Table 1.
Summary of identified publications for this review.
| Objective | Publication date | Authors | Input data (CMR sequences) | Dataset | Methodological considerations | Sample | Evaluation |
|---|---|---|---|---|---|---|---|
| To distinguish between myocarditis and MI | 19.12.19 | Di Noto et al. (43) | Cines, SSFP, T2, LGE | University Heart Center, University Hospital Zurich | Exponential filter, wavelet transform filter, recursive feature elimination | 173 patients (111 MI, 62 Myocarditis) | Training (90%); Testing (10%). Nested 10-fold cross validation |
| To predict the presence of LGE in myocarditis patients | 16.09.22 | Cavallo et al. (44) | Short Axis (STIR, PSIR), T1, T2, LGE | Policlinico Tor Vergata, Rome, Italy | Weka data mining platform Correlation-based Feature Selection Ensemble ML |
19 patients | Training (70%); Testing (30%) 10-fold cross validation |
| To distinguish myocarditis from healthy controls | 04.01.22 | Sharifrazi et al. (45) | Cine, T1, T2, LGE (PSIR) | Z-Alizadeh Sani myocarditis dataset | Convolutional neural network combined with k means clustering | 32 myocarditis patients, 15 healthy controls | Training (90%); Testing (10%) |
| To distinguish myocarditis from healthy controls | 30.06.22 | Moravvej et al. (46) | Cine, T1, T2, LGE (PSIR) | Z-Alizadeh Sani myocarditis dataset | Reinforcement learning with population-based weights strategy | 586 myocarditis patients, 307 healthy controls | 5-fold cross validation |
| To distinguish myocarditis from healthy controls | 26.10.22 | Jafari et al. (47) | Cine, T1, T2, LGE (PSIR) | Z-Alizadeh Sani myocarditis dataset | Turbulence Neural Transformer (TNT) architecture Explainable-based Grad Cam method |
32 myocarditis patients, 15 healthy controls | 10-fold cross validation |
| To distinguish myocarditis from healthy controls | 08.04.22 | Ghareeb et al. (48) | Cines, T1, T2, LGE(PSIR) | Heart Hospital, Hamad Medical Corporation, Qatar | Unsupervised learning- K means clustering, Bayesian factor analysis | 169 patients | Silhouette score |
NB The reported sample sizes differ between the 3 studies using the Z-Alizadeh Sani myocarditis dataset– we are unable to ascertain the reasons for this and it is unclear in the manuscripts.