Researchers have developed new self-supervised learning methods for denoising low-dose CT scans, a crucial step for reducing radiation exposure in medical imaging. One approach, Progressive $\mathcal{J}$-Invariant Learning, uses a step-wise mechanism and noise injection to improve denoising efficiency and performance, outperforming existing self-supervised methods on a Mayo LDCT dataset. Another method, Neighbor2Inverse, adapts the Neighbor2Neighbor principle for phase-contrast CT, creating denoising networks from subsampled projections to preserve structural details while suppressing noise, showing promise for both specialized and clinical CT applications. AI
Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →
IMPACT Advances in self-supervised denoising could enable safer medical imaging by reducing radiation dose without sacrificing image quality.
RANK_REASON Two arXiv papers detail novel self-supervised learning methods for low-dose CT denoising.