Awesome-Radiology-Report-Generation serves as a comprehensive, curated collection of research papers, datasets, tools, and evaluation metrics focused on automated radiology report generation. It aims to provide a centralized resource for researchers and practitioners in the field, facilitating quick access to state-of-the-art advancements and foundational resources for developing and evaluating AI models in medical imaging interpretation.

How It Works

This project functions as a comprehensive, curated bibliography and resource hub for the field of radiology report generation. It systematically organizes a vast collection of academic papers, datasets, tools, and evaluation metrics, categorizing them by publication year, conference, and specific sub-domains such as foundation models and survey papers. This structured approach allows researchers and practitioners to efficiently discover and access relevant state-of-the-art advancements and foundational resources.

Quick Start & Requirements

This repository is a curated list of research resources and does not contain runnable code or a deployable application. Therefore, there are no installation instructions, prerequisites, or setup requirements. Users are directed to the individual papers and code repositories linked within the list for specific implementation details.

Highlighted Details

• Extensive coverage of foundation models for medical AI, including vision-language models and large language models adapted for radiology tasks.
• A comprehensive catalog of datasets, ranging from large-scale chest X-ray collections (e.g., MIMIC-CXR, CheXpert Plus) to specialized datasets for pathology and CT imaging.
• Detailed listing of evaluation metrics and benchmarks designed to assess the quality, factuality, and clinical utility of generated radiology reports.
• Inclusion of relevant tools and libraries, such as CXR-Report-Metric, RadGraph, and torchxrayvision, aiding in the development and assessment pipeline.

Maintenance & Community

The repository is maintained by mk-runner, with contact information provided via email (kangliu422@gmail.com) and WeChat (kangliu422) for contributions or inquiries. The collection appears to be actively updated, as evidenced by the inclusion of papers from 2025.

Licensing & Compatibility

As this repository is a curated list of external resources, it does not specify a singular license for the collection itself. Users must refer to the individual licenses of the linked papers, code repositories, and datasets for their respective terms of use and compatibility.

Limitations & Caveats

This repository serves as a reference list and does not provide direct access to runnable code or deployable models. Users must navigate to the linked external resources for implementation details. The rapid pace of research in this domain necessitates continuous updates to maintain comprehensiveness.