ECG Heartbeat Classification Project

Introduction

Electrocardiogram (ECG) recordings are critical for diagnosing various heart conditions. This project focuses on developing a classification model to distinguish between normal heartbeats and myocardial infarctions using ECG data. The dataset consists of 100 samples, each containing 96 data points representing the electrical activity recorded during one heartbeat.

Methodology

Data Preprocessing and Exploration:

Loading and Initial Processing: Loaded the dataset with 100 samples labeled as -1 (normal heartbeat) and +1 (myocardial infarction). Addressed missing values and ensured data integrity.

Data Visualization: Conducted exploratory data analysis using histograms, box plots, and line plots to understand class distribution and identify patterns.

SMOTE Oversampling: Applied SMOTE to address class imbalance, enhancing the model's ability to learn from minority class instances.

Normalization and Train-Test Split:

Normalized the data to ensure feature scaling consistency.

Split the dataset into training and test sets for model evaluation.

Model Training and Evaluation:

Algorithms Used: Trained multiple classification algorithms, including K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Random Forest, Convolutional Neural Networks (CNN), and Long Short-Term Memory (LSTM) networks.

Advanced Preprocessing: Utilized advanced filtering methods like the Savitzky-Golay filter and Bandpass filter to improve signal quality and reduce noise.

Feature Extraction: Employed Wavelet Transformation for feature extraction from ECG signals.

Model Selection:

Evaluated models based on performance metrics such as accuracy, precision, recall, and F1-score.

Best Model: The Support Vector Machine (SVM) was selected as the best-performing model, demonstrating superior evaluation metrics.

Conclusion

This project explored various preprocessing techniques and classification algorithms to optimize the performance of ECG heartbeat classification. The SVM model was ultimately selected for its superior performance, underscoring the significance of thorough data preprocessing and careful model selection in ECG classification tasks.

Challenges Encountered

Managing class imbalance in the dataset.

Reducing noise and enhancing signal quality.

Selecting and fine-tuning models for optimal performance.