielded a lower correlation (R-squared = 0.597). Conclusion This research demonstrates that preoperative diagnostic factors can be quantified to predict intraoperative complexity. While the current scoring system shows promise in identifying high-risk cases for blood loss, further refinement with larger, more diverse datasets is required to improve its transferability.

It is known that memory weakens with age; however, there are promising signs that music therapy can be used as a tool to delay the neurodegeneration that occurs in Alzheimer’s disease. Music is thought to facilitate episodic memory, which is impaired in AD; thus, various findings suggest music as a potential therapeutic intervention to improve cognition in older adults. Music has the ability to increase emotional arousal, which is a key factor in modulating episodic memory within the hippocampus. Older adults show gist versus detail trade-offs in memory, often recalling the gist at the expense of the details. Mnemonic Similarity Tasks (MSTs) are a useful tool to assess hippocampal function in older adults, as they tax hippocampal pattern separation by testing memory for lure items that are highly similar to target items. The current study aims to investigate the effect of music-induced emotional arousal on memory in older adults. Our results will help determine how we can use music therapy as a tool to improve cognitive decline with age. If music-induced emotional arousal increases performance on the MST compared to silence, this is a promising sign of music’s potential as an aid for neurodegenerative disorders.

Individuals with conditions impairing or removing motor function often retain motor imagery neural activity associated with the movements they can no longer perform effectively. This study demonstrates the potential of reliably detecting and classifying pre-motor signals into specific motor intentions in real-time using a non-invasive electroencephalography (EEG) headset. Raw EEG data was collected via OpenBCI while a study participant watched a video with cycles consisting of a 5 second preparation phase, a 5 second movement imagining phase, and a 10 second rest phase. During the imagery phase, the participant was instructed on screen to imagine moving their left hand, right hand, both feet, or tongue. Data epochs from the imagery phase were extracted for classification with a 1-40 Hz bandpass filter and running windows used for preprocessing. Data augmentation was performed to improve data diversity and volume with amplitude manipulation and gaussian noise being used. Validation was strictly performed with unaltered data segments. The EEG Net-GRU model achieved a peak validation accuracy of 0.720. These results demonstrate that classified EEG signals can be effectively applied to assistive devices, such as an online multidirectional cursor for individuals with impaired motor functions. By associating neural pre-motor activity with intent, this study shows how assistive technology can become more responsive for individuals with significant motor impairments, ultimately improving their quality of life.

Introduction. Seizure onset during key periods of language development may negatively affect language outcomes in epilepsy. We examined if earlier age of onset predicts poorer language performance in adults with temporal lobe epilepsy (TLE), and if outcomes differ by seizure side or sex. Methods. Seventy-three patients with medial TLE undergoing presurgical evaluation at Emory University completed neuropsychological testing. Thirty-seven (50.7%) had left-sided and 36 (49.3%) had right-sided seizure onset based on video-EEG. Performance on the D-KEFS letter fluency subtest was measured. Results. Multiple regression analysis was significant [F(5,67)=2.62, p=.032], with age of onset as the only significant predictor (B=0.319, p=.034), explaining 16.3% of variance (R² =.163). Later onset was associated with higher letter fluency scores. Sex, language lateralization, and seizure side were not significant predictors. Independent t-tests showed no significant differences between left and right seizure onset [t(68) = -0.59, p = 0.56] or between males and females [t(71) = 0.41, p = 0.68]. Discussion. Age of seizure onset plays a key role in language outcomes in TLE, and later onset is associated with better performance. These findings highlight the need for speech/language intervention at critical developmental periods, focus on language development, and perhaps early consideration of surgical interventions during malleable periods of brain development for those with intractable seizures.

Congenital heart disease (CHD) is the most common birth defect worldwide, occurring in about 1% of births. Due to alterations in blood flow to the developing brain, CHD has been associated with neurodevelopmental dysfunction and cognitive deficits, which can have lifelong implications. Previous studies exploring the connection between CHD and the brain have focused on the differences between babies with and without CHD. Our study compares babies with CHD to babies with CHD and abnormal neurological findings. Under Dr. Yalda Afshar’s guidance, Dr. Sonam Mokha and I utilized the UC Fetal Consortium (UCFC)’s CHD database of maternal and fetal outcomes from January 2018 to December 2024 and collected data concerning CHD diagnoses and head imaging results, in order to better understand how CHD impacts the course of brain injury. The objective of our retrospective study is to use this data to conduct a descriptive statistical analysis investigating the population of neonates born with CHD and structural brain abnormalities. We divided the cases in two groups based on whether they exhibited normal or abnormal head imaging, then compared the demographics and outcomes of the two groups by conducting t-testing and two-proportion testing on various population statistics in R. We have completed data collection and are currently working on performing statistical analysis. Based on our preliminary results, we expect to find significant differences in birth weight, head circumference, survival rate, and rates of feeding tube dependence.