Page 64 - AIH-1-4

Artificial Intelligence in Health A fuzzy system for heartbeat classification

Table 3. Performance evaluation results of testing data using the VTMA classification method
ANFIS TP TN FP FN Accuracy (%) Sensitivity (%) Specificity (%) Precision (%) F1-score (%)
NSR 35 175 0 0 100 100 100 100 100
LBBB 34 172 1 3 98.09 91.89 99.42 97.14 94.44
RBBB 35 175 0 0 100 100 100 100 100
PVC 33 172 2 3 97.62 91.67 98.85 94.29 92.96
APC 35 175 0 0 100 100 100 100 100
PB 55 275 0 0 100 100 100 100 100
Abbreviations: ANFIS: Adaptive neuro-fuzzy inference system; APC: Atrial premature condition; FN: False negative; FP: False positive; LBBB: Left
bundle branch block; NSR: Normal sinus rhythm; PVC: Premature ventricular contraction; TN: True negative; TP: True positive.

Table 4. Performance evaluation results of checking data using the VTMA classification method
ANFIS TP TN FP FN Accuracy (%) Sensitivity (%) Specificity (%) Precision(%) F -score (%)
1
NSR 10 50 0 0 100 100 100 100 100
LBBB 9 48 1 2 95 81.82 97.96 90 85.71
RBBB 10 50 0 0 100 100 100 100 100
PVC 10 47 0 3 95 76.92 100 100 86.96
APC 10 50 0 0 100 100 100 100 100
PB 10 50 0 0 100 100 100 100 100
Abbreviations: ANFIS: Adaptive neuro-fuzzy inference system; APC: Atrial premature condition; FN: False negative; FP: False positive; LBBB: Left
bundle branch block; NSR: Normal sinus rhythm; PB: Paced beat; PVC: Premature ventricular contraction; TN: True negative; TP: True positive.

Table 5. Comparison of the proposed methods and related works
Performance evaluation Method Accuracy (%) Sensitivity (%) Specificity (%) Precision (%) F -score (%) Number of classes
1
[1] TERMA- FrFT 82.2 84.25 - 89.03 80.23 6
[2] DCNN 98.63 92.41 99.06 92.86 92.63 5
[4] CNN 89.3 - - - 89.10 1
[20] ANFIS 97.75 97.75 99.25 - - 4
[21] ANFIS 98.39 - 99.67 - - 6
[25] CNN-RNN 95.90 95.90 96.34 - - 5
[26] 2-D CNN 97.3 89.3 98 - - 2
This work VTMA 98.33 93.12 99.66 98.33 95.44 6
Abbreviations: ANFIS: Adaptive neuro-fuzzy inference system; CNN: Convolutional neural network; DCNN: Deep convolutional neural network;
RNN: Recurrent neural network; TERMA- FrFT: Two event-related moving averages-fractional Fourier transform; VTMA: Variable-threshold
multi-adaptive neuro-fuzzy system.

also covering six classes, the same as our proposed method. through subtractive clustering. Each member of the output
Overall, the proposed work outperforms the mentioned vector is assigned a specific heart rate, with the output
studies, as presented in Table 5. pulses divided into six categories. An input data matrix
of properties is generated and joined to an output vector,
5. Conclusion resulting in a system that classifies six types of heartbeats
This research presents a VTMA for detecting ECG using fuzzy logic and neural learning.
abnormalities. The process begins with selecting ECG Acknowledgments
records for classification. Next, the ECG signals undergo
normalization using a subtractive clustering algorithm. None.
To pre-process the data, low-pass and high-pass filters are
applied to eliminate noise. Following this, input properties Funding
are extracted and prepared for the neural-fuzzy system None.

Volume 1 Issue 4 (2024) 58 doi: 10.36922/aih.3367