Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Aligning Medical Images with General Knowledge from Large Language Models

Xiao Fang, Yi Lin, Dong Zhang, Kwang-Ting Cheng, Hao Chen

TL;DR

This work proposes ViP, a novel visual symptom-guided prompt learning framework for medical image analysis, which facilitates general knowledge transfer from CLIP, and demonstrates that ViP can outperform state-of-the-art methods on two challenging datasets.

Abstract

Pre-trained large vision-language models (VLMs) like CLIP have revolutionized visual representation learning using natural language as supervisions, and demonstrated promising generalization ability. In this work, we propose ViP, a novel visual symptom-guided prompt learning framework for medical image analysis, which facilitates general knowledge transfer from CLIP. ViP consists of two key components: a visual symptom generator (VSG) and a dual-prompt network. Specifically, VSG aims to extract explicable visual symptoms from pre-trained large language models, while the dual-prompt network utilizes these visual symptoms to guide the training on two learnable prompt modules, i.e., context prompt and merge prompt, which effectively adapts our framework to medical image analysis via large VLMs. Extensive experimental results demonstrate that ViP can outperform state-of-the-art methods on two challenging datasets.

Aligning Medical Images with General Knowledge from Large Language Models

TL;DR

This work proposes ViP, a novel visual symptom-guided prompt learning framework for medical image analysis, which facilitates general knowledge transfer from CLIP, and demonstrates that ViP can outperform state-of-the-art methods on two challenging datasets.

Abstract

Pre-trained large vision-language models (VLMs) like CLIP have revolutionized visual representation learning using natural language as supervisions, and demonstrated promising generalization ability. In this work, we propose ViP, a novel visual symptom-guided prompt learning framework for medical image analysis, which facilitates general knowledge transfer from CLIP. ViP consists of two key components: a visual symptom generator (VSG) and a dual-prompt network. Specifically, VSG aims to extract explicable visual symptoms from pre-trained large language models, while the dual-prompt network utilizes these visual symptoms to guide the training on two learnable prompt modules, i.e., context prompt and merge prompt, which effectively adapts our framework to medical image analysis via large VLMs. Extensive experimental results demonstrate that ViP can outperform state-of-the-art methods on two challenging datasets.
Paper Structure (12 sections, 3 equations, 5 figures, 2 tables)

This paper contains 12 sections, 3 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Overall Pipeline
  4. Visual Symptom Generator (VSG)
  5. Dual-Prompt Network
  6. Experiments
  7. Dataset and Implementation Details
  8. Comparisons with State-of-the-art Methods
  9. Ablation Study
  10. Conclusion
  11. Acknowledgments.
  12. Disclosure of Interests.

Figures (5)

  • Figure 1: Overview of ViP, which consists of a visual symptom generator (VSG) and a dual-prompt network. The visual symptoms predicted by VSG are used as inputs for downstream networks (marked by the blue dashed line).
  • Figure 2: Example visual symptoms generated by GPT-4 achiam2023gpt.
  • Figure 3: (I) Zero-shot CLIP with category name or visual symptoms as text inputs. (II) Diagnostic process based on cosine similarity scores between images and visual symptoms.
  • Figure 4: Failure cases in the zero-shot experiment.
  • Figure 5: Ablation study comparing with different types of knowledge.