Enhanced Breast Cancer Diagnosis Using Multimodal Feature Fusion with Radiomics and Transfer Learning

. 2025 Aug 28;15(17):2170. doi: 10.3390/diagnostics15172170

Algorithm 2 Image augmentation pipeline

Require:
Image $I_{aligned}$
Ensure:
Augmented Image $\hat{I}$
1:
$I \leftarrow tf . image . random_flip_left_right (I)$
2:
$I \leftarrow tf . image . random_flip_up_down (I)$
3:
$I \leftarrow tf . image . random_brightness (I, 0.3)$
4:
$I \leftarrow tf . image . random_contrast (I, 0.8, 1.2)$
5:
$I \leftarrow tf . image . random_saturation (I, 0.8, 1.2)$
6:
$I \leftarrow tf . image . random_hue (I, 0.02)$
7:
$σ \leftarrow tf . random . uniform ([], 0.1, 2.0)$
8:
$I \leftarrow tf . image . gaussian_filter 2 d (I, (3, 3), σ)$
9:
$s \leftarrow tf . shape (I) [: 2] \times tf . random . uniform ([], 0.8, 1.0)$
10:
$I \leftarrow tf . image . random_crop (I, s)$
11:
$I \leftarrow tf . image . resize (I, tf . shape (I) [: 2])$
12:
$I^{'} \leftarrow tf . random . normal (tf . shape (I), 0.0, 0.05)$
13:
$I \leftarrow tf . clip_by_value (I + I^{'}, 0.0, 1.0)$
14:
return Augmented Image $\hat{I}$