Table 4.
Comparison between proposed model and relevant studies.
| Study | Splitting Strategy | Method | Performance | |||
|---|---|---|---|---|---|---|
| Sen | Spe | Score | Acc | |||
| Serbes et al. [15] | official 60/40 | SVM | - | - | - | 49.86% |
| Jakovljević and Lončar-Turukalo [16] | official 60/40 | HMM | - | - | 39.56% | - |
| Kochetov et al. [18] | interpatient 5-fold CV |
RNN | 58.4% | 73% | 65.7% | - |
| Chambres et al. [19] | official 60/40 | HMM | 42.32% | 56.69% | 39.37% | 49.50% |
| SVM | 48.90% | 77.80% | 49.86% | 49.98% | ||
| Liu et al. [20] | random 75/25 | CNN | - | - | - | 81.62% |
| Minami et al. [21] | official 60/40 | CNN | 28% | 81% | 54% | - |
| Ma et al. [22] | official 60/40 | bi-ResNet | 31.12% | 69.20% | 50.16% | 52.79% |
| random 10-fold CV |
58.54% | 80.06% | 69.30% | 67.44% | ||
| Acharya and Basu [24] | interpatient 80/20 | CNN-RNN | 48.63% | 84.14% | 66.38% | - |
| Demir et al. [25] | 10-fold CV | VGG16 | - | - | - | 63.09% |
| SVM | - | - | - | 65.5% | ||
| Saraiva et al. [26] | random 70/30 | CNN | - | - | - | 74.3% |
| Gairola et al. [27] | official 60/40 | CNN | 40.1% | 72.3% | 56.2% | - |
| interpatient 80/20 | 53.7% | 83.3% | 68.5% | - | ||
| Ma et al. [28] | official 60/40 | ResNet + NL | 41.32% | 63.20% | 52.26% | - |
| interpatient 5-fold CV | 63.69% | 64.73% | 64.21% | - | ||
| Yang et al. [29] | official 60/40 | ResNet + SE + SA | 17.84% | 81.25% | 49.55% | - |
| Pham et al. [30] | official 60/40 | CNN-MoE | 26% | 68% | 47% | - |
| random 5-fold CV |
68% | 90% | 79% | - | ||
| Nguyen and Pernkopf [31] | official 60/40 | ResNet | 37.24% | 79.34% | 58.29% | - |
| This work | official 60/40 | CNN-LSTM with FL | 47.37% | 82.46% | 64.92% | 73.69% |
| Interpatient 10-fold CV | 52.78% | 84.26% | 68.52% | 76.39% | ||
| LOOCV | 60.29% | - | - | 74.57% | ||