Assessment of different optimizers on recurrent neural networks for electrocardiogram (ECG) classification Available to Purchase

Biomedical signals are collections of data considering psychological signals, they can be in the form of sound like opening and closing of heart valve, respiratory sound, and many others along with some electrical signals like Electrocardiogram (ECG), Magnetic resonance imaging (MRI), Gait Rhythm (GR), electrooculogram (EOG), Electromyography (EMG), all of these can be obtained using medical observatory machines. Many times, Doctor’s fail to detect the presence of a disease manually as a result, some progressive diseases worsen a person’s condition to an unrecoverable phase. With the advancement of technology, it has been made possible for us to analyse such data using ML algorithms. To classify the disease, biological signals were obtained, denoised, and feature extraction has been performed. In this experiment, an ECG was analysed to determine whether a person had cardiac disease. The early detection of heart disease made possible by this study will assist in saving a great number of lives.Almost around 20 thousand people have died in 2014 just because of heart attacks nationwide and the numbers reached to 28 thousand in 2019. The data also indicate that such occurrences have steadily increased across all age categories, except for those under 14 and those between the ages of 14 and 18. Beginning in 2016, NCRB began releasing information on these deaths by age group. CDC analyzed that a cardiovascular disease-related death occurs in the US every Thirty-Four second. In the United States, around seven lakh deaths in 2020 were caused due to heart related diseases—1  in every 5 fatalities. So, there was a need to build an ML model that will be efficient enough in detecting heart diseases at an early stage by analyzing the biomedical signals from the heart, that are known as ECG or ELECTROCARDIOGRAM signals. Many machine learning optimizers namely SGD, AdaDelta, Adagrad, Nadam, FTRL, Adam were studied and used in this research work. A comparative analysis was made between these optimizers to choose the most efficient one. Out of all these, the ADAM optimizer outperformed and gave us 98.90% accuracy.