Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Rise of AI Language Pathologists: Exploring Two-level Prompt Learning for Few-shot Weakly-supervised Whole Slide Image Classification

Linhao Qu, Xiaoyuan Luo, Kexue Fu, Manning Wang, Zhijian Song

TL;DR

This work addresses the problem of Few-shot Weakly Supervised WSI Classification (FSWC) by proposing TOP, a two-level prompt learning MIL framework that fuses GPT-4-derived language priors with CLIP-based vision-language representations. By introducing instance prompt guided pooling and a bag-level prompt group, TOP effectively aggregates patch-level features into robust bag representations and supports few-shot learning under MIL constraints. Empirical results on Camelyon16, TCGA-Lung, and a Cervical cancer WSI dataset show state-of-the-art bag- and instance-level performance across 1–16 labeled bags per class, with substantial improvements over strong baselines. The approach highlights the potential of language-prior-informed prompts to boost pathology WSI classification in data-scarce scenarios, offering a practical path toward scalable AI-assisted pathology.

Abstract

This paper introduces the novel concept of few-shot weakly supervised learning for pathology Whole Slide Image (WSI) classification, denoted as FSWC. A solution is proposed based on prompt learning and the utilization of a large language model, GPT-4. Since a WSI is too large and needs to be divided into patches for processing, WSI classification is commonly approached as a Multiple Instance Learning (MIL) problem. In this context, each WSI is considered a bag, and the obtained patches are treated as instances. The objective of FSWC is to classify both bags and instances with only a limited number of labeled bags. Unlike conventional few-shot learning problems, FSWC poses additional challenges due to its weak bag labels within the MIL framework. Drawing inspiration from the recent achievements of vision-language models (V-L models) in downstream few-shot classification tasks, we propose a two-level prompt learning MIL framework tailored for pathology, incorporating language prior knowledge. Specifically, we leverage CLIP to extract instance features for each patch, and introduce a prompt-guided pooling strategy to aggregate these instance features into a bag feature. Subsequently, we employ a small number of labeled bags to facilitate few-shot prompt learning based on the bag features. Our approach incorporates the utilization of GPT-4 in a question-and-answer mode to obtain language prior knowledge at both the instance and bag levels, which are then integrated into the instance and bag level language prompts. Additionally, a learnable component of the language prompts is trained using the available few-shot labeled data. We conduct extensive experiments on three real WSI datasets encompassing breast cancer, lung cancer, and cervical cancer, demonstrating the notable performance of the proposed method in bag and instance classification. All codes will be available.

The Rise of AI Language Pathologists: Exploring Two-level Prompt Learning for Few-shot Weakly-supervised Whole Slide Image Classification

TL;DR

This work addresses the problem of Few-shot Weakly Supervised WSI Classification (FSWC) by proposing TOP, a two-level prompt learning MIL framework that fuses GPT-4-derived language priors with CLIP-based vision-language representations. By introducing instance prompt guided pooling and a bag-level prompt group, TOP effectively aggregates patch-level features into robust bag representations and supports few-shot learning under MIL constraints. Empirical results on Camelyon16, TCGA-Lung, and a Cervical cancer WSI dataset show state-of-the-art bag- and instance-level performance across 1–16 labeled bags per class, with substantial improvements over strong baselines. The approach highlights the potential of language-prior-informed prompts to boost pathology WSI classification in data-scarce scenarios, offering a practical path toward scalable AI-assisted pathology.

Abstract

This paper introduces the novel concept of few-shot weakly supervised learning for pathology Whole Slide Image (WSI) classification, denoted as FSWC. A solution is proposed based on prompt learning and the utilization of a large language model, GPT-4. Since a WSI is too large and needs to be divided into patches for processing, WSI classification is commonly approached as a Multiple Instance Learning (MIL) problem. In this context, each WSI is considered a bag, and the obtained patches are treated as instances. The objective of FSWC is to classify both bags and instances with only a limited number of labeled bags. Unlike conventional few-shot learning problems, FSWC poses additional challenges due to its weak bag labels within the MIL framework. Drawing inspiration from the recent achievements of vision-language models (V-L models) in downstream few-shot classification tasks, we propose a two-level prompt learning MIL framework tailored for pathology, incorporating language prior knowledge. Specifically, we leverage CLIP to extract instance features for each patch, and introduce a prompt-guided pooling strategy to aggregate these instance features into a bag feature. Subsequently, we employ a small number of labeled bags to facilitate few-shot prompt learning based on the bag features. Our approach incorporates the utilization of GPT-4 in a question-and-answer mode to obtain language prior knowledge at both the instance and bag levels, which are then integrated into the instance and bag level language prompts. Additionally, a learnable component of the language prompts is trained using the available few-shot labeled data. We conduct extensive experiments on three real WSI datasets encompassing breast cancer, lung cancer, and cervical cancer, demonstrating the notable performance of the proposed method in bag and instance classification. All codes will be available.
Paper Structure (18 sections, 9 equations, 2 figures, 7 tables)

This paper contains 18 sections, 9 equations, 2 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Multiple Instance Learning for WSI Classification
  4. Few Shot Classification
  5. Vision-language Models and Prompt Learning
  6. Preliminaries
  7. Problem Formulation
  8. Vision-Language Pre-training and Few-shot Prompt Learning
  9. Method
  10. Construction of Instance and Bag-level Prompt
  11. Instance Prompt Guided Pooling
  12. Experiment
  13. Datasets, Evaluation Metrics and Comparison Methods
  14. Implementation Details
  15. Results on the Camelyon 16 Dataset
  16. ...and 3 more sections

Figures (2)

  • Figure 1: A. Existing WSI classification tasks. B. Our proposed FSWC task based on language models. C. Existing prompt learning paradigm using pre-trained V-L models, where the parameters of the V-L models are always frozen. D. Our proposed Two-level Prompt Learning paradigm.
  • Figure 2: Framework of our proposed TOP.