The impact of vaginal bromocriptine on reducing pain and menstrual bleeding in women with adenomyosis: a randomized controlled trial.

Electroencephalography (EEG) provides a way to monitor and measure brain activity by recording the electrical signals in the brain which result from mental activities and help us understand how we think or feel [, ]. Although this modality is often used for the early diagnosis of neurological diseases, it has also been widely applied in cross-disciplinary areas of artificial intelligence, education, health and human-machine interactions [–]. In this context, EEG signals provide comprehensive and valuable information about mental processes by directly measuring the electrical activity of the brain []. In particular, the detection of cognitive disorders and emotion recognition are among the popular research areas using EEG signals []. In recent years, mental performance measurement with EEG signals has also become an increasingly popular approach []. The number of applications using objective measures of mental states and performance is numerous and very important. As such, several potential application areas have arisen, such as identifying the abilities of individuals, the early diagnosis of cognitive disorders and improving educational processes [, ]. Several spatial and geometric approaches have been proposed for EEG feature extraction in the literature. Common Spatial Patterns (CSP) [] and its variants extract discriminative spatial filters by maximizing variance ratios between classes; however, CSP relies on frequency band pre-selection and requires training-phase optimization of spatial filter parameters, resulting in O(Ch) computational complexity for covariance matrix decomposition. Riemannian geometry-based methods [] operate on the manifold of symmetric positive definite (SPD) covariance matrices and have shown strong performance in BCI applications; yet they require covariance matrix estimation, which introduces quadratic complexity O(Ch × L) and sensitivity to short epochs and noise. Functional connectivity measures such as phase locking value (PLV) [] and coherence [] capture inter-channel relationships but require frequency decomposition and are computationally intensive for real-time applications. Pattern-based approaches including Local Binary Pattern (LBP) adaptations for EEG [] capture local signal texture but typically operate on individual channels independently without modeling cross-channel interactions. In contrast, the proposed DiffPat method offers several distinct advantages over these existing approaches: (i) DiffPat is entirely deterministic and training-free—it requires no parameter optimization, spatial filter learning, or covariance estimation; (ii) it operates with linear time complexity O(L), making it significantly more efficient than CSP (O(Ch)), Riemannian (O(Ch × L)), and connectivity-based methods; (iii) unlike single-channel pattern methods (e.g., LBP), DiffPat inherently captures inter-channel relationships through its transition table mechanism (Eqs. –), generating a 32 × 32 connectivity-like representation without explicit frequency decomposition; (iv) the transition table structure directly supports explainability, as each feature maps to a specific channel-to-channel interaction, enabling DLob-based symbolic interpretation and cortical connectome visualization—a capability absent in CSP, Riemannian, and most connectivity-based methods. These properties position DiffPat as a lightweight, transparent, and interpretable alternative to existing spatial feature extraction paradigms.