Automated accurate schizophrenia detection system using Collatz pattern technique with EEG signals. (September 2021)

• Record Type: Journal Article
• Authors: Baygin, Mehmet; Yaman, Orhan; Tuncer, Turker; Dogan, Sengul; Barua, Prabal Datta; Acharya, U. Rajendra
• Is Part Of: Biomedical signal processing and control. Volume 70(2021)
• Publication Date: 2021-09
• Subjects: Collatz pattern -- Schizophrenia diagnosis -- EEG processing -- Maximum absolute pooling -- Iterative NCA; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.102936
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: New generation EEG classification model is presented using the Collatz pattern. Maximum absolute pooling is utilized as a decomposition method. Obtained state-of-art results for schizophrenia classification. Universal success of this model is shown using two datasets. Abstract: Background: Schizophrenia (SZ) is one of the prevalent mental ailments worldwide and is manually diagnosed by skilled medical professionals. Nowadays electroencephalogram (EEG) signals-based machine learning methods have been proposed to help medical professionals. Materials and method: In this work, we have proposed a novel Collatz conjecture-based automated schizophrenia detection model using EEG signals. The objectives of the presented model are to show the feature generation ability of conjecture-based structure and present a highly accurate EEG-based schizophrenia detection model with low time burden. Our presented model comprises three stages. (i) New feature generation function is presented using Collatz Conjecture, and named as Collatz pattern. Combination of Collatz pattern and maximum absolute pooling decomposer, a new multilevel feature generation method is employed to extract both low-level and high-level features. (ii) The iterative neighborhood component analysis (INCA) is employed on the selected features to select the clinically significant features. (iii) The chosen features are fed to k nearest neighbors (KNN) classifier for automated deetction of SZ. Results: Our developed Collatz conjecture-based automated SZ detection model is validated using two public schizophrenia databases with 19 and 10 channels corresponding to database-1 (DB1) and database-2 (DB2) datasets, respectively. We have obtained the classification accuracy of 99.47% and 93.58% for DB1 and DB2 datasets, respectively, with ten-fold cross-validation strategy. Conclusions: Our developed model is accurate and robust in detecting SZ using EEG signals. Our deevloped automated system is ready for clinical usage in hospitals and polyclinics to assist clinicians in their diagnosis as an adjunct tool.