Table 1 Table of the state-of-the-art performances achieved in previous works about NSCLC recurrence prediction.
N. of patients | Dataset | Model | Performances | |
|---|---|---|---|---|
Wang et al.51 | 157 | Private | Handcrafted Radiomic features based | Acc = 0.85 |
Aonpong et al.50 | 88 | Public | CNN + gene-expression based | AUC = 0.77 Acc = 0.83 |
Kim et al.49 | 326 | Public | CNN based + Handcrafted Radiomic based + Clinical based | AUC = 0.77 Acc = 0.73 |
Hindocha et al.52 | 657 | Private | Clinical based | AUC = 0.69 |
Bove et al.14 | 144 | Public | CNN based + Clinical based | AUC = 0.83 Acc = 0.79 |
Our proposed model | Public | CNN + Transformer based | AUC = 0.91 Acc = 0.89 | |
Our proposed model | 144 | ViT + Transformer based | AUC = 0.90 Acc = 0.86 |
