├── .history ├── README_20250625214617.md ├── README_20250625214623.md ├── README_20250625214635.md ├── README_20250625214611.md ├── README_20250625215919.md ├── README_20250625214812.md ├── README_20250625215203.md ├── README_20250625220008.md ├── README_20250625214804.md ├── README_20250625220309.md ├── README_20250625231735.md ├── README_20250625231736.md ├── README_20250625231832.md ├── README_20250625231728.md ├── README_20250625231733.md ├── README_20250625214751.md ├── README_20250625214747.md └── README_20250625220230.md ├── static └── image │ ├── fig1.png │ └── wsi-agents.png └── README.md /.history/README_20250625214617.md: -------------------------------------------------------------------------------- 1 | # WSI-Agents -------------------------------------------------------------------------------- /.history/README_20250625214623.md: -------------------------------------------------------------------------------- 1 | # WSI-Agents -------------------------------------------------------------------------------- /.history/README_20250625214635.md: -------------------------------------------------------------------------------- 1 | # WSI-Agents -------------------------------------------------------------------------------- /.history/README_20250625214611.md: -------------------------------------------------------------------------------- 1 | # WSI-Agents 2 | -------------------------------------------------------------------------------- /static/image/fig1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/XinhengLyu/WSI-Agents/HEAD/static/image/fig1.png -------------------------------------------------------------------------------- /static/image/wsi-agents.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/XinhengLyu/WSI-Agents/HEAD/static/image/wsi-agents.png -------------------------------------------------------------------------------- /.history/README_20250625215919.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://arxiv.org/abs/placeholder) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image (WSI) analysis that integrates specialized functional agents with robust verification mechanisms to enhance both task-specific accuracy and multi-task versatility. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 🏗️ System Overview 14 | 15 | WSI-Agents consists of three main modules: 16 | - **Task Allocation Module**: Assigns tasks to specialized expert agents 17 | - **Verification Mechanism**: Internal consistency + external knowledge validation 18 | - **Summary Module**: Synthesizes results with visual interpretation maps 19 | 20 | ![WSI-Agents Workflow](static/image/fig1.png) 21 | *Need: Figure 1 - The workflow diagram* 22 | 23 | ![WSI-Agents Architecture](static/image/figure2.png) 24 | *Need: Figure 2 - Detailed system architecture* 25 | 26 | --- 27 | 28 | ## 🎯 Key Features 29 | 30 | - 🎯 **Specialized Expert Agents**: Morphology, diagnosis, treatment planning, and report generation 31 | - ✅ **Dual Verification System**: Internal consistency checking + external knowledge validation 32 | - 🧠 **Knowledge-Driven**: Leverages pathology knowledge bases and WSI foundation models 33 | - 📊 **Superior Performance**: Outperforms existing WSI MLLMs and medical agents by 10-17% 34 | - 🔍 **Visual Interpretation**: Generates comprehensive attention maps for explainable analysis 35 | 36 | --- 37 | 38 | ## 📊 Results Highlights 39 | 40 | - **WSI-Bench**: Achieves 70.3% average performance, surpassing previous methods 41 | - **WSI-VQA**: Reaches 60.0% accuracy, outperforming WSI-LLaVA (55.0%) and others 42 | - **Report Generation**: 44.0% accuracy with significant improvements across all metrics 43 | - **Consistent Improvements**: 10-17% performance gains across diverse pathology tasks 44 | 45 | --- 46 | 47 | ## 📂 Resources 48 | 49 | 🚀 **Code**: Coming Soon 50 | 📄 **Paper**: Coming Soon 51 | 52 | --- -------------------------------------------------------------------------------- /.history/README_20250625214812.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://arxiv.org/abs/placeholder) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image (WSI) analysis that integrates specialized functional agents with robust verification mechanisms to enhance both task-specific accuracy and multi-task versatility. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 🏗️ System Overview 14 | 15 | WSI-Agents consists of three main modules: 16 | - **Task Allocation Module**: Assigns tasks to specialized expert agents 17 | - **Verification Mechanism**: Internal consistency + external knowledge validation 18 | - **Summary Module**: Synthesizes results with visual interpretation maps 19 | 20 | ![WSI-Agents Workflow](static/image/figure1.png) 21 | *Need: Figure 1 - The workflow diagram* 22 | 23 | ![WSI-Agents Architecture](static/image/figure2.png) 24 | *Need: Figure 2 - Detailed system architecture* 25 | 26 | --- 27 | 28 | ## 🎯 Key Features 29 | 30 | - 🎯 **Specialized Expert Agents**: Morphology, diagnosis, treatment planning, and report generation 31 | - ✅ **Dual Verification System**: Internal consistency checking + external knowledge validation 32 | - 🧠 **Knowledge-Driven**: Leverages pathology knowledge bases and WSI foundation models 33 | - 📊 **Superior Performance**: Outperforms existing WSI MLLMs and medical agents by 10-17% 34 | - 🔍 **Visual Interpretation**: Generates comprehensive attention maps for explainable analysis 35 | 36 | --- 37 | 38 | ## 📊 Results Highlights 39 | 40 | - **WSI-Bench**: Achieves 70.3% average performance, surpassing previous methods 41 | - **WSI-VQA**: Reaches 60.0% accuracy, outperforming WSI-LLaVA (55.0%) and others 42 | - **Report Generation**: 44.0% accuracy with significant improvements across all metrics 43 | - **Consistent Improvements**: 10-17% performance gains across diverse pathology tasks 44 | 45 | --- 46 | 47 | ## 📂 Resources 48 | 49 | 🚀 **Code**: Coming Soon 50 | 📄 **Paper**: Coming Soon 51 | 52 | --- -------------------------------------------------------------------------------- /.history/README_20250625215203.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://arxiv.org/abs/placeholder) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image (WSI) analysis that integrates specialized functional agents with robust verification mechanisms to enhance both task-specific accuracy and multi-task versatility. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 🏗️ System Overview 14 | 15 | WSI-Agents consists of three main modules: 16 | - **Task Allocation Module**: Assigns tasks to specialized expert agents 17 | - **Verification Mechanism**: Internal consistency + external knowledge validation 18 | - **Summary Module**: Synthesizes results with visual interpretation maps 19 | 20 | ![WSI-Agents Workflow](static/image/figure1.png) 21 | *Need: Figure 1 - The workflow diagram* 22 | 23 | ![WSI-Agents Architecture](static/image/figure2.png) 24 | *Need: Figure 2 - Detailed system architecture* 25 | 26 | --- 27 | 28 | ## 🎯 Key Features 29 | 30 | - 🎯 **Specialized Expert Agents**: Morphology, diagnosis, treatment planning, and report generation 31 | - ✅ **Dual Verification System**: Internal consistency checking + external knowledge validation 32 | - 🧠 **Knowledge-Driven**: Leverages pathology knowledge bases and WSI foundation models 33 | - 📊 **Superior Performance**: Outperforms existing WSI MLLMs and medical agents by 10-17% 34 | - 🔍 **Visual Interpretation**: Generates comprehensive attention maps for explainable analysis 35 | 36 | --- 37 | 38 | ## 📊 Results Highlights 39 | 40 | - **WSI-Bench**: Achieves 70.3% average performance, surpassing previous methods 41 | - **WSI-VQA**: Reaches 60.0% accuracy, outperforming WSI-LLaVA (55.0%) and others 42 | - **Report Generation**: 44.0% accuracy with significant improvements across all metrics 43 | - **Consistent Improvements**: 10-17% performance gains across diverse pathology tasks 44 | 45 | --- 46 | 47 | ## 📂 Resources 48 | 49 | 🚀 **Code**: Coming Soon 50 | 📄 **Paper**: Coming Soon 51 | 52 | --- -------------------------------------------------------------------------------- /.history/README_20250625220008.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://arxiv.org/abs/placeholder) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image (WSI) analysis that integrates specialized functional agents with robust verification mechanisms to enhance both task-specific accuracy and multi-task versatility. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 🏗️ System Overview 14 | 15 | WSI-Agents consists of three main modules: 16 | - **Task Allocation Module**: Assigns tasks to specialized expert agents 17 | - **Verification Mechanism**: Internal consistency + external knowledge validation 18 | - **Summary Module**: Synthesizes results with visual interpretation maps 19 | 20 | ![WSI-Agents Workflow](static/image/fig1.png) 21 | *Need: Figure 1 - The workflow diagram* 22 | 23 | ![WSI-Agents Architecture](static/image/wsi-agents.png) 24 | *Need: Figure 2 - Detailed system architecture* 25 | 26 | --- 27 | 28 | ## 🎯 Key Features 29 | 30 | - 🎯 **Specialized Expert Agents**: Morphology, diagnosis, treatment planning, and report generation 31 | - ✅ **Dual Verification System**: Internal consistency checking + external knowledge validation 32 | - 🧠 **Knowledge-Driven**: Leverages pathology knowledge bases and WSI foundation models 33 | - 📊 **Superior Performance**: Outperforms existing WSI MLLMs and medical agents by 10-17% 34 | - 🔍 **Visual Interpretation**: Generates comprehensive attention maps for explainable analysis 35 | 36 | --- 37 | 38 | ## 📊 Results Highlights 39 | 40 | - **WSI-Bench**: Achieves 70.3% average performance, surpassing previous methods 41 | - **WSI-VQA**: Reaches 60.0% accuracy, outperforming WSI-LLaVA (55.0%) and others 42 | - **Report Generation**: 44.0% accuracy with significant improvements across all metrics 43 | - **Consistent Improvements**: 10-17% performance gains across diverse pathology tasks 44 | 45 | --- 46 | 47 | ## 📂 Resources 48 | 49 | 🚀 **Code**: Coming Soon 50 | 📄 **Paper**: Coming Soon 51 | 52 | --- -------------------------------------------------------------------------------- /.history/README_20250625214804.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://arxiv.org/abs/placeholder) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image (WSI) analysis that integrates specialized functional agents with robust verification mechanisms to enhance both task-specific accuracy and multi-task versatility. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 🏗️ System Overview 14 | 15 | WSI-Agents consists of three main modules: 16 | - **Task Allocation Module**: Assigns tasks to specialized expert agents 17 | - **Verification Mechanism**: Internal consistency + external knowledge validation 18 | - **Summary Module**: Synthesizes results with visual interpretation maps 19 | 20 | ![WSI-Agents Workflow](static/image/figure1.png) 21 | *Need: Figure 1 - The workflow diagram* 22 | 23 | ![WSI-Agents Architecture](static/image/figure2.png) 24 | *Need: Figure 2 - Detailed system architecture* 25 | 26 | --- 27 | 28 | ## 🎯 Key Features 29 | 30 | - 🎯 **Specialized Expert Agents**: Morphology, diagnosis, treatment planning, and report generation 31 | - ✅ **Dual Verification System**: Internal consistency checking + external knowledge validation 32 | - 🧠 **Knowledge-Driven**: Leverages pathology knowledge bases and WSI foundation models 33 | - 📊 **Superior Performance**: Outperforms existing WSI MLLMs and medical agents by 10-17% 34 | - 🔍 **Visual Interpretation**: Generates comprehensive attention maps for explainable analysis 35 | 36 | --- 37 | 38 | ## 📊 Results Highlights 39 | 40 | - **WSI-Bench**: Achieves 70.3% average performance, surpassing previous methods 41 | - **WSI-VQA**: Reaches 60.0% accuracy, outperforming WSI-LLaVA (55.0%) and others 42 | - **Report Generation**: 44.0% accuracy with significant improvements across all metrics 43 | - **Consistent Improvements**: 10-17% performance gains across diverse pathology tasks 44 | 45 | --- 46 | 47 | ## 📂 Resources 48 | 49 | 🚀 **Code**: Coming Soon 50 | 📄 **Paper**: Coming Soon 51 | 📊 **Dataset**: Coming Soon 52 | 🎯 **Pre-trained Models**: Coming Soon 53 | 54 | --- -------------------------------------------------------------------------------- /.history/README_20250625220309.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://arxiv.org/abs/placeholder) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image analysis that bridges the gap between accuracy and versatility in digital pathology through specialized agents and robust verification mechanisms. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 📂 Resources 14 | 15 | 🚀 **Code**: Coming Soon 16 | 📄 **Paper**: Coming Soon 17 | 18 | --- 19 | 20 | ## 🏗️ System Overview 21 | 22 | WSI-Agents employs a collaborative multi-agent approach to address the accuracy-versatility trade-off in WSI analysis. The system integrates specialized expert agents with comprehensive verification mechanisms to ensure clinical accuracy while maintaining multi-task capabilities. 23 | 24 | ![WSI-Agents Workflow](static/image/fig1.png) 25 | 26 | The architecture consists of three core components: a task allocation module that assigns specialized expert agents, verification mechanisms that ensure accuracy through internal consistency and external knowledge validation, and a summary module that synthesizes final results with visual interpretation. 27 | 28 | ![WSI-Agents Architecture](static/image/wsi-agents.png) 29 | 30 | --- 31 | 32 | ## 🎯 Key Innovations 33 | 34 | - **Multi-Agent Collaboration**: Specialized agents for morphology, diagnosis, treatment planning, and report generation 35 | - **Dual Verification**: Internal consistency checking combined with external knowledge validation 36 | - **Knowledge Integration**: Leverages pathology knowledge bases and WSI foundation models 37 | - **Visual Interpretation**: Comprehensive attention maps for explainable analysis 38 | 39 | --- 40 | 41 | ## 📊 Performance Highlights 42 | 43 | - **WSI-Bench**: Achieves superior performance across morphological analysis, diagnosis, and treatment planning 44 | - **WSI-VQA**: 60.0% accuracy, outperforming existing WSI MLLMs by significant margins 45 | - **Report Generation**: Substantial improvements in clinical report quality and accuracy 46 | - **Consistent Gains**: 10-17% performance improvements across diverse pathology tasks 47 | 48 | --- 49 | -------------------------------------------------------------------------------- /.history/README_20250625231735.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://drive.google.com/file/d/1T8aTBL_-JIZpKoRbvvYmoj2JisXssQMr/view) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image analysis that bridges the gap between accuracy and versatility in digital pathology through specialized agents and robust verification mechanisms. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 📂 Resources 14 | 15 | 🚀 **Code**: Coming Soon 16 | --- 17 | 18 | ## 🏗️ System Overview 19 | 20 | WSI-Agents employs a collaborative multi-agent approach to address the accuracy-versatility trade-off in WSI analysis. The system integrates specialized expert agents with comprehensive verification mechanisms to ensure clinical accuracy while maintaining multi-task capabilities. 21 | 22 | ![WSI-Agents Workflow](static/image/fig1.png) 23 | 24 | 25 | The architecture consists of three core components: a task allocation module that assigns specialized expert agents, verification mechanisms that ensure accuracy through internal consistency and external knowledge validation, and a summary module that synthesizes final results with visual interpretation. 26 | 27 | ![WSI-Agents Architecture](static/image/wsi-agents.png) 28 | 29 | --- 30 | 31 | ## 🎯 Key Innovations 32 | 33 | - **Multi-Agent Collaboration**: Specialized agents for morphology, diagnosis, treatment planning, and report generation 34 | - **Dual Verification**: Internal consistency checking combined with external knowledge validation 35 | - **Knowledge Integration**: Leverages pathology knowledge bases and WSI foundation models 36 | - **Visual Interpretation**: Comprehensive attention maps for explainable analysis 37 | 38 | --- 39 | 40 | ## 📊 Performance Highlights 41 | 42 | - **WSI-Bench**: Achieves superior performance across morphological analysis, diagnosis, and treatment planning 43 | - **WSI-VQA**: 60.0% accuracy, outperforming existing WSI MLLMs by significant margins 44 | - **Report Generation**: Substantial improvements in clinical report quality and accuracy 45 | - **Consistent Gains**: 10-17% performance improvements across diverse pathology tasks 46 | 47 | --- 48 | -------------------------------------------------------------------------------- /.history/README_20250625231736.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://drive.google.com/file/d/1T8aTBL_-JIZpKoRbvvYmoj2JisXssQMr/view) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image analysis that bridges the gap between accuracy and versatility in digital pathology through specialized agents and robust verification mechanisms. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 📂 Resources 14 | 15 | 🚀 **Code**: Coming Soon 16 | 17 | --- 18 | 19 | ## 🏗️ System Overview 20 | 21 | WSI-Agents employs a collaborative multi-agent approach to address the accuracy-versatility trade-off in WSI analysis. The system integrates specialized expert agents with comprehensive verification mechanisms to ensure clinical accuracy while maintaining multi-task capabilities. 22 | 23 | ![WSI-Agents Workflow](static/image/fig1.png) 24 | 25 | 26 | The architecture consists of three core components: a task allocation module that assigns specialized expert agents, verification mechanisms that ensure accuracy through internal consistency and external knowledge validation, and a summary module that synthesizes final results with visual interpretation. 27 | 28 | ![WSI-Agents Architecture](static/image/wsi-agents.png) 29 | 30 | --- 31 | 32 | ## 🎯 Key Innovations 33 | 34 | - **Multi-Agent Collaboration**: Specialized agents for morphology, diagnosis, treatment planning, and report generation 35 | - **Dual Verification**: Internal consistency checking combined with external knowledge validation 36 | - **Knowledge Integration**: Leverages pathology knowledge bases and WSI foundation models 37 | - **Visual Interpretation**: Comprehensive attention maps for explainable analysis 38 | 39 | --- 40 | 41 | ## 📊 Performance Highlights 42 | 43 | - **WSI-Bench**: Achieves superior performance across morphological analysis, diagnosis, and treatment planning 44 | - **WSI-VQA**: 60.0% accuracy, outperforming existing WSI MLLMs by significant margins 45 | - **Report Generation**: Substantial improvements in clinical report quality and accuracy 46 | - **Consistent Gains**: 10-17% performance improvements across diverse pathology tasks 47 | 48 | --- 49 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://conferences.miccai.org/2025/en/default.asp) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://drive.google.com/file/d/1T8aTBL_-JIZpKoRbvvYmoj2JisXssQMr/view) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image analysis that bridges the gap between accuracy and versatility in digital pathology through specialized agents and robust verification mechanisms. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 📂 Resources 14 | 15 | 🚀 **Code**: Coming Soon 16 | 17 | --- 18 | 19 | ## 🏗️ System Overview 20 | 21 | WSI-Agents employs a collaborative multi-agent approach to address the accuracy-versatility trade-off in WSI analysis. The system integrates specialized expert agents with comprehensive verification mechanisms to ensure clinical accuracy while maintaining multi-task capabilities. 22 | 23 | ![WSI-Agents Workflow](static/image/fig1.png) 24 | 25 | 26 | The architecture consists of three core components: a task allocation module that assigns specialized expert agents, verification mechanisms that ensure accuracy through internal consistency and external knowledge validation, and a summary module that synthesizes final results with visual interpretation. 27 | 28 | ![WSI-Agents Architecture](static/image/wsi-agents.png) 29 | 30 | --- 31 | 32 | ## 🎯 Key Innovations 33 | 34 | - **Multi-Agent Collaboration**: Specialized agents for morphology, diagnosis, treatment planning, and report generation 35 | - **Dual Verification**: Internal consistency checking combined with external knowledge validation 36 | - **Knowledge Integration**: Leverages pathology knowledge bases and WSI foundation models 37 | - **Visual Interpretation**: Comprehensive attention maps for explainable analysis 38 | 39 | --- 40 | 41 | ## 📊 Performance Highlights 42 | 43 | - **WSI-Bench**: Achieves superior performance across morphological analysis, diagnosis, and treatment planning 44 | - **WSI-VQA**: 60.0% accuracy, outperforming existing WSI MLLMs by significant margins 45 | - **Report Generation**: Substantial improvements in clinical report quality and accuracy 46 | - **Consistent Gains**: 10-17% performance improvements across diverse pathology tasks 47 | 48 | --- 49 | -------------------------------------------------------------------------------- /.history/README_20250625231832.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://conferences.miccai.org/2025/en/default.asp) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://drive.google.com/file/d/1T8aTBL_-JIZpKoRbvvYmoj2JisXssQMr/view) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image analysis that bridges the gap between accuracy and versatility in digital pathology through specialized agents and robust verification mechanisms. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 📂 Resources 14 | 15 | 🚀 **Code**: Coming Soon 16 | 17 | --- 18 | 19 | ## 🏗️ System Overview 20 | 21 | WSI-Agents employs a collaborative multi-agent approach to address the accuracy-versatility trade-off in WSI analysis. The system integrates specialized expert agents with comprehensive verification mechanisms to ensure clinical accuracy while maintaining multi-task capabilities. 22 | 23 | ![WSI-Agents Workflow](static/image/fig1.png) 24 | 25 | 26 | The architecture consists of three core components: a task allocation module that assigns specialized expert agents, verification mechanisms that ensure accuracy through internal consistency and external knowledge validation, and a summary module that synthesizes final results with visual interpretation. 27 | 28 | ![WSI-Agents Architecture](static/image/wsi-agents.png) 29 | 30 | --- 31 | 32 | ## 🎯 Key Innovations 33 | 34 | - **Multi-Agent Collaboration**: Specialized agents for morphology, diagnosis, treatment planning, and report generation 35 | - **Dual Verification**: Internal consistency checking combined with external knowledge validation 36 | - **Knowledge Integration**: Leverages pathology knowledge bases and WSI foundation models 37 | - **Visual Interpretation**: Comprehensive attention maps for explainable analysis 38 | 39 | --- 40 | 41 | ## 📊 Performance Highlights 42 | 43 | - **WSI-Bench**: Achieves superior performance across morphological analysis, diagnosis, and treatment planning 44 | - **WSI-VQA**: 60.0% accuracy, outperforming existing WSI MLLMs by significant margins 45 | - **Report Generation**: Substantial improvements in clinical report quality and accuracy 46 | - **Consistent Gains**: 10-17% performance improvements across diverse pathology tasks 47 | 48 | --- 49 | -------------------------------------------------------------------------------- /.history/README_20250625231728.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://drive.google.com/file/d/1T8aTBL_-JIZpKoRbvvYmoj2JisXssQMr/view) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image analysis that bridges the gap between accuracy and versatility in digital pathology through specialized agents and robust verification mechanisms. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 📂 Resources 14 | 15 | 🚀 **Code**: Coming Soon 16 | 📄 **Paper**: Coming Soon 17 | 18 | --- 19 | 20 | ## 🏗️ System Overview 21 | 22 | WSI-Agents employs a collaborative multi-agent approach to address the accuracy-versatility trade-off in WSI analysis. The system integrates specialized expert agents with comprehensive verification mechanisms to ensure clinical accuracy while maintaining multi-task capabilities. 23 | 24 | ![WSI-Agents Workflow](static/image/fig1.png) 25 | 26 | 27 | The architecture consists of three core components: a task allocation module that assigns specialized expert agents, verification mechanisms that ensure accuracy through internal consistency and external knowledge validation, and a summary module that synthesizes final results with visual interpretation. 28 | 29 | ![WSI-Agents Architecture](static/image/wsi-agents.png) 30 | 31 | --- 32 | 33 | ## 🎯 Key Innovations 34 | 35 | - **Multi-Agent Collaboration**: Specialized agents for morphology, diagnosis, treatment planning, and report generation 36 | - **Dual Verification**: Internal consistency checking combined with external knowledge validation 37 | - **Knowledge Integration**: Leverages pathology knowledge bases and WSI foundation models 38 | - **Visual Interpretation**: Comprehensive attention maps for explainable analysis 39 | 40 | --- 41 | 42 | ## 📊 Performance Highlights 43 | 44 | - **WSI-Bench**: Achieves superior performance across morphological analysis, diagnosis, and treatment planning 45 | - **WSI-VQA**: 60.0% accuracy, outperforming existing WSI MLLMs by significant margins 46 | - **Report Generation**: Substantial improvements in clinical report quality and accuracy 47 | - **Consistent Gains**: 10-17% performance improvements across diverse pathology tasks 48 | 49 | --- 50 | -------------------------------------------------------------------------------- /.history/README_20250625231733.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://drive.google.com/file/d/1T8aTBL_-JIZpKoRbvvYmoj2JisXssQMr/view) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image analysis that bridges the gap between accuracy and versatility in digital pathology through specialized agents and robust verification mechanisms. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 📂 Resources 14 | 15 | 🚀 **Code**: Coming Soon 16 | 📄 **Paper**: Coming Soon 17 | 18 | --- 19 | 20 | ## 🏗️ System Overview 21 | 22 | WSI-Agents employs a collaborative multi-agent approach to address the accuracy-versatility trade-off in WSI analysis. The system integrates specialized expert agents with comprehensive verification mechanisms to ensure clinical accuracy while maintaining multi-task capabilities. 23 | 24 | ![WSI-Agents Workflow](static/image/fig1.png) 25 | 26 | 27 | The architecture consists of three core components: a task allocation module that assigns specialized expert agents, verification mechanisms that ensure accuracy through internal consistency and external knowledge validation, and a summary module that synthesizes final results with visual interpretation. 28 | 29 | ![WSI-Agents Architecture](static/image/wsi-agents.png) 30 | 31 | --- 32 | 33 | ## 🎯 Key Innovations 34 | 35 | - **Multi-Agent Collaboration**: Specialized agents for morphology, diagnosis, treatment planning, and report generation 36 | - **Dual Verification**: Internal consistency checking combined with external knowledge validation 37 | - **Knowledge Integration**: Leverages pathology knowledge bases and WSI foundation models 38 | - **Visual Interpretation**: Comprehensive attention maps for explainable analysis 39 | 40 | --- 41 | 42 | ## 📊 Performance Highlights 43 | 44 | - **WSI-Bench**: Achieves superior performance across morphological analysis, diagnosis, and treatment planning 45 | - **WSI-VQA**: 60.0% accuracy, outperforming existing WSI MLLMs by significant margins 46 | - **Report Generation**: Substantial improvements in clinical report quality and accuracy 47 | - **Consistent Gains**: 10-17% performance improvements across diverse pathology tasks 48 | 49 | --- 50 | -------------------------------------------------------------------------------- /.history/README_20250625214751.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://arxiv.org/abs/placeholder) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image (WSI) analysis that integrates specialized functional agents with robust verification mechanisms to enhance both task-specific accuracy and multi-task versatility. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 🏗️ System Overview 14 | 15 | WSI-Agents consists of three main modules: 16 | - **Task Allocation Module**: Assigns tasks to specialized expert agents 17 | - **Verification Mechanism**: Internal consistency + external knowledge validation 18 | - **Summary Module**: Synthesizes results with visual interpretation maps 19 | 20 | ![WSI-Agents Workflow](static/image/figure1.png) 21 | *Need: Figure 1 - The workflow diagram* 22 | 23 | ![WSI-Agents Architecture](static/image/figure2.png) 24 | *Need: Figure 2 - Detailed system architecture* 25 | 26 | --- 27 | 28 | ## 🎯 Key Features 29 | 30 | - 🎯 **Specialized Expert Agents**: Morphology, diagnosis, treatment planning, and report generation 31 | - ✅ **Dual Verification System**: Internal consistency checking + external knowledge validation 32 | - 🧠 **Knowledge-Driven**: Leverages pathology knowledge bases and WSI foundation models 33 | - 📊 **Superior Performance**: Outperforms existing WSI MLLMs and medical agents by 10-17% 34 | - 🔍 **Visual Interpretation**: Generates comprehensive attention maps for explainable analysis 35 | 36 | --- 37 | 38 | ## 📊 Results Highlights 39 | 40 | - **WSI-Bench**: Achieves 70.3% average performance, surpassing previous methods 41 | - **WSI-VQA**: Reaches 60.0% accuracy, outperforming WSI-LLaVA (55.0%) and others 42 | - **Report Generation**: 44.0% accuracy with significant improvements across all metrics 43 | - **Consistent Improvements**: 10-17% performance gains across diverse pathology tasks 44 | 45 | --- 46 | 47 | ## 📂 Resources 48 | 49 | 🚀 **Code**: Coming Soon 50 | 📄 **Paper**: Coming Soon 51 | 📊 **Dataset**: Coming Soon 52 | 🎯 **Pre-trained Models**: Coming Soon 53 | 54 | --- 55 | 56 | ## 📄 Citation 57 | 58 | ```bibtex 59 | @inproceedings{wsi-agents2025, 60 | title={WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis}, 61 | author={Anonymized Authors}, 62 | booktitle={MICCAI}, 63 | year={2025} 64 | } 65 | ``` 66 | 67 | --- 68 | 69 | ## 📞 Contact 70 | 71 | For questions and collaborations, please contact: email@anonymized.com -------------------------------------------------------------------------------- /.history/README_20250625214747.md: -------------------------------------------------------------------------------- 1 | # WSI-Agents 2 | 3 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 4 | 5 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 6 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://arxiv.org/abs/placeholder) 7 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 8 | 9 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image (WSI) analysis that integrates specialized functional agents with robust verification mechanisms to enhance both task-specific accuracy and multi-task versatility. 10 | 11 | 🎉 **Accepted at MICCAI 2025** 12 | 13 | --- 14 | 15 | ## 🏗️ System Overview 16 | 17 | WSI-Agents consists of three main modules: 18 | - **Task Allocation Module**: Assigns tasks to specialized expert agents 19 | - **Verification Mechanism**: Internal consistency + external knowledge validation 20 | - **Summary Module**: Synthesizes results with visual interpretation maps 21 | 22 | ![WSI-Agents Workflow](static/image/figure1.png) 23 | *Need: Figure 1 - The workflow diagram* 24 | 25 | ![WSI-Agents Architecture](static/image/figure2.png) 26 | *Need: Figure 2 - Detailed system architecture* 27 | 28 | --- 29 | 30 | ## 🎯 Key Features 31 | 32 | - 🎯 **Specialized Expert Agents**: Morphology, diagnosis, treatment planning, and report generation 33 | - ✅ **Dual Verification System**: Internal consistency checking + external knowledge validation 34 | - 🧠 **Knowledge-Driven**: Leverages pathology knowledge bases and WSI foundation models 35 | - 📊 **Superior Performance**: Outperforms existing WSI MLLMs and medical agents by 10-17% 36 | - 🔍 **Visual Interpretation**: Generates comprehensive attention maps for explainable analysis 37 | 38 | --- 39 | 40 | ## 📊 Results Highlights 41 | 42 | - **WSI-Bench**: Achieves 70.3% average performance, surpassing previous methods 43 | - **WSI-VQA**: Reaches 60.0% accuracy, outperforming WSI-LLaVA (55.0%) and others 44 | - **Report Generation**: 44.0% accuracy with significant improvements across all metrics 45 | - **Consistent Improvements**: 10-17% performance gains across diverse pathology tasks 46 | 47 | --- 48 | 49 | ## 📂 Resources 50 | 51 | 🚀 **Code**: Coming Soon 52 | 📄 **Paper**: Coming Soon 53 | 📊 **Dataset**: Coming Soon 54 | 🎯 **Pre-trained Models**: Coming Soon 55 | 56 | --- 57 | 58 | ## 📄 Citation 59 | 60 | ```bibtex 61 | @inproceedings{wsi-agents2025, 62 | title={WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis}, 63 | author={Anonymized Authors}, 64 | booktitle={MICCAI}, 65 | year={2025} 66 | } 67 | ``` 68 | 69 | --- 70 | 71 | ## 📞 Contact 72 | 73 | For questions and collaborations, please contact: email@anonymized.com -------------------------------------------------------------------------------- /.history/README_20250625220230.md: -------------------------------------------------------------------------------- 1 | # 🤖 WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis 2 | 3 | [![MICCAI 2025](https://img.shields.io/badge/MICCAI-2025-blue)](https://miccai2025.org/) 4 | [![Paper](https://img.shields.io/badge/Paper-arXiv-red)](https://arxiv.org/abs/placeholder) 5 | [![License](https://img.shields.io/badge/License-MIT-green)](LICENSE) 6 | 7 | **WSI-Agents** is a novel collaborative multi-agent system for whole slide image analysis that bridges the gap between accuracy and versatility in digital pathology through specialized agents and robust verification mechanisms. 8 | 9 | 🎉 **Accepted at MICCAI 2025** 10 | 11 | --- 12 | 13 | ## 📂 Resources 14 | 15 | 🚀 **Code**: Coming Soon 16 | 📄 **Paper**: Coming Soon 17 | --- 18 | 19 | ## 🏗️ System Overview 20 | 21 | WSI-Agents employs a collaborative multi-agent approach to address the accuracy-versatility trade-off in WSI analysis. The system integrates specialized expert agents with comprehensive verification mechanisms to ensure clinical accuracy while maintaining multi-task capabilities. 22 | 23 | ![WSI-Agents Workflow](static/image/fig1.png) 24 | 25 | The architecture consists of three core components: a task allocation module that assigns specialized expert agents, verification mechanisms that ensure accuracy through internal consistency and external knowledge validation, and a summary module that synthesizes final results with visual interpretation. 26 | 27 | ![WSI-Agents Architecture](static/image/wsi-agents.png) 28 | 29 | --- 30 | 31 | ## 🎯 Key Innovations 32 | 33 | - **Multi-Agent Collaboration**: Specialized agents for morphology, diagnosis, treatment planning, and report generation 34 | - **Dual Verification**: Internal consistency checking combined with external knowledge validation 35 | - **Knowledge Integration**: Leverages pathology knowledge bases and WSI foundation models 36 | - **Visual Interpretation**: Comprehensive attention maps for explainable analysis 37 | 38 | --- 39 | 40 | ## 📊 Performance Highlights 41 | 42 | - **WSI-Bench**: Achieves superior performance across morphological analysis, diagnosis, and treatment planning 43 | - **WSI-VQA**: 60.0% accuracy, outperforming existing WSI MLLMs by significant margins 44 | - **Report Generation**: Substantial improvements in clinical report quality and accuracy 45 | - **Consistent Gains**: 10-17% performance improvements across diverse pathology tasks 46 | 47 | --- 48 | 49 | ## 📄 Citation 50 | 51 | ```bibtex 52 | @inproceedings{wsi-agents2025, 53 | title={WSI-Agents: A Collaborative Multi-Agent System for Multi-Modal Whole Slide Image Analysis}, 54 | author={Anonymized Authors}, 55 | booktitle={MICCAI}, 56 | year={2025} 57 | } 58 | ``` 59 | 60 | --- 61 | 62 | ## 📞 Contact 63 | 64 | For questions and collaborations, please contact: email@anonymized.com --------------------------------------------------------------------------------