Minimum and Maximum Pattern-Based Self-Organized Feature Engineering: Fibromyalgia Detection Using Electrocardiogram...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Electrocardiogram (ECG) signal analysis is crucial for diagnosing various medical conditions.
• Developing efficient and simple feature extraction techniques is essential for improving ECG classification accuracy.

Purpose of the Study:

• To introduce a novel and straightforward feature extraction function called the minimum and maximum pattern (MinMaxPat).
• To evaluate the classification performance of the proposed MinMaxPat function within a comprehensive feature engineering model for ECG signals.

Main Methods:

• The MinMaxPat function divides ECG signals into overlapping blocks to identify minimum and maximum value indices.
• A feature map in base 16 is generated from these indices, and its histogram forms a 256-length feature vector.
• A feature engineering model incorporating MinMaxPat, cumulative weight-based iterative neighborhood component analysis (CWINCA) for feature selection, and a t-algorithm-based k-nearest neighbors (tkNN) classifier was developed.

Main Results:

• The MinMaxPat-based feature engineering model was applied to a public ECG fibromyalgia dataset.
• The model achieved classification accuracy exceeding 80% using both leave-one-record-out (LORO) cross-validation (CV) and 10-fold CV.
• Three distinct cases were analyzed, demonstrating consistent high performance.

Conclusions:

• The proposed MinMaxPat feature extraction method, integrated into a simple model, yields high classification performance for ECG signals.
• The findings highlight the surprising effectiveness of this straightforward approach for ECG signal classification tasks.