Electrocardiogram (ECG) is a diagnostic method that helps to assess and record the electrical impulses of heart. The traditional methods in the extraction of ECG features is inneffective for avoiding the computational abstractions in the ECG signal. The cardiologist and medical specialist find numerous difficulties in the process of traditional approaches. The specified restrictions are eliminated in the proposed classifier. The fundamental aim of this work is to find the R-R interval. To analyze the blockage, different approaches are implemented, which make the computation as facile with high accuracy. The information are recovered from the MIT-BIH dataset. The retrieved data contain normal and pathological ECG signals. To obtain a noiseless signal, Gabor filter is employed and to compute the amplitude of the signal, DCT-DOST (Discrete cosine based Discrete orthogonal stock well transform) is implemented. The amplitude is computed to detect the cardiac abnormality. The R peak of the underlying ECG signal is noted and the segment length of the ECG cycle is identified. The Genetic algorithm (GA) retrieves the primary highlights and the classifier integrates the data with the chosen attributes to optimize the identification. In addition, the GA helps in performing hereditary calculations to reduce the problem of multi-target enhancement. Finally, the RBFNN (Radial basis function neural network) is applied, which diminishes the local minima present in the signal. It shows enhancement in characterizing the ordinary and anomalous ECG signals.

Automatic electrocardiogram analysis is the best practice for recording the functions of the heart by positioning the electrodes at the external area of the skin. The research on ECG device is focused by various researchers in recent years [

In [

The proposed system focuses on the blockage area to detect the R-R interval in ECG signal as represented in

Gabor filter is a type of linear filters and its response for impulse signal is characterized as a Gaussian function [

To define the result of signal propagation in frequency domain, the unpredictable theory has to equal the constant value.

In 2D type, the time variable t is supplanted by spatial coordinates (x, y), and the frequency f is superseded by space variables (u, v). In most cases, the 2D Gabor function is evaluated as follows:

In the frequency domain,

The standard deviation of the elliptical Gaussian is represented as

_{u} is computed by using the equation,

_{v} is evaluated by using,

This method uses the DCT-DOST scheme to examine the time domain representation of the ECG signal and to naturally distinguish the R-peak. In the case of DOST, the signal loses its structure during the coefficient truncation. However, it withstands against the coefficient truncations with DCT. The DCT includes all the frequencies to reduce the unpredictability. The DCT-DOST shows essential coefficients at lower frequencies.

The linear S transform fill the gap among fourier and wavelet transforms. The S transfer of a signal h(t) is,

Window’s width is expressed as,

_{0}) is a 1D time function that demonstrates the magnitude change with time for a fixed frequency. The DOST of h (KT) is,

where n extends from 1, 2,…N-1.

The proposed work’s main goal is to automatically find the peak value of R. To detect the R peak, every heartbeat segment consists 105 patterns as per the R top identification and 151 patterns are generated after the retrieval of R-peak. A sum of 256 patterns is taken to find the extension of cardiac pulse. The advantage of determining the length of every cardiac pulse is to accurately detect the R top. The entire process is depicted in

After the retrieval of noiseless image, the DCT-DOST approach is applied for performing peak identification. Initially, the sample frequency is 100 hertz. It is split into five intervals to accurately locate the R-R interval. It is real value transformation and it is positioned in space to minimize the time. It includes no negative frequency. Only positive frequencies are used and there is no symmetry coefficient. Hence, the higher frequencies have to be converted as frequency space during segmentation. Since the DCT-DOST contains no negative frequencies, the frequency width for any signal of length 2 N is,

N_{1 = 1} and

N_{i }=_{ }2^{i−2} for 2 ≤

The DCT-DOST method is,

The info ECG signal is propagated via N point DCT. This level produces the coefficients A_{1}, A_{2}, …, A_{n}. The acquired coefficients are split into sub bands [2^{0}, 2^{1}, 2^{2}, ……2^{n−1}. For each sub band,

In ECG signal, the feature extraction helps to figure out the amplitude and interval values of P-QRS-T segment in ECG. This work aims to determine the R-R interval and to extract the morphological highlights. By utilizing highlight extraction, 19 transient highlights including PQ, RR and PT interim and 3 morphological highlights are extricated from the ECG signal as portrayed in

The maximal and minimal points for each beat of the ECG signal are captured by using morphological highlights. The equation is,

The least value and most value point are figured out in the first and next R peaks. Then it is normalized by taking the esteems between 0 and 1.

Features, which describe the position of P, Q, R, S, T peak and QRS duration are computed by using the initial position of the Q-wave in the end of the S-wave. The QRS complex is computed, which is highly significant in the detection of abnormality.

Step 2: Identify the duration of QRS complex waveform.

Step 3: Execute the wavelet analysis

Step 4: Calculate the coefficients by using wavelet decomposition.

Step 5: Identify R peak location in the signal by taking 60% of its value as threshold.

Step 6: Identify Q point by finding the smallest value ranging from Rloc-50 to Rloc-10.

Step 7: Identify S point by finding the smallest value ranging from Rloc+5 to Rloc+50.

Step 8: Identify T point by finding the highest value ranging from Rloc+25 to Rloc+100.

Step 9: Compute the duration of QRS complex by using the equation,

Step 10: Find X=QRS.

False negative Detection of QRS complex by using,

Premature ventricular complexes

Low amplitude.

False positive Detection by using,

Negative QRS complexes

Low SNR

This QRS algorithm is helpful to extract the R-R interval. It is performed by using the heart rate variability (HRV). It is an interval among two sequential R peaks and it is measured by,^{th} wave.

The next step is to reduce the number of features. It’s done with the aid of a genetic algorithm. It is utilized to improve the features for identifying ECG signals. The structure of this algorithm is signified in

And finally it applies a fitness function, which is computed by,

N stands for the number of outputs, t stands for the goal output, and out stands for the actual output. Positive and negative values may be present in the fitness function. As a result, we can’t use fitness benefit directly. The selection operator is used to identify the best features associated with the highest fitness value and passes them over to the next generation. The crossover operator swaps the selected individuals chromosomes to produce offspring chromosomes.

The final operator is then used to notify the bits in the chromosome. The probability that the chromosome in the n^{th} position will be estimated is calculated using,

The GA algorithm aids in the optimization of neural network results, and it works well to achieve high precision, sensitivity, and specificity, as well as providing output with better classification. The classification is performed by RBFNN.

^{n} vector of real numbers. The network’s result is R^{n} →

where the neurons present in the hidden layer is represented as N, C_{i} is the centre vector and a_{i} is the neuron’s weight. The parameters a_{i}, c_{i} and β_{i} aid to optimize the fitness between φ and the signal.

A typical RBF of the scalar input vector which is a first layer is,

Normalized and de-normalized forms of the generated input are also possible. It is discovered to be in non-normalized state. The equation is,

where,

This input layer expression is expressed as,

where,

In the de-normalized form

In the normalized form

The probability density function among the input and the output layer is estimated,

The output y given an input x as

where, the conditional prospect of y specified x is signified as P (y|x).

For performing classification, the trained and test datasets are obtained from MIT-BIH database. Nearly 80% of data are chosen for training and 20% is considered for testing. The training dataset is represented as

The output of the training dataset is Y_{i} and time prediction is done by predicting the successive value and features of a sequence,

The entire work is implemented in MATLAB to analyze the ECG signals. The MIT-BIH dataset is used to validate. The RBFNN classifier is trained by using the aforementioned dataset and the performance is examined for the sample ECG signal. The expected outcome for the ECG signals at each stages of the proposed method is exhibited for detailed analysis. The ECG specimen image taken is elaborated for 50,000 samples. A sample ECG signal is shown in

The electromyogram noise, Gaussian noise and low frequency noises are excluded by the Gabor filter. In addition, the texture features of ECG signal are analysed. In comparison with the input signal, the output of Gabor is more precise and accurate as depicted in

The distance between the R-peak values is estimated by finding the absolute values. When the heart’s electrical function is assumed as a vector, it is easy to analyze the trajectory of the vectors peak. The signal ECG is considered as projection of the heart’s electrical vector as depicted in

The energies in the ECG signal is gathered by using DCT-DOST to represent the most important coefficient at low frequency. The features that are extracted using the DCT-DOST approach indicate the time-recurrence attributes of ECG signal. From

The traditional filtering minimizes the signal noise by delaying the QRS components. The zero phase filtering minimizes phase distortion and provides a compromise among filtering and data retention. The output of the zero phase filter is depicted in

The ECG portion is composed of 112 patterns before the occurrence of R top and 144 patterns after the occurrence of R top. An aggregate of 256 patterns is chosen to find the length of every occasion relating to window size. To consolidate the majority of data with respect to each heart occasion, the length of each event is chosen. These unbalanced time–recurrence coefficients have to be processed for the ECG signal to represent the morphological qualities. The segmentation result of DCT-DOST is shown in

The moving average filter is utilized to remove high frequency noises from the ECG signal by computing the running mean on the predetermined window length. The R-top in the ECG signal is smoothed around 33% of its unique height. The output of this filter is represented in

The QRS wave of the ECG is detected by using zero crossing point detection approach. The dominant and low frequency contents in the ECG are roughly estimated as represented in

The R top discovery in ECG is used to analyze heart anomalies and pulse fluctuation. The primary request separation of the sign is utilized to store the incline data of the genuine pinnacles.

The enhanced performances is achieved with the slope index than the high recurrence index, which is depicted in

The QRS detection ensures the efficient extraction of beat interval and the abnormalities in the heart function. The improvement in the QRS sections are executed by the proposed technique to eliminate the pattern meandering. In this paper, the QRS fiducial focuses are detected to perceive the R point using by QRS complex.

The RR-interim is resolved to obtain the dynamic qualities of the ECG signal. The mean RR interim features are determined by averaging the RR interims of the previous 3-minimum RR interval in a specific occasion, which is highlighted in

Similarly, the neighborhood RR features are inferred by averaging all the RR-interims of the previous episodes of a specific occasion. The neighborhood and mean highlights indicate the mean qualities. These 4 highlights are connected to the morphological list of ECG signal.

The performance of this methodology is analogized with the traditional methods like CNN(Convolutional Neural Network) and SVM (Support Vector Machine). With a maximum accuracy of 98.5%, the accuracy of this system outperforms other approaches, which is portrayed in

The sensitivity shows the true positive value of the classification. It’s calculated as the percentage of positives, which are correctly categorised. With a maximum sensitivity of 98.3%, it outperforms the CNN and SVM, which own the maximum sensitivity of 92% and 86% respectively.

The proposed method’s specificity values change in a zig-zag pattern as the number of samples is increased. With a maximum specificity of 99%, the proposed method delivers better performance than CNN and SVM, which have the maximum of 93% and 95.6% respectively. The compariron outcome is represented in

The measure of various contents in the ECG signal like class, sinus rhythm, artifact, ventricular tachycardia, atrial brillation, bigeminy and PVC (Premature Ventricular Contractions) are computed in terms of R, P, S and F1. From

Aggregate accuracy comparison | |||
---|---|---|---|

Model | Training | Validation | Test |

Baseline–LSTM |
66.8% |
66.3% |
65.6% |

Stacked unidirectional-LSTM |
80.5% |
78.1% |
79.2% |

Deep residual-CNN | 84.7% | 75.3% | 74.7% |

Combined unidirectional LSTM-CNN |
83.4% |
77.7% |
79.6% |

Proposed RBFNN | 99% | 84.4% | 98.5% |

Classification metrics comparison | ||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|

Rhythm | BDLSTM | Residual | LSTM-CNN | Proposed-RBFNN | ||||||||||||

Class | R | P | S | F1 | R | P | S | F1 | R | P | S | F1 | R | P | S | F1 |

Sinus rhythm | 0.82 | 0.83 | 0.94 | 0.84 | 0.64 | 0.88 | 0.86 | 0.76 | 0.79 | 0.80 | 0.95 | 0.79 | 0.85 | 0.87 | 0.96 | 0.89 |

Artifact/noice | 0.88 | 0.82 | 0.94 | 0.83 | 0.89 | 0.97 | 0.94 | 0.82 | 0.81 | 0.83 | 0.94 | 0.81 | 0.89 | 0.85 | 0.92 | 0.84 |

Ventricular tachycardia | 0.16 | 0.51 | 0.95 | 0.26 | 0.48 | 0.92 | 0.96 | 0.08 | 0.56 | 0.57 | 0.97 | 0.43 | 0.55 | 0.34 | 0.94 | 0.67 |

Atrial brillation | 0.81 | 0.83 | 0.94 | 0.82 | 0.78 | 0.93 | 0.92 | 0.76 | 0.73 | 0.69 | 0.89 | 0.84 | 0.88 | 0.81 | 0.97 | 0.81 |

Bigeminy | 0.72 | 0.65 | 0.82 | 0.67 | 0.89 | 0.98 | 0.98 | 0.16 | 0.67 | 0.67 | 0.96 | 0.55 | 0.84 | 0.83 | 0.91 | 0.80 |

Pvc | 0.78 | 0.76 | 0.88 | 0.76 | 0.78 | 0.93 | 0.93 | 0.83 | 0.79 | 0.77 | 0.92 | 0.72 | 0.81 | 0.82 | 0.95 | 0.89 |

The training, validation and testing efficiencies of the proposed approach are compared with the conventional methods. The training efficiency of this present method is higher than the other methods.

From

F1 score class comparison | ||||
---|---|---|---|---|

Rhythm class | BDLSTM | RESIDUAL | LSTM-CNN | Proposed-RBFNN |

Sinus rhythm | 0.812 | 0.734 | 0.793 | 0.883 |

Artifact/noise | 0.834 | 0.818 | 0.843 | 0.923 |

Ventricular tachycardia | 0.265 | 0.169 | 0.417 | 0.721 |

Atrial Brillation | 0.837 | 0.763 | 0.764 | 0.852 |

Bigeminy | 0.663 | 0.136 | 0.553 | 0.754 |

Pvc | 0.769 | 0.821 | 0.724 | 0.912 |

Overall | 0.813 | 0.728 | 0.742 | 0.902 |

By considering the classification methods, the performance is improved as shown in

F1 score class comparison | ||||||||
---|---|---|---|---|---|---|---|---|

Rhythm class | BDLSTM | Residual | LSTM-CNN | Proposed-RBFNN | ||||

Sinus rhythm | 0.812 | 0.612 | 0.734 | 0.692 | 0.793 | 0.702 | 0.883 | 0.813 |

Artifact/noise | 0.834 | 0.734 | 0.818 | 0.746 | 0.843 | 0.774 | 0.923 | 0.874 |

Ventricular tachycardia | 0.265 | 0.065 | 0.169 | 0.085 | 0.417 | 0.145 | 0.721 | 0.835 |

Atrial Brillation | 0.837 | 0.337 | 0.763 | 0.797 | 0.764 | 0.717 | 0.852 | 0.857 |

Bigeminy | 0.663 | 0.263 | 0.136 | 0.073 | 0.553 | 0.523 | 0.754 | 0.873 |

Pvc | 0.769 | 0.669 | 0.821 | 0.709 | 0.724 | 0.709 | 0.912 | 0.879 |

The proposed work enhances the diagnosis accuracy by eliminating the redundant and noise highlights. The specified algorithm provides sensitivity and accuracy above 98.5%. These algorithms are computationally facile and aids in the processing of massive set of database. By this work, the artifacts are detected with extreme accuracy. It gives better acknowledgement performance than the other existing frameworks.