Research topic of signal processing PPT for help

Research topic of signal processing PPT

1. Presentation on progress report Pooja Sharma Enrollment Id: 150001001224 PhD scholar, CSE Supervisor ‘s name : Dr. Shail Kumar Dinkar

2. Progress Summary Course work completed. State of Art seminar completed. Publications: Sharma, P., & Dinkar , S. K. (2022). A Linearly Adaptive Sine–Cosine Algorithm with Application in Deep Neural Network for Feature Optimization in Arrhythmia Classification using ECG Signals.  Knowledge-Based Systems ,  242 , 108411. 2. Sharma, P., Dinkar , S. K., & Gupta, D. V. (2021). A novel hybrid deep learning method with cuckoo search algorithm for classification of arrhythmia disease using ECG signals.  Neural Computing and Applications ,  33 (19), 13123-13143.

3. Publications contd. 3. Sharma, P., Dinkar , S. K., & Deep, K. (2021). Optimized Convolutional Neural Network-Based Classification of Arrhythmia Disease Using ECG Signals. In  Soft Computing for Problem Solving  (pp. 299-310). Springer, Singapore. 4. Sharma, P., Gupta, D. V., & Jangra , S. (2019). Ecg signal based arrhythmia detection system using optimized hybrid classifier.  Int. J. Innov . Technol. Explor . Eng.(IJITEE) .

4. Publications contd. 5. Sharma, P., & Gupta, D. V. (2018). Disease Classification from ECG Signal Using R-Peak Analysis with Artificial Intelligence.  International Journal of Signal Processing, Image Processing and Pattern Recognition ,  11 (3), 29-40. 6. Sharma, P., & Gupta, D. V. A Review and Comparative analysis of Heart using Electrocardiogram (ECG) International journal of Research in Electronics and Comuter Engineering ,Volume 5, Issue 3,July - September, 2017,291-299

5. A Linearly Adaptive Sine-Cosine Algorithm with Application in Deep Neural Network for feature optimization in Arrhythmia Classification using ECG Signals Pooja Sharma, Shail Kumar Dinkar

6. Abstract This work proposes a novel variant namely Linear Adaptive Sine Cosine Algorithm (LA-SCA) of newly developed metaheuristic called Sine Cosine Algorithm (SCA). The developed variant utilizes a linearly adaptive operator that is based on the number of generations, followed by an Opposition based Learning (OBL) model applied around the best solution. The applied mechanisms help LA-SCA to overcome the slow convergence problem in SCA and capable of achieving appropriate balance between macro and micro search ability of SCA. The proposed LA-SCA is utilized to offer an intelligent Deep Neural Network (DNN) mechanism for improving the feature extraction of ECG signal for efficient classification of arrhythmia diseases.

7. Introduction Time ahead detection and diagnosis of an abnormality proves to be a very significant achievement in the healthcare sector. Despite medical advances, the World Health Organization reported 17.9 million death cases due to cardiovascular diseases in 2016, which reflects compromised detection of cardiac diseases leading to unsatisfactory medical care. This stats reflects 31% of the global deaths, out of which 85% were due to stroke and heart attack [1 ]. It is a vital organ that works as a muscular pump that circulates blood throughout the body whose condition can be monitored using ECG (Electrocardiogram) [ 2 ]. The popularity of ECG as a medical device is also rising because of the ease it offers for testing while requiring the least amount of training. However, its interpretation requires a high level of expertise. ECG tracks the rhythm, and the rate of heartbeat tracing the blood flow through the heart [ 3 ]. It is generalized that healthy heartbeat 65 to 80 times per minute [ 4 ]. Arrhythmia is a generalized term used to denote any disturbances in the heart pulse. There are two fundamental types of arrhythmias, namely Bradycardia and Tachycardia [ 5-6] .

8. Introduction 60 65 70 75 80 85 90 95 100 Tachycardia > 100 60 < Healthy heart beat< 80 Brady arrhythmia < 60 Heart Beats Pattern Heart Beats per Minute Pattern of Heart Beat Pulse Rate Bradycardia arrhythmia is diagnosed when the heart rate is too low under 60 beats for consistently. Tachycardia is the point at which the heart rate is too quick more than 100 pulse rates every minute.

9. Cardiac physiology Heart chambers with blood circulation

10. Electrocardiogram Systematic representation of ECG waveform

11. Electrocardiogram Features Description Duration time It measures the time among two successive R peaks. 0.6 to 1.2 S The spreading of the electrical impulse from the right atrium to the left atrium generates the P wave. < 80 ms It is measured the and ends at the initial point of 120ms to 200ms It has a larger amplitude than the due to the larger muscle mass of the ventricle than atria. 82ms to 120 ms It represents the connecting region present between and . It is a region that reflects ventricles depolarization events. NA It reflects ventricles repolarization event and is peaked in abnormal conditions. 160ms It is hypothesized to reflect repolarization of the interventricular region. It is often absent or has very low amplitude. NA Features Description Duration time It measures the time among two successive R peaks. 0.6 to 1.2 S The spreading of the electrical impulse from the right atrium to the left atrium generates the P wave. < 80 ms 120ms to 200ms 82ms to 120 ms NA It reflects ventricles repolarization event and is peaked in abnormal conditions. 160ms It is hypothesized to reflect repolarization of the interventricular region. It is often absent or has very low amplitude. NA

12. Motivation and Contribution In clinical practice, morphological alterations in the ECG wave frequency are routinely analyzed to detect abnormalities and the type of arrhythmia . the process of analyzing ECG is not an easy task because, at times, a precise interpretation may require exploring every heart of the subject or the patient . ECG waveform follows a non-linear or sinusoidal dynamic behaviour that gets altered more in arrhythmia as compared to linear components . Major work in this paper can be concluded as : A unique DNN-based technique with support Vector Machine (SVM) is utilized to classify the different data samples of ECG into various types of disease related to arrhythmia for the prediction classifier. For function optimization, LA-SCA fitness function is used to identify the best feature population information that is forwarded to the next step. The proposed methodology utilizing LA-SCA with DNN+SVM is compared with original SCA and previous related work available in literature.

13. Literature Review Authors Description of the work Iqbal et al., 2018 [7] Authors had proposed Deep Deterministic Learning (DDL) to address the identification of cardiac diseases based on the ECG signal evaluation. The approach utilized the ANN framework for pattern recognition and classification. The researchers employed available datasets in addition to manually collected samples. The DDL based approach demonstrated an overall accuracy of 98% Chen et al., 2018 [8] Authors dedicated their work towards arrhythmia identification based on ECG signal evaluation followed by classification under four classes, namely, normal beats(N), ventricular ectopic beats (VEBs), supra-ventricular ectopic beats (SVEBs), and fusion of ventricular and normal (F). The work involved principal component analysis (PCA) with dynamic time warping (DTW) for extracting segmented features. 19 segmented arrhythmia features were used in RBF based SVM to detect heart-related diseases with an overall accuracy of 97.80% Yıldırım et al., 2018   [9] Authors had implemented a deep neural network to classify ECG signals into a set of 13, 15, and 17 arrhythmia classes. The work demonstrated an average accuracy of 91.33% while classifying signals into 17 classes with an average computation time of 0.015s required by each sample. However, the work demonstrated higher accuracy of 95.2% (13 classes) and 92.51% (15 classes) for a smaller number of classes is involved. Therefore work needs improvement as classification strength decreases for more classes

14. Literature Review Authors Description of the work Zairi et al., 2019  [10] Author had taken advantage of field-programmable gates array to minimize ECG features while involving DWT to reduce feature dimension for offer better arrhythmia classification using Multilayer Perceptron (MLP) with 98.3% accuracy Liu et al., 2019 [11] Author had proposed ECG signal classification with radial basis probabilistic process neural network (RBPPNN) among various heart diseases. The work demonstrated ECG among ten diseases with a classification accuracy of 75.52% and with 86.75% efficiency, specifically for sinus arrhythmia Wang et al., 2020  [12] Author had proposed a fully connected neural network (F-CNN) to achieved better performance for arrhythmia classification using MIT arrhythmia and MIT supraventricular arrhythmia database. The proposed classifier had a two-layered architecture for the classification of ECG signals into five classes. Each layer exhibited independent and fully connected neural networks. The experiments are conducted based on AAMI standards. The achieved overall classification accuracy is 93.4% by the F-CNN architecture.

15. Literature Review Authors Description of the work Çınar and Tuncer (2020) [13] Author presented the hybrid deep learning architecture on the basis of Alexnet-SVM. To analyze the performance, the classification of the signal first done by SVM, K-nearest Neighbor (KNN), and Long Short Time Memory (LSTM) produced an accuracy of 68.75%, 65.63%, and 90.67%, which has improved to 96.77% by hybrid approach Cai et al., 2020 [14] Author introduced a deep learning-based detection approach for the most common heart arrhythmia, namely; Atrial fibrillation (AF). To identify AF in ECG signal, presented schemes construct a novel one-dimensional deep densely connected neural network (DDNN). Simulation performed using a twelve-lead ECG recording dataset and exhibits accuracy of 99.35%. For classification of Normal Sinus Rhythm (NSR) Abnormal Arrhythmia (ARR) and Congestive Heart Failure (CHF) ECG signals

16. Proposed Methodology .

17. Preprocessing Stage Pre-processing of signal

18. Preprocessing stage ECG signal sample Preprocessed ECG signals using DWT

19. Feature Extraction stage Identification of signal for detection of  

20. Feature Optimization stage The features of ECG signals in the form of have been detected in the last stage. In the present section, the main task is to select the appropriate feature sets according to the type of disease . It has been established that the extracted features may include some irrelevant features that may be redundant or least significant for the classification process . In the present work, LA SCA is used to select the most relevant features . The fitness of each candidate solution ( ) is determined using an objective function ( ). Steps involved in optimizing the features representing are given in Algorithm3  

21. Training and Classification Deep Neural Network Model

22. Training and Classification Architecture of FFBPNN

23. Input Layer Hidden Layer Output Layer features of of Brady-arrhythmia 10 Hidden Layers Brady-arrhythmia features of of Premature or extra heartbeat Premature or extra heartbeat features of of Tachycardia Tachycardia features of of Ventricular arrhythmia Ventricular arrhythmia features of of Normal Normal Input Layer Hidden Layer Output Layer 10 Hidden Layers Brady-arrhythmia Premature or extra heartbeat Tachycardia Ventricular arrhythmia Normal Training and Classification Layer information of FFBPNN

24. Training and Classification

25. Training and Classification Illustration of the proposed methodology

26. Performance Evaluation The performance of the proposed work is computed in terms of precision, sensitivity, accuracy and execution time measurements. The metrics are calculated as follows using confusion matrix parameters. Where, True Positive, True Negative, False Positive and False Negative Execution time is measure to exhibit how quickly the applied algorithm can determine classification in terms of time.  

27. Result and Discussion Description of Utilized System Sr. No. Components Description 1 Software MATLAB 2016a 2 Toolboxes Data Acquisition Toolbox, Signal Processing Toolbox, Optimization Toolbox, Neural Network Toolbox 3 CPU Intel(R) Core ™ [email protected] GHz 4 Operating System Windows 10 (64 bit) 5 RAM 4GB 6 Hard Drive 240GB

28. Tested classifiers. Accuracy Training set Testing set Whole data Random Tree [97] 96.87 96.47 96.84 JRip [98] 97.24 96.14 97.32 SVM [99] 97.08 96.37 97.18 DNN 97.39 96.82 97.47 SVM+ DNN with LA-SCA 99.25 99.11 99.18 Result and Discussion (Accuracy Analysis )

29. Comparison with Existing W ork Sr. No. Research Work Neural Network Variants Accuracy (%) 1 Proposed SVM+DNN+LA-SCA 99.29 2 Sharma P, Shail Kumar SVM + FFBPNN 98.53% 3 Wang et al., 2020 F-CNN 93.4% 4 Zairi et al., 2019 MLP 98.3% 5 Singh et al., 2019 FFBPNN 95% 6 Bhagyalakshmi et al., 2018 SVNN 96.96% 7 Iqbal et al., 2018 ANN 98% 8 Isin and Ozdalili , 2017 Deep CNN 92% Accuracy comparison against existing neural network variants for arrhythmia classification

30. C onclusion The results show the superiority of proposed LA-SCA over original SCA and other methods. The proposed LA-SCA is utilized to optimize the feature extraction of ECG signal while using with an intelligent deep neural network(DNN) and support vector machine(SVM). This work presents an accurate ECG signal classification methodology to precisely classify ECG samples into sixteen classes of arrhythmia. DWT is used to denoise ECG samples at the beginning. The work is motivated by R peak’s unique ability to recognize QRS-complexes, which is then followed by SVM based signal optimization. This work authorizes that the proposed SVM + DNN with LA-SCA obtains 99.29% accuracy, 97.51% sensitivity and 98.66% specificity over whole data respectively which show better performance as compared to another classifier. Thus, according to a comparative analysis of the proposed work with existing approaches for the categorization of arrhythmia based on neural network architecture, the suggested work outperforms recent neural network enhancements developed to increase the performance of ECG classifications.

31. References World Health Organization. Cardiovascular Diseases (accessed 15.01.2021.). Available online: https://www.who.int/en/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds). Berkaya , S. K., Uysal , A. K., Gunal , E. S., Ergin , S., Gunal , S., & Gulmezoglu , M. B. (2018). A survey on ECG analysis. Biomedical Signal Processing and Control, 43, 216-235. Agarwal, S., Krishnamoorthy , V., & Pratiher , S. (2016, September). ECG signal analysis using wavelet coherence and s-transform for classification of cardiovascular diseases. In 2016 International conference on advances in computing, communications and informatics (ICACCI) (pp. 2765-2770). IEEE. Elhaj , F. A., Salim , N., Harris, A. R., Swee , T. T., & Ahmed, T. (2016). Arrhythmia recognition and classification using combined linear and nonlinear features of ECG signals. Computer methods and programs in biomedicine, 127, 52-63. Pławiak , P. (2018). Novel methodology of cardiac health recognition based on ECG signals and evolutionary-neural system. Expert Systems with Applications, 92, 334-349. Raj, S., & Ray, K. C. (2018). Sparse representation of ECG signals for automated recognition of cardiac arrhythmias. Expert systems with applications, 105, 49-64 . Iqbal, U., Wah , T. Y., ur Rehman , M. H., Mujtaba , G., Imran, M., & Shoaib , M. (2018). Deep deterministic learning for pattern recognition of different cardiac diseases through the internet of medical things. Journal of medical systems, 42(12), 252. Chen, X., Wang, Y., & Wang, L. (2018). Arrhythmia recognition and classification using ECG morphology and segment feature analysis. IEEE/ACM transactions on computational biology and bioinformatics, 16(1), 131-138.

32. References Yıldırım , Ö., Pławiak , P., Tan, R. S., & Acharya, U. R. (2018). Arrhythmia detection using deep convolutional neural network with long duration ECG signals. Computers in biology and medicine, 102, 411-420 . Zairi , H., Talha , M. K., Meddah , K., & Slimane , S. O. (2019). FPGA-based system for artificial neural network arrhythmia classification. Neural Computing and Applications, 1-16. Liu, K., Xu , S., & Feng, N. (2019). A radial basis probabilistic process neural network model and corresponding classification algorithm. Applied Intelligence, 49(6), 2256-2265. Wang, H., Shi, H., Lin, K., Qin, C., Zhao, L., Huang, Y., & Liu, C. (2020). A high-precision arrhythmia classification method based on dual fully connected neural network. Biomedical Signal Processing and Control, 58, 101874 . Çınar , A., & Tuncer , S. A. (2020). Classification of normal sinus rhythm, abnormal arrhythmia and congestive heart failure ECG signals using LSTM and hybrid CNN-SVM deep neural networks.  Computer Methods in Biomechanics and Biomedical Engineering , 1-12 Cai , W., Chen, Y., Guo , J., Han, B., Shi, Y., Ji , L., ... & Luo , J. (2020). Accurate detection of atrial fibrillation from 12-lead ECG using deep neural network.  Computers in biology and medicine ,  116 , 103378.

33. Kindly allow for thesis submission within one month

34. Thank you