Click the image to visit the challenge homepage

Join Our Pathology Report Generation Challenge!

We are thrilled to announce our lab's Pathology Report Generation Challenge, a groundbreaking initiative to advance automated pathology report generation using cutting-edge vision-language models! Hosted in collaboration with leading medical centers, this challenge leverages a diverse dataset of 20,500 gigapixel-scale whole-slide images (WSIs) from six medical institutions across five countries—Korea, Japan, India, Turkey, and Germany. Our goal is to develop AI models that produce clinically accurate and meaningful pathology reports while addressing racial and ethnic diversity in medical AI.

Why Participate?

• Impactful Research: Contribute to improving diagnostic accuracy and addressing labor shortages in pathology through innovative AI solutions.

• Unique Dataset: Work with a multicultural dataset featuring Pan-Asia and European cases, fostering inclusive medical AI technologies.

• Advanced Evaluation: Test your models against standardized datasets and clinically relevant metrics, moving beyond traditional NLP metrics like BLEU or ROUGE.

• Real-World Application: Explore how automated reports can integrate into diagnostic workflows, guided by expert clinical assessments.

Challenge Highlights

• Training Data: Access ~20,500 cases via the TrainDataset (available now, with additional servers added to ensure smooth downloads).

• Test Phases: Compete in Test Phase 1 (open now, deadline: August 1, 2025) and Test Phase 2 (opens August 2, 2025).

• Flexible Rules: Use pre-trained weights without restrictions, but additional training datasets are prohibited.

• Debug Phase: Validate your submission format before the deadline on July 10, 2025.

• Extended Deadlines: All timelines have been extended by two weeks due to recent server maintenance (check the Important Dates section).

Important Dates (All times 10:00 AM KST)

• Training Data Release: May 20, 2025 ✅

• Registration Deadline: July 4, 2025 ✅

• Debug Phase: July 4, 2025 – July 10, 2025 ✅

• Test Phase 1: July 11, 2025 – August 1, 2025 ✅

• Test Phase 2: August 2, 2025 – August 23, 2025

• Winner Announcement: September 9, 2025

Key Information

• Data Access: Training and Test1 datasets are available for download. Check the Reg2025-TrainDataset and Reg2025-TestDataset pages for details.

• Server Updates: All servers are fully restored post-maintenance. Additional servers have been added to handle high traffic. Note: A brief FTP downtime is scheduled for May 26, 2025 (9:00–10:00 AM KST).

• Data Fixes: Some TIFF files had issues (color distortion, corruption). Corrected files are available in /REG_train_revised (servers 104–107) and REG_test1_revised. Check corrupted_id.csv and revised_list.csv for details. Local fixes: tiffset -s 262 2 FILENAME.tiff (corrupted files) or tiffset -s 262 6 PIT_file_name.tiff (color issues).

• Evaluation: Use our Evaluation Code on GitHub to ensure your submissions meet the required format.

• Submissions: Follow the guidelines at reg2025.grand-challenge.org.

Get Involved!

This is your chance to push the boundaries of medical AI and make a meaningful impact on global healthcare. Register now at reg2025.grand-challenge.org and join us in shaping the future of pathology!

For any questions, contact our team via the challenge website. We look forward to seeing your innovative solutions!