The Classification of Un-preprocessed ECG Waveforms through the Application of the Hierarchical Temporal Memory Model
Normal and abnormal cardiac function of the human heart can be analyzed through the application of ECG waveform processing and evaluation. Although traditionally the interpretation of these waveforms remains largely a manual effort, as computing power has increased, so too has the application of computational methods for ECG evaluation and classification. One computational method previously applied was the Artificial Neural Network but its application required the addition of signal pre-processing and feature extraction. Recently, a new computational model, Hierarchical Temporal Memory (HTM), has become available for research. This model itself is an attempt to replicate the structural and algorithmic properties of the neocortex of the human brain, thus allowing the utilization of images in both the learning and test databases. It should be noted, current HTM research has advanced from the development of vision hierarchies applied to the recognition of simple line pictographs to vision hierarchies capable of pattern recognition employing photographic images. This study advances the state-of-the-art in ECG waveform classification by developing a system that outperforms currently available classifiers. Furthermore, it advances our understanding of the HTM classification model by gaining a greater understanding of how its various components influence the model’s ability to learn, together with its correlation to human learning and by showing it can be successfully applied to the classification of waveforms.^
John M Casarella,
"The Classification of Un-preprocessed ECG Waveforms through the Application of the Hierarchical Temporal Memory Model"
(January 1, 2013).
ETD Collection for Pace University.