Table 3.
Comparison between this work and state-of-the-art MI detection methods.
| Method | Accuracy % | Sensitivity % | Specificity % | Database/ Remark |
Subjects/ Records Number |
Approach |
|---|---|---|---|---|---|---|
| Murugan [53] | NA | 92.30 | 94.30 | European ST-T/Beat | 90 records | Ant-Miner algorithm |
| Jinhopark [54] | NA | 95.70 | 95.30 | European ST-T/Beat | 90 records | Kernel density estimation (DWT based) with SVM |
| Safdarian [14] | 94.00 | NA | NA | PTB/Beat | 290 subjects | T wave integral with KNN, PNN and ANN |
| B. Liu [16] | 94.40 | NA | NA | PTB/Beat | 52 Normal 148 MI |
ECG polynomial fitting algorithm PolyFit-based ECG |
| Jian Wang [48] | 89.00 | 91.70 | 81.50 | Hospital data collection | 167 patients | Deep learning-based scheme |
| L. Sun [13] | NA | 91.00 | 85.00 | PTB/Beat | 52 Normal 238 MI |
ST segment, Polynomial fitting with KNN |
| Acharya U.R. [3] | 98.50 | 99.70 | 98.50 | PTB/Beat | 52 Normal 148 CAD |
DCT features based |
| Murthy [17] | 90.51 | 96.19 | NA | European ST-T/Beat | 16 MI | Statistical analysis with PCA and SVM |
| M. Arif [9] | 98.30 | 97.00 | 99.60 | PTB/Beat | 52 Normal 148 MI |
KNN, Time domain feature extraction |
| J.H. Tan [20] | 99.85 | 99.84 | 99.85 | Fantasia, PTB Single lead |
52 Normal 238 MI |
8-layers stacked CNN-LSTM with Blindfold |
| P. Barmpoutis [6] | 99.70 | - | - | PTB/Beat | 290 subjects | mapping of Grassmannian and Euclidean features into a Hilbert space |
| V.K. Sudarshan [22] | 99.86 | 99.78 | 99.94 | MIT-BIH Normal, Fantasia, and BIDMC/2-s Frame | 73 subjects | Dual tree complex WT coefficients features with KNN |
| W.S. Kim [49] | NA | 84.60 | 91.50 | Collected data/HRV | 20 Normal 64 Patients |
HRV time and frequency measurements |
| E.S. Jayachan-dran [50] | 95.00 | NA | NA | MIT-BIH/Beat | 6 Normal 2 MI |
Time domain analysis |
| S.G. Al-Kindi [52] | 93.70 | 85.00 | 100.00 | PTB/ST-segments | 20 Normal 20 MI |
ST segment analysis by DWT |
| L.N. Sharma [18] | 96.00 | 93.00 | 99.00 | PTB/Frame | 52 Normal 238 MI |
ST segment analysed by DWT, KNN and SVM |
| Kamal Jafarian [55] | 98.43 | 98.50 | 98.37 | PTB/ST-segments | 52 Normal 148 MI |
CNN scheme used with DWT and PCA based features |
| This work | 99.09 | 99.49 | 98.44 | European ST-T, Fantasia, and Collected data/minute | 92 Normal 266 MI |
PR and ST feature extraction by using Choi-Williams and classified by Multi-Class SVM |