Research Information System
2. Ulusal Nörogörüntüleme Kongresi, Ankara, Turkey, 11 - 13 September 2025, pp.75, (Summary Text)
Objective: The aim of this study is to classify EEG data obtained simultaneously with fMRI in eyes open (EO) and eyes closed (EC) conditions using a Long Short-Term Memory (LSTM) neural network. Methods: The study included 200 healthy volunteer participants. Data were recorded simultaneously with an MRI-compatible EEG system, which comprised 33 channels, including one ECG recording. Preprocessing steps were applied to the EEG data to improve signal quality and enhance model learning. The ECG channel was then removed from the signal. During the segmentation phase, the data was divided into 2-second epochs and fed into the LSTM model as input. A 5-fold cross-validation method was used for model training and evaluation. This method provided a more reliable test of the model's generalizability and accuracy. At each fold, the model's performance was evaluated using accuracy, precision, recall, and F1 score metrics. Results: The performance of the LSTM-based model was evaluated using a 5-fold cross-validation method. The model achieved very good performance with 96.14% accuracy, 94.83% sensitivity, 96.33% recall, and 95.56% F1 score values. Conclusion: The results confirm that the LSTM architecture can successfully learn time-dependent EEG patterns and can be used as an effective method for classifying such mental states. Furthermore, the automatic classification of simple mental states, such as eye open/closed, has applicability in many areas, including clinical diagnostics, brain-computer interfaces, and assessment of consciousness.