A PET/CT Cross-Modal Wavelet Fusion and Pseudo-Mask Guided Network With Frozen SAM Decoder for Multiple Myeloma Segmentation

Key Points

• The study aims to enhance PET/CT segmentation for multiple myeloma using a new network design.
• Developed a cross-modal wavelet fusion approach.
• Implemented a pseudo-mask guided network with a frozen SAM decoder.
• Evaluated the model's performance across different datasets.
• Achieved strong accuracy in segmentation.
• Showed generalizability across multiple datasets.
• Improved computational efficiency compared to existing methods.