利用蒙特卡罗树搜索优化医学图像分割神经网络架构

A novel medical image segmentation framework, MNAS-Unet, is proposed, which integrates Monte Carlo Tree Search (MCTS) with Neural Architecture Search (NAS). MNAS-Unet dynamically explores promising network architectures through MCTS, significantly enhancing the efficiency and accuracy of architecture search. It also optimizes the structures of DownSC and UpSC units, enabling rapid and precise model adjustments. Experimental results demonstrate that MNAS-Unet outperforms NAS-Unet and other state-of-the-art models in segmentation accuracy across several medical image datasets, including PROMISE12, Ultrasound Nerve, and CHAOS. Furthermore, compared to NAS-Unet, MNAS-Unet reduces the architecture search budget by 54% with early stopping at 139 epochs.