Researchers have developed a new unsupervised method for aligning multi-modal medical images, crucial for image-guided therapies. This technique, called Invertible Neural Network-based Registration (INNReg), uses an image-to-image translation network to convert images into a common modality, followed by a registration network for alignment. INNReg incorporates an Invertible Neural Network with a local attention mechanism to maintain geometric consistency and a novel barrier loss function based on Normalized Mutual Information to improve accuracy. Experiments on MRI T1/T2 and MRI/CT datasets demonstrated INNReg's effectiveness. AI
RANK_REASON The item is an academic paper detailing a new method for medical image registration. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →