Algorithm 1. Pseudocode for brain tumor segmentation based on multimodal MRI. ReU-Net-Multimodal Brain Tumor Segmentation

Input: Multimodal 3D MRI scans (M_t1, M_t1ce, M_t2, M_flair) Output: Segmentation maps (S_wt, S_tc, S_et) Procedure ReU_Net_Segmentation(M_t1, M_t1ce, M_t2, M_flair): 1. Multimodal Feature Recombination F_recombined = Multimodal_Feature_Recombination_Module(M_t1, M_t1ce, M_t2, M_flair) 2. Encoder Path (3D U-Net based) Encoder_skips = [] Current_features = F_recombined For each encoder_stage: Current_features = ConvReXt_Module(Current_features) Current_features = DGA_Module(Current_features) Add Current_features to Encoder_skips If not last_encoder_stage: Current_features = ConvReXt_Downsampling_Module(Current_features) Bottleneck Bottleneck_features = Current_features 3. Decoder Path Current_features = Bottleneck_features For each decoder_stage: Current_features = Upsample_Module(Current_features) Skip_features_processed = Dense_Multiscale_Module(Encoder_skips.pop()) Current_features = Concatenate(Current_features, Skip_features_processed) Current_features = Dense_Multiscale_Module(Current_features) // Decoder feature refinement 4. Output Segmentation Heads S_wt = Output_Conv_WT(Current_features) S_tc = Output_Conv_TC(Current_features) S_et = Output_Conv_ET(Current_features) Return S_wt, S_tc, S_et End Procedure