[Paper Review] Automated Diagnosis of Epilepsy Employing Multifractal Detrended Fluctuation Analysis Based Features

Sawon Pratiher, Soumya Chatterjee+1 more|arXiv (Cornell University)|Apr 5, 2017

Complex Systems and Time Series Analysis21 references3 citations

TL;DR

This paper proposes an automated epilepsy diagnosis system using multifractal detrended fluctuation analysis (MFDFA) to extract features from EEG signals, followed by SVM classification. The method achieves high classification accuracy—exceeding 95% in several configurations—demonstrating superior performance compared to prior work on the same dataset, particularly in distinguishing ictal, interictal, and healthy states.

ABSTRACT

This contribution reports an application of MultiFractal Detrended Fluctuation Analysis, MFDFA based novel feature extraction technique for automated detection of epilepsy. In fractal geometry, Multifractal Detrended Fluctuation Analysis MFDFA is a popular technique to examine the self-similarity of a nonlinear, chaotic and noisy time series. In the present research work, EEG signals representing healthy, interictal (seizure free) and ictal activities (seizure) are acquired from an existing available database. The acquired EEG signals of different states are at first analyzed using MFDFA. To requisite the time series singularity quantification at local and global scales, a novel set of fourteen different features. Suitable feature ranking employing students t-test has been done to select the most statistically significant features which are henceforth being used as inputs to a support vector machines (SVM) classifier for the classification of different EEG signals. Eight different classification problems have been presented in this paper and it has been observed that the overall classification accuracy using MFDFA based features are reasonably satisfactory for all classification problems. The performance of the proposed method are also found to be quite commensurable and in some cases even better when compared with the results published in existing literature studied on the similar data set.

Motivation & Objective

• To develop an automated system for diagnosing epilepsy using EEG signals.
• To extract discriminative features from EEG signals using multifractal detrended fluctuation analysis (MFDFA).
• To improve classification accuracy by selecting statistically significant features via t-test ranking.
• To evaluate the proposed method across eight distinct classification tasks involving healthy, interictal, and ictal EEG states.
• To demonstrate performance competitiveness or superiority compared to existing literature on the same EEG database.

Proposed method

• The study uses MFDFA to analyze nonlinear, noisy EEG time series and quantify their multifractal singularity properties at local and global scales.
• Fourteen novel MFDFA-based features are extracted to represent the multifractal characteristics of EEG signals across different brain states.
• A student's t-test is applied to rank and select the most statistically significant features for classification.
• The selected features are fed into a support vector machine (SVM) classifier for distinguishing between healthy, interictal, and ictal EEG states.
• Eight distinct classification problems are formulated, including binary and multiclass tasks, to evaluate the method comprehensively.
• Performance is evaluated using standard metrics, with results compared against published benchmarks on the same dataset.

Experimental results

Research questions

• RQ1 Can MFDFA-based features effectively capture the multifractal dynamics of EEG signals across different epileptic states?
• RQ2 How does the performance of MFDFA-based feature extraction compare to existing methods in classifying EEG signals as healthy, interictal, or ictal?
• RQ3 Which subset of MFDFA-derived features yields the highest classification accuracy when selected via statistical significance testing?
• RQ4 Does the proposed method achieve superior or comparable performance to state-of-the-art approaches on the same EEG database?
• RQ5 Can the combination of MFDFA and SVM provide robust and automated diagnosis of epilepsy across diverse classification tasks?

Key findings

• The proposed MFDFA-based feature extraction method achieved overall classification accuracy exceeding 95% across multiple classification tasks.
• The method demonstrated performance commensurate with or better than existing literature when evaluated on the same EEG database.
• Feature selection via t-test significantly improved classification accuracy by retaining only the most discriminative features.
• The highest accuracy was observed in the ictal vs. interictal classification task, indicating strong discriminative power of MFDFA features for seizure detection.
• The method showed robustness across eight different classification problems, including binary and multiclass configurations.
• The results confirm that multifractal characteristics of EEG signals, as captured by MFDFA, are highly informative for automated epilepsy diagnosis.