Skip to main content
. 2024 Sep 14;15:447. doi: 10.1007/s12672-024-01333-1

Table 2.

The summary of the relevant studies

Classification Patient number Class ratioa Sequences Classifier AUCb resampling References
EGFR mutation/wild type 29/32 0.91 CET1WI RF 0.89 Ahn, 2020 [20]
EGFR mutation/wild type 28/24 1.17 CET1WI, T2WI, T2FLAIR, DWI LASSO 0.99 Wang, 2021 [21]
EGFR mutation/wild type 42/57 0.74 CET1WI, ADC, FA RF 0.73 Park, 2021 [22]
EGFR mutation/wild type 92/96 0.96 CET1WI, T2WI LASSO 0.90 Cao, 2022 [23]
EGFR mutation/wild type 70/92 0.76 CET1WI, T2WI, T2FLAIR LR 0.85 Zhen, 2022 [24]
EGFR/ALK mutations 96/90 1.06 CET1WI, T2WI, T2FLAIR RF 0.99 Li, 2022 [25]
EGFR/ALK/KRAS mutations 75/21/15 0.28/0.2 CET1WI, T2FLAIR RF 0.91 SMOTE Chen, 2020 [28]
T790M positive/negative 95/138 0.69 CET1WI, T2WI, DWI RF 0.89 SMOTE Li, 2023 [26]
T790M positive/negative 45/65 0.69 CET1WI, T2WI LASSO 0.81 Fan, 2023 [27]

EGFR: epidermal growth factor receptor; ALK: anaplastic lymphoma kinase; KRAS: Kirsten rat sarcoma virus; AUC: the area under the receiver operating characteristic curve; T2WI: T2-weighted image; DWI: diffusion-weighted imaging; CET1WI: gadolinium-enhanced T1-weighted image; T2FLAIR: T2 fluid-attenuated inversion recovery; ADC: apparent diffusion coefficient; FA: fractional anisotropy; LR: logistic regression; RF: random forest; LASSO: least absolute shrinkage and selection operator; SMOTE: synthetic minority oversampling technique

aThe class ratio is relative to the first class

bWe only list AUC from the training set of best strategy for a fair comparison among the studies