๐‘๐ž๐ฌ๐ž๐š๐ซ๐œ๐กไธจAnatomically-Grounded Radiology Report Generation by Liu Chang๐‘๐ž๐ฌ๐ž๐š๐ซ๐œ๐กไธจAnatomically-Grounded Radiology Report Generation by Liu Chang

๐‘๐ž๐ฌ๐ž๐š๐ซ๐œ๐กไธจAnatomically-Grounded Radiology Report Generation

Liu Chang

Liu Chang

AI Researcher & Co-author
๐Œ๐จ๐ญ๐ข๐ฏ๐š๐ญ๐ข๐จ๐ง & ๐‚๐ก๐š๐ฅ๐ฅ๐ž๐ง๐ ๐ž: Image-report pairing alone provides weak anatomical grounding. To improve this issue, we introduce S2D-Align, which progressively adds reference reports and anatomical phrases, with a memory adapter sharing information across alignment stages.
๐‘๐ž๐ฌ๐ฎ๐ฅ๐ญ๐ฌ & ๐•๐ž๐ง๐ฎ๐ž๐ฌ: Improved radiology report generation on MIMIC-CXR and IU X-Ray through auxiliary-guided fine-tuning. Published at AAAI 2026.
๐‹๐ข๐ง๐ค๏ผšhttps://ojs.aaai.org/index.php/AAAI/article/download/40335/44296
Like this project

Posted Sep 15, 2026

S2D-Align progressively aligns reference reports and anatomical phrases using a memory adapter. Published at AAAI 2026.