A-Lite UACNet: Hybrid Attention U-Net with Active Contour for Efficient Brain Tumor Segmentation

Accurate segmentation of the brain tumor is important for diagnostic as well as treatment purposes, but robust segmentation within strict computational budgets is difficult because of large imbalances, small tumor regions, and insufficient tumor boundaries. This paper proposes A-Lite-UACNet, a lightweight efficient deep learning architecture for robust segmentation inspired from active contour models for rapid Brain Tumor Delineation. Furthermore, proposed system utilizes the concept of Omni learning for both tumor segmentation as well as tumor image categorization, which stabilizes the entire feature representation for better performance. The proposed sys- tem utilizes the lightweight version of the encoder portion based on the MobileNetV2 model for efficient feature representation with UACNet for better segmentation accuracy within lower computational time with strict computational budgets. Moreover, its lightweight model minimizes both memory consumption as well as increased latency during inference for fast inference with strict computational expenses for efficient segmentation within realtime medical settings with the proposed BRISC-2025 dataset for ultra-fast inference with high precision, as described in detail within this paper for confirmation of proposed efficient segmentation with ultra-fast inference performance for efficient medical segmentation within the real-time medical domain. Conclusion