Researchers have developed a novel physics-driven framework for self-supervised low-dose computed tomography (LDCT) denoising. This method explicitly models the mixed Poisson-Gaussian noise inherent in LDCT measurements, a limitation in many existing self-supervised techniques. The framework separates noise components, processes them through distinct thinning operations to create independent noise realizations, and uses these to train an image-domain network. Experiments on simulated and real LDCT data demonstrate significant improvements over current self-supervised methods, achieving performance comparable to supervised approaches. AI
RANK_REASON The cluster contains a research paper published on arXiv detailing a new method for medical image processing. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
