Research

Paper

AI LLM February 26, 2026

PRIMA: Pre-training with Risk-integrated Image-Metadata Alignment for Medical Diagnosis via LLM

Authors

Yiqing Wang, Chunming He, Ming-Chen Lu, Mercy Pawar, Leslie Niziol, Maria Woodward, Sina Farsiu

Abstract

Medical diagnosis requires the effective synthesis of visual manifestations and clinical metadata. However, existing methods often treat metadata as isolated tags, failing to exploit the rich semantic knowledge embedded in clinical descriptions. We propose PRIMA (Pre-training with Risk-integrated Image-Metadata Alignment), a framework that integrates domain-specific knowledge into multi-modal representation learning. We first curate an expert corpus of risk-disease correlations via Retrieval-Augmented Generation (RAG) to refine Clinical ModernBERT, embedding diagnostic priors into the text encoder. To bridge the modality gap, we introduce a dual-encoder pre-training strategy utilizing DINOv3 and our refined BERT, optimized by a suite of four complementary loss functions. These losses are designed to capture multi-granular semantic alignment and handle the ambiguity of clinical correlations through soft labels. Finally, we leverage Qwen-3 to fuse these aligned features for precise disease classification. Extensive experiments demonstrate that PRIMA effectively harmonizes pixel-level features with abstract clinical expertise, significantly outperforming other state-of-the-art methods. Notably, our framework achieves superior robustness without the need for massive data collection or exhaustive computational resources. Our code will be made public upon acceptance.

Metadata

arXiv ID: 2602.23297
Provider: ARXIV
Primary Category: cs.CV
Published: 2026-02-26
Fetched: 2026-02-27 04:35

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2602.23297v1</id>\n    <title>PRIMA: Pre-training with Risk-integrated Image-Metadata Alignment for Medical Diagnosis via LLM</title>\n    <updated>2026-02-26T18:07:52Z</updated>\n    <link href='https://arxiv.org/abs/2602.23297v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2602.23297v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Medical diagnosis requires the effective synthesis of visual manifestations and clinical metadata. However, existing methods often treat metadata as isolated tags, failing to exploit the rich semantic knowledge embedded in clinical descriptions. We propose PRIMA (Pre-training with Risk-integrated Image-Metadata Alignment), a framework that integrates domain-specific knowledge into multi-modal representation learning. We first curate an expert corpus of risk-disease correlations via Retrieval-Augmented Generation (RAG) to refine Clinical ModernBERT, embedding diagnostic priors into the text encoder. To bridge the modality gap, we introduce a dual-encoder pre-training strategy utilizing DINOv3 and our refined BERT, optimized by a suite of four complementary loss functions. These losses are designed to capture multi-granular semantic alignment and handle the ambiguity of clinical correlations through soft labels. Finally, we leverage Qwen-3 to fuse these aligned features for precise disease classification. Extensive experiments demonstrate that PRIMA effectively harmonizes pixel-level features with abstract clinical expertise, significantly outperforming other state-of-the-art methods. Notably, our framework achieves superior robustness without the need for massive data collection or exhaustive computational resources. Our code will be made public upon acceptance.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CV'/>\n    <published>2026-02-26T18:07:52Z</published>\n    <arxiv:primary_category term='cs.CV'/>\n    <author>\n      <name>Yiqing Wang</name>\n    </author>\n    <author>\n      <name>Chunming He</name>\n    </author>\n    <author>\n      <name>Ming-Chen Lu</name>\n    </author>\n    <author>\n      <name>Mercy Pawar</name>\n    </author>\n    <author>\n      <name>Leslie Niziol</name>\n    </author>\n    <author>\n      <name>Maria Woodward</name>\n    </author>\n    <author>\n      <name>Sina Farsiu</name>\n    </author>\n  </entry>"
}