跳到正文
arXiv cs.AI· Robbie Holland, Ashwin Kumar, Mateus A. Esmeraldo, David Le, Andrea Fisher, Jiazhen Pan, Louisa Fay, Magdalini Paschali, Vasiliki Sideri-Lampretsa, Ates Fettahoglu, Stefania Moroianu, Philipp Wesp, Ivan Lopez, Tom Nuno Wolf, McKenzie S. White, Eduardo Pontes Reis, Fatima Rodriguez, Curtis P. Langlotz, Vasilis Syrgkanis, Imon Banerjee, Akshay S. Chaudhari, Sergios Gatidis·· 10 小时前AI 评分48

Glyph:从临床数据中挖掘 AI 科学发现的可解释性引擎

Glyph: An Interpretability Engine for AI-Driven Scientific Discovery from Clinical Data

AI 导读

研究团队推出 Glyph,一个利用基础模型和稀疏自编码器从多模态临床数据中提取特征、并将其解读为临床概念的多阶段框架。Glyph 在 41,628 名患者的腹部 CT、放射报告和实验室数据中识别出 397 个预测性概念,获得 4,463 项 Bonferroni 显著关联,多数关联可在跨机构留出队列中自动复现。这些概念与文献来源标记形成互补,提升了预测能力,Glyph 已开源发布。

正文 · 原文

Authors:Robbie Holland, Ashwin Kumar, Mateus A. Esmeraldo, David Le, Andrea Fisher, Jiazhen Pan, Louisa Fay, Magdalini Paschali, Vasiliki Sideri-Lampretsa, Ates Fettahoglu, Stefania Moroianu, Philipp Wesp, Ivan Lopez, Tom Nuno Wolf, McKenzie S. White, Eduardo Pontes Reis, Fatima Rodriguez, Curtis P. Langlotz, Vasilis Syrgkanis, Imon Banerjee, Akshay S. Chaudhari, Sergios Gatidis

View PDF HTML (experimental)

Abstract:Automated scientific discovery has the potential to advance understanding of disease mechanisms and accelerate identification of early disease markers. Manual data annotation is infeasible at population scale, motivating methods that automatically extract pre-defined phenotypes from multimodal and unstructured clinical data. Such pre-defined feature sets determine and limit which associations these approaches can discover. This limitation extends to agentic AI systems that use these tools to query population-scale data. In contrast, foundation models (FMs) trained on large volumes of clinical data can perform tasks beyond their training objectives. This suggests their latent embeddings may represent novel or unexplored features that could yield insights into disease mechanisms, yet such latent features remain difficult to interpret. To address this challenge, we introduce Glyph, an automated multistage framework that extracts features from multimodal and unstructured clinical data using FMs and sparse autoencoders, interprets them as coded clinical concepts, and evaluates their value for clinical scientific discovery. We applied Glyph to abdominal CT volumes, CT radiology reports and laboratory measurements from 41,628 patients across different institutions. We identified 397 predictive concepts supported by 4,463 Bonferroni-significant associations. Moreover, a majority of discovered associations were automatically reproducible in held-out cohorts across different sites. Glyph concepts provided complementary associations to literature-derived markers, leading to gains in predictive capability. More broadly, these findings suggest that foundation model features revealed through interpretability methods could expand clinical understanding of disease. We release Glyph at this https URL.
Subjects: Artificial Intelligence (cs.AI)
Cite as: arXiv:2611.08922 [cs.AI]
  (or arXiv:2611.08922v2 [cs.AI] for this version)
  https://doi.org/10.48550/arXiv.2611.08922

arXiv-issued DOI via DataCite

Submission history

From: Robbie Holland [view email]
[v1] Fri, 6 Nov 2026 03:08:38 UTC (17,322 KB)
[v2] Wed, 18 Nov 2026 07:09:55 UTC (23,781 KB)

来源:arXiv cs.AI · arxiv.org