Quantitative EEG Assessment of Dependence-Related Neurophysiological Patterns Using Rule- and Score-Based Modeling in Substance Use Disorders

1. Introduction

Substance use disorders (SUDs) represents a complex neurobehavioral disorder underpinned by maladaptive neural plasticity and dysregulation of cortical arousal systems. Electroencephalography (EEG) [1], with its high temporal resolution, provides a powerful non-invasive means of capturing these alterations in oscillatory brain dynamics. Over the past two decades, EEG studies have consistently demonstrated that individuals with SUDs exhibit characteristic deviations in resting-state and task-related spectral activity, particularly within the alpha, beta, and theta frequency bands [2,3,4].

Among the most replicated findings is a pattern of increased beta activity accompanied by reduced alpha power, reflecting cortical hyperarousal and impaired inhibitory control. Sokhadze et al. [2] observed that substance-dependent individuals display excessive beta synchronization and attenuated alpha rhythms, indicative of excitatory dominance within the cortical network. Similarly, Rangaswamy et al. [5] linked heightened beta power to relapse vulnerability and deficient self-regulation. Complementary results from Ceballos et al. [6] and Bel-Bahar et al. [7] further support that this imbalance persists even during resting conditions, suggesting that hyperarousal constitutes a stable neurophysiological trait rather than a transient withdrawal effect.

The theta-to-beta ratio (TBR) has emerged as an integrative marker of cortical activation and attentional regulation. Low TBR values (< 1.0) are typically associated with dominance of high-frequency beta oscillations over slower theta waves, a pattern linked to impulsivity and hypervigilance [8,9]. Conversely, normalization or elevation of TBR corresponds to improved inhibitory control and relaxed attentional states [10]. These findings outline a “hyperaroused–disinhibited” EEG phenotype frequently observed in chronic substance users.

Another sensitive marker of cortical regulation is alpha blocking, the suppression of alpha power upon eye opening [11]. In healthy adults, strong alpha blocking reflects adaptive sensory gating and intact cortical inhibition. In contrast, impaired or absent alpha suppression among dependent individuals points to weakened top-down inhibitory control and reduced cortical reactivity [7]. Supporting evidence from Fathi et al. [12] demonstrates that restoring alpha coherence and reducing excessive beta power through neurofeedback interventions can normalize cortical excitability and enhance abstinence outcomes.

Building upon this body of evidence, we implemented a rule-based quantitative EEG framework to identify dependence-related neurophysiological patterns. Each EEG record was evaluated across multiple spectral parameters, including Theta/Beta Ratio < 1.0, Alpha Relative Power < 0.30, Beta Relative Power > 0.25, Hyperarousal Index ≥ 2.0, and Alpha Blocking = Impaired. EEGs meeting all these criteria were classified as Dependence-Likely, representing the hyperaroused–disinhibited cortical state consistently described across prior studies. This standardized computational approach enables interpretable and reproducible detection of d fıstıkependence-linked EEG features and offers a promising neurophysiological framework for precision assessment and monitoring in addiction research.

2. Materials and Methods

2.1. Ethical Consideration

The study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of {Blinded University} (protocol code: 24-177, date of approval: 2024-10-17).

2.2. Study Design and Participants

This retrospective study included EEG recordings and clinical data from 47 individuals diagnosed with SUDs who were admitted to the addiction treatment unit of Moodist Hospital between 2024-2025. EEG data from patients diagnosed with SUDs who presented to our psychiatry clinic were retrospectively included in the study if their records contained available EEG recordings. Patients with a history of epilepsy, neurodegenerative disorders, or other neurological or systemic conditions that could potentially affect EEG patterns were excluded. This approach ensured that the analyses specifically reflected electrophysiological alterations associated with substance use.

Demographic and clinical information was obtained from patient medical records and included age, gender, educational and marital status, employment and economic status, physical health, history of psychiatric treatment or medication use, history of suicide attempt, and type and frequency of substance use. Additional variables such as previous addiction treatment and family history of addiction were also recorded.

All participants were diagnosed with SUDs according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria [13]. Individuals with major neurological disorders, severe head trauma, or incomplete EEG data were excluded from the analysis.

2.3. EEG Recording Procedure

EEG recordings were obtained using a Neurosoft Neuro-NRS “Neuron-Spectrum” system (Neurosoft, Ivanovo, Russia) with 19 scalp electrodes placed according to the international 10–20 system. Data were stored in European Data Format (.edf) files [14]. All recordings were visually inspected for quality, and channels containing excessive noise or artifacts were excluded prior to analysis. EEG data preprocessing and quantitative analyses were performed using MNE-Python (v0.14) [15].

Each EEG session consisted of seven standardized segments: background activity, eyes open, eyes closed, hyperventilation, after hyperventilation, flash stimulation, and after flash stimulation [16]. Background, eyes open, and eyes closed segments were successfully obtained in all cases (100%). Hyperventilation and after-hyperventilation segments were present in 95.7% of recordings, while flash stimulation and after-flash stimulation segments were available in 91.5% and 61.7% of cases, respectively (shown in Table 1).

The mean durations of EEG segments were as follows: background activity – 149 s, eyes open – 73 s, eyes closed – 222 s, hyperventilation – 172 s, after hyperventilation – 277 s, flash stimulation – 333 s, and after flash stimulation – 473 s (shown in Figure 1). Segment boundaries were visually confirmed to ensure consistent epoching across participants. Participants were instructed to remain still, avoid blinking during stimulation, and maintain regular breathing to minimize artifacts throughout the recording.

2.4. Channel Selection and Signal Analysis

EEG recordings were obtained from 19 scalp electrodes following the international 10–20 system. For the present analysis, a subset of frontal and central electrodes (FZ–CZ, FP1–F3, FP1–F7, FP2–F4, FP2–F8, F3–C3, F4–C4, C3–P3, C4–P4, CZ–PZ) was selected [16]. This selection was based on prior evidence indicating that frontal regions are critically involved in executive control, attention, and inhibitory processing—functions frequently impaired in addiction—while central regions reflect sensorimotor and cortical arousal activity, providing complementary insights into brain regulation [17].

The raw EEG signals were preprocessed using MNE-Python. After filtering to remove artifacts and noise, power spectral density (PSD) was estimated using Welch’s method (Welch, 1967) as implemented in the MNE-Python library [18,19]. From the PSD, absolute and relative power values were extracted for the delta (1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), and beta (13–30 Hz) frequency bands. Additionally, the theta/beta power ratio (TBR) was calculated as an indicator of cortical arousal and attentional control.

To facilitate statistical analysis and reduce feature dimensionality, band powers from all frontal and central channels were averaged separately, resulting in mean frontal and central values for each frequency band and for the TBR metric (shown in Table 2). These aggregated measures were then exported to an Excel dataset for further group-level statistical comparisons. A schematic overview of the complete EEG data processing and feature extraction workflow is illustrated in Figure 2.

2.5. Derived EEG Feature Extraction

From the processed EEG data, a set of derived quantitative indices was calculated to capture individual differences in cortical reactivity, arousal balance, and dependence-related neurophysiological patterns. The most relevant measures included the Alpha Blocking Ratio (eyes-closed / eyes-open alpha power), the Theta/Beta Ratio (TBR) as an indicator of attentional control and cortical arousal, the Hyperarousal Index (beta / alpha power ratio), and the Cortical Reactivity Index (CRI) reflecting alpha modulation during stimulation and recovery periods.

In addition, composite variables such as the Dependence Likelihood and Homeostatic Recovery Score were derived to represent higher-order patterns related to physiological regulation and addiction-related EEG responses. A complete list of the derived EEG variables, their definitions, and calculation methods is provided in Table 3.

2.6. Dependence Likelihood Determination and Scoring Framework

To quantify addiction-related neurophysiological patterns, a composite measure termed Dependence Likelihood was established. This index integrates multiple EEG-derived parameters associated with cortical arousal, inhibitory control, and self-regulatory balance, domains frequently altered in SUDs.

The rule-based definition was as follows:

“Yes”: (TBR < 1.0) AND (Background Activity Alpha Relative Mean < 0.30) AND (Background Activity Beta Relative Mean > 0.25) AND (Hyperarousal Index ≥ 2.0) AND (Alpha Blocking Intact = No)

“No”: Otherwise

Each threshold reflects established physiological evidence: low theta/beta ratio and low alpha power indicate insufficient cortical inhibition; high beta power and elevated hyperarousal index indicate excessive excitatory drive; and absent alpha blocking denotes deficient cortical modulation. Collectively, these parameters represent a state of hyperarousal with impaired inhibitory regulation, characteristic of dependence-related EEG profiles.

In addition to the rule-based classification, a continuous Dependence Score was calculated to quantify the multidimensional contribution of EEG features on a normalized scale (0–1). Variable definitions and normalization approaches are presented in Table 4.

The composite score was computed as:

Dependence Score = 0.25 × (1 − TBR_norm) + 0.25 × Beta_norm + 0.25 × HRI_norm + 0.15 × (1 − Alpha_norm) + 0.10 × AlphaBlocking_def

This formulation weights inhibitory deficits (low TBR and low alpha) and excitatory dominance (high beta and hyperarousal) more heavily, while incorporating alpha blocking as a modulatory factor.

For interpretability, a combined decision rule was applied:

Dependence-related if Dependence Score ≥ 0.7 or Dependence Likelihood = “Yes”

Borderline if 0.5 ≤ Dependence Score < 0.7

Normal pattern if Dependence Score < 0.5

This hybrid system allows both rule-based and quantitative assessment, improving classification robustness by aligning categorical thresholds with continuous neurophysiological gradients. (Figure 3) The descriptive results are summarized as Dependence Likelihood Result in Table 5.

2.7. Statistical Analyses and Tools

All statistical analyses were performed using Wistats v3.0 (WisdomEra Corp., Istanbul, Turkey), an integrated Python-based platform for statistical evaluation and machine learning. The software incorporates key analytical libraries including SciPy, scikit-learn, and statsmodels, allowing seamless execution of both classical and data-driven analyses within the same computational environment.

Descriptive statistics were calculated for all demographic, clinical, and EEG-derived variables. Comparative analyses between groups or EEG-based categories were performed using appropriate parametric or non-parametric tests according to data distribution, as assessed by Shapiro–Wilk, skewness, and kurtosis tests. Chi-square or Fisher’s exact tests were used for categorical variables, while independent-samples t-test, Mann–Whitney U, or Kruskal–Wallis tests were applied for continuous variables. Correlation analyses were conducted using Spearman’s or Pearson’s coefficients as appropriate. Statistical significance was set at p < 0.05.

3. Results

3.1. Participant Characteristics

A total of 47 patients diagnosed with SUDs were included in the study. Case characteristics are presented in Table 6. The majority of participants were male (76.6%), with a mean age of 31.6 years. Most participants were single (59.6%), had completed university or high school education (53.2% and 27.7%, respectively), and reported medium-to-high economic status (93.6%). Approximately half of the sample were employed (53.2%) and living with their families (46.8%). A psychiatric treatment history was present in 83.0% of cases, and 72.3% reported previous psychiatric drug use. Suicide attempts were reported by 6.4%. Family history of addiction was found in 12.8% of participants. The most common addictive substances were alcohol (48.9%), cannabis (42.6%), and cocaine (23.4%); poly-substance use (≥2 substances) was observed in 42.5% of cases.

3.2. Derived EEG Variables

Quantitative EEG analysis revealed that all participants demonstrated low alpha power status (100%) and high hyperarousal index values (mean = 3.45). The dependence-related EEG pattern was observed in 87.2% of cases, while 8.5% exhibited borderline patterns and 4.3% showed normal findings. The mean dependence score was 0.86, indicating pronounced cortical arousal and imbalance between alpha and beta activity. (Figure 4)

Regarding specific EEG indices, 53.2% of patients showed a hyperaroused dependence pattern, whereas 46.8% were classified as balanced. The alpha blocking ratio averaged 1.60, and 46.8% of participants retained intact alpha blocking. After-hyperventilation alpha recovery was low in 61.7%, and after-flash alpha rebound was absent in 70.2%.

3.3. Segment-Based Spectral Power Analysis

PSD analyses calculated by Welch’s method from frontal and central electrodes demonstrated the following mean theta/beta ratios across conditions: background 0.63, eyes open 0.54, eyes closed 0.54, hyperventilation 0.37, after hyperventilation 0.55, flash stimulation 0.65, and after flash stimulation 0.53. Alpha relative power remained consistently low (0.07–0.09) across all conditions, while beta relative power was elevated (0.22–0.29), consistent with a hyperarousal profile. Gamma and delta relative powers were generally low (0.02–0.15). These findings reflect sustained cortical activation and reduced inhibitory regulation during both baseline and stimulation segments. (Figure 5)

Collectively, EEG findings in patients with SUDs indicate a predominant hyperarousal pattern, characterized by decreased alpha activity, elevated beta power, and low theta/beta ratios during hyperventilation. The quantitative indices support the presence of dependence-related cortical dysregulation, consistent with neurophysiological models of addictive behavior (shown in Table 5).

4. Discussion

The present study identified a consistent hyperaroused–disinhibited cortical profile among individuals with SUDs, characterized by low alpha power, elevated beta activity, and reduced theta/beta ratios across EEG segments. Quantitative indices revealed that 87.2% of participants exhibited dependence-related EEG patterns, while the mean Dependence Score (0.86) indicated pronounced cortical excitation and diminished inhibitory regulation. These findings align with the theoretical framework of addiction as a state of sustained cortical hyperarousal and impaired top-down control, supporting the potential of EEG-based biomarkers for objective assessment of dependence severity.

Prior EEG studies in addiction have consistently reported increased beta synchronization and reduced alpha rhythms, reflecting excessive cortical excitability and deficient inhibitory control [2,3,6]. Similarly, the TBR has emerged as a sensitive marker of attentional regulation and arousal balance, where low TBR values (<1.0) indicate impulsivity and hypervigilance [8,9]. The current findings replicate and extend these results by demonstrating that such spectral imbalances persist across multiple stimulation conditions, including hyperventilation and flash segments, suggesting a trait-like neurophysiological signature rather than a transient state effect.

The present findings regarding the theta/beta ratio (TBR) align with recent evidence emphasizing its role as a sensitive marker of cortical arousal and attentional regulation. A study titled “The EEG Theta/Beta Ratio: A Marker of Arousal or Cognitive Processing Capacity?” demonstrated that lower TBR values are associated with increased cortical excitation, impulsivity, and hypervigilance, reflecting a dominance of fast-frequency beta oscillations over slower theta rhythms [20]. This interpretation supports the current observation that participants exhibiting reduced TBR values also presented elevated beta power and diminished alpha activity, consistent with a hyperaroused–disinhibited cortical profile. Together, these findings reinforce the view that diminished TBR may represent a stable electrophysiological correlate of impaired inhibitory control in substance use disorders.

Unlike prior research that analyzed single spectral parameters or opaque multivariate models, this study introduces a transparent, mathematically defined composite score integrating inhibitory (TBR, alpha) and excitatory (beta, hyperarousal) features with a modulatory alpha-blocking term. The Dependence Score formula

(0.25×(1−TBR_norm) + 0.25×Beta_norm + 0.25×HRI_norm + 0.15×(1−Alpha_norm) + 0.10×AlphaBlocking_def)

enables interpretable quantification of cortical dysregulation on a 0–1 scale, with explicit thresholds for clinical interpretability. To our knowledge, comparable weighted and normalized scoring systems have not been explicitly described in addiction EEG literature [3,6].

The combined rule-based and continuous scoring approach offers multiple advantages:

Interpretability: Clear weighting and thresholding allow replication and clinical translation.

Granularity: Borderline scoring (0.5–0.7) captures intermediate regulation states often missed by binary classification.

Reproducibility: The explicit formula enables comparison across datasets and laboratories.

This framework thus bridges the gap between traditional spectral EEG analysis and clinically usable biomarkers, potentially supporting relapse risk assessment and personalized treatment monitoring.

While the weighting scheme was guided by empirical rationale, it remains heuristic and should be validated in larger, independent cohorts. The sample size was modest and restricted to patients with established SUD; future studies should examine whether the same scoring framework discriminates between active users, abstinent individuals, and healthy controls. Moreover, optimization of weights via machine-learning algorithms and evaluation of test–retest reliability would strengthen the robustness of this model.

5. Conclusions

In conclusion, this study provides empirical evidence that cortical hyperarousal and reduced inhibitory regulation dominate the EEG profile of individuals with SUD. The proposed Dependence Score and decision-rule system offers a novel, interpretable, and standardized method for quantifying these neurophysiological patterns. Future validation studies may establish its value as an EEG-based biomarker for clinical monitoring and treatment response in addiction research.