Input Modalities (MRI Image, CT Image, PET Image)

Edge Detection (Enhances boundaries in each input)

Cross-Modal Attention (Fuses MRI, CT, PET features by learning inter-modality relations)

Dynamic Feature Gating (Weights modalities dynamically based on relevance for segmentation)

EfficientNet-B4 Encoder + UNet Decoder (EfficientUNet)

Pixel-wise Tumor Segmentation