Abstract:

In the medical sector, automatic segmentation procedures are often employed for detecting multiple diseases using diverse forms of medical records such as Magnetic Resonance images (MRI), Computed Tomography (CT) scans, etc. However, it is hard to ensure superior segmentation performance over multi-modal data due to the inherent heterogeneity of medical images. Traditional segmentation models frequently struggle due to poor generalizability, when handling multimodal data. Furthermore, employing multi-modal medical data to guarantee precise segmentation performance frequently requires advanced encoder-decoder models. This modification is made by introducing diverse encoding or decoding units for processing each specific modality. In this work, an advanced attention mechanism-based deep learning network is implemented to effectively learn complex heterogeneous features in medical images. From the standard public databases, the required multi-modal images are gathered, and it is directly offered to the segmentation model for performing multi-modal segmentation. The segmentation process is executed using Multi-head Gated cross Attention Fusion Encoder-based Adaptive Transformer Mobile-UNet++ with Consistency Loss Function (MGAFE-ATMU++-CLF), which provides robust and efficient multi-modal segmentation. The segmentation accuracy over multi-modal images is enhanced through the optimization of parameters from the MGAFE-ATMU++ model using the Renovated Puma Optimizer (RPO), which adjusts the model parameters to provide better effectiveness. The developed MGAFE-ATMU++-CLF model primarily focuses on robust feature extraction, effective cross-modal information fusion, dynamic adaptation, and enhanced boundary delineation. The experimental outcomes with the suggested network are validated with the standard works to ensure its segmentation effectiveness.

Description:

.