Heartbeat Classification

Classify ECG heartbeat signals into predefined categories based on heartbeat abnormality by transforming time series to images

Datasets:

• MIT-BIH Dataset (https://www.physionet.org/content/mitdb/1.0.0/): contains 5 categories: N, S, V, F, Q (see https://arxiv.org/pdf/1805.00794.pdf for details), 87k train records, 21k test records
• PTB Dataset (https://www.physionet.org/content/ptbdb/1.0.0/): contains 2 categories: Normal and Abnormal, 14k records

ECG heartbeat signals are time series data (187 signals per heartbeat in the given data)

Approaches:

• Gramian Angular Field (GAF)
• Recurrence Plot (RP)
• Markov Transition Field (MTF)
• Intermediate fusion