Diagnosing epileptic seizures using combined features from independent components and prediction probability from EEG...

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Epileptic seizures are neurological events with significant injury risks.
• Early detection of epileptic seizures is vital for timely treatment and improved patient outcomes.
• Advanced artificial intelligence approaches are needed for early epileptic seizure disorder detection.

Purpose of the Study:

• To design a novel ensemble approach for high-performance, early detection of epileptic seizures.
• To investigate the efficacy of a proposed approach using electroencephalography (EEG) data.
• To address the research gap by combining features for improved seizure detection performance.

Main Methods:

• A novel ensemble approach, Fast, Independent Component Analysis Random Forest (FIR), was developed.
• The FIR model extracts independent components and prediction probability features from EEG data.
• The proposed model integrated Independent Component Analysis (ICA) with prediction probability to enhance accuracy.

Main Results:

• The FIR ensemble model demonstrated superior performance compared to original features.
• The FIR combined with Support Vector Machine (FIR+SVM) achieved an accuracy of 98.4% for epileptic seizure detection.
• The use of combined features significantly improved detection performance over single feature sets.

Conclusions:

• The proposed FIR approach shows potential for the early diagnosis of epileptic seizures.
• This AI-driven method can enhance detection and facilitate timely interventions in the medical industry.
• The study highlights the effectiveness of ensemble methods and combined features in improving seizure detection accuracy.