Table 3 Summary of potential findings based on model performance across different input formats.
From: Focal cortical dysplasia (type II) detection with multi-modal MRI and a deep-learning framework
Aspects | Classification tasks | MRI | DL models | Remarks | |
|---|---|---|---|---|---|
Modes (Iacs) | Planes | ||||
FCD-II | FCD-II vs. HCs | T1w | ax(T1w) | Xception/ DenseNet201 | • Xception and DenseNet201 show the best performance. |
Subtypes | FCD-IIa vs. FCD-IIb | FLAIR/T1w | Sg (FLAIR/T1w) | DenseNet201 | • DenseNet201 displays remarkable performance. |
Age | Adolescents vs. HCs | FLAIR/T1w | ax(FLAIR)/ cr(T1w) | DenseNet201 | • The DenseNet201 depicts remarkable outcomes. • The axial plane of both MRI modes performs better in FCD-II lesion detection. |
Adults vs. HCs | T1w | ax(T1w) | |||
IIa vs. IIb (in adults) | FLAIR/T1w | - | |||
Sex | Female patients vs. HCs | T1w | cr (FLAIR/T1w) | Xception/ DenseNet201 | • Coronal and sagittal planes showed more distinct and diverse features. • DenseNet201 recorded the highest performance scores. |
Males vs. HCs | T1w | sg (T1w) | DenseNet201 | ||
Females vs. Males | T1w | cr/ sg (T1w) | |||
IIa vs. IIb (in females) | FLAIR/T1w | - | |||
IIa vs. IIb (in males) | T1w | - | |||
Brain hemisphere | FCD-II in left vs. right | FLAIR/T1w | sg (FLAIR)/ ax (T1w) | DenseNet201 | • FCD-II lesions in axial and sagittal planes were well distinguished. • The DenseNet201 showed promising results. |
IIb in left vs. right | FLAIR/T1w | ax (T1w) | |||
Brain lobes | frontal vs. parietal | T1w | ax (T1w) | DenseNet201 | • Well detected in T1w MRI mode and the axial imaging plane. • DenseNet201 produced noteworthy outcomes. |
IIa vs. IIb (in frontal) | T1w | sg (FLAIR)/ax (T1w) | |||
