Analyzing EEG Data During Opium Addiction Treatment Using a Fuzzy Logic-Based Machine Learning Model

Elnaz DehAbadi¹Fateme Ayşin Anka²Fateme Vafaei^3,4Hossein Lanjanian^5*Sajjad Nematzadeh⁶Mahsa Torkamanian-Afshar⁶Nazanin Aghahosseinzargar¹Farzad Kiani^7,8Peyman Hassani Abharian^9*

• ¹Garmsar Branch, Islamic Azad University, Garmsar, Iran
• ²Political Sciences and Public Administration Dept., Faculty of Economics, Administrative and Social Sciences, Istinye University, Istanbul, Türkiye
• ³Department of Biomedical Engineering, Shahed University, Tehran, Iran
• ⁴Basir Eye Health Research Center, Iran University of Medical Sciences, Tehran, Iran
• ⁵Research Institute for Endocrine Science, Shahid Beheshti University of Medical Sciences, Tehran, Iran
• ⁶Software Engineering department, engineering faculty, Istanbul Topkapi University, Istanbul, Türkiye
• ⁷Computer Engineering Dept., Faculty of Engineering, Fatih Sultan Mehmet Vakif University, Istanbul, Türkiye
• ⁸Data Science Application and Research Center, Fatih Sultan Mehmet Vakif University, istanbul, Türkiye
• ⁹Institute for Cognitive Science Studies (IRICSS), Department of Cognitive Psychology and Rehabilita-tiont, Tehran, Iran

Abstract Background: Reliable noninvasive tools for assessing substance abuse treatment and predicting outcomes remain a challenge. We believe EEG-derived complexity measures may have a direct link to clinical diagnosis. To this aim, our study involved a psychological investigation of four groups of current and former male opium addicts. Furthermore, we propose a machine learning (ML) model incorporating fuzzy logic to analyze EEG data and identify neural complexity changes associated with opium addiction. Method: Male participants were categorized into four groups: active addicts, those with less than three days of treatment, those treated for over two weeks, and healthy controls. Psychological assessments evaluate mental health and addiction status. EEG data were collected using standardized electrode placement, preprocessed to remove noise, and analyzed using the Higuchi Fractal Dimension(HFD) to quantify neural complexity. Feature selection methods and ML classifiers were applied to identify key patterns distinguishing addiction stages. Results: Distress levels varied significantly across groups and persisted post-quitting. Addicts exhibited poorer general health than controls, though treatment led to improvements. Significant differences in neural complexity were observed in brain regions linked to attention, memory, and executive function. The ML model effectively classified addiction stages based on EEG-derived features. Conclusion: This study demonstrates the potential of ML and fuzzy logic in assessing addiction-related neural dynamics, offering insights into opioid addiction's pathophysiology. The findings highlight the promise of brainwave-based biomarkers for personalized addiction diagnosis and treatment monitoring.