Scalogram Image-Based Texture Features for Robust Epileptic Seizure Detection Under Noisy Conditions

Accurate detection of epileptic seizure events is an extremely challenging problem due to the complex and highly variable nature of seizure activity, which can manifest differently in each individual. This study employs texture-based descriptors—Local Binary Patterns (LBP), Gray Level Co-occurrence Matrix (GLCM), and Histogram of Oriented Gradients (HOG)—extracted from scalogram images and classified using a Random Forest model, providing both interpretable and efficient alternative to deep learning models. The proposed approach is evaluated on two publicly available benchmark datasets, demonstrating near-perfect classification accuracy and F1-score in distinguishing epileptic from healthy EEG segments and effectively classifying ictal, interictal, and pre-ictal states. This task of identifying ictal, pre-and interictal states becomes comparatively more challenging, as it requires the model to differentiate between subtle variations in brain activity that occur during different phases of the seizure cycle. Further, the model maintains its strong performance even under noisy conditions, with classification accuracy unaffected by the addition of Gaussian noise up to an SNR of -10 dB. Notably, texture-based features have shown significant potential for EEG-based seizure detection, reinforcing their relevance for real-world applications.