MICCAI 2025 Lighthouse Challenge: Brain Tumor Segmentation Cluster of Challenges (BraTS)

Authors are listed alphabetically.

The Brain Tumor Segmentation (BraTS) Cluster of Challenges 2025 is a collaborative effort with the "AI for Response Assessment in Neuro-Oncology" (AI-RANO) cooperative group and leading clinical societies, including RSNA, ASNR, and ESNR. This partnership aims to establish and promote clinically relevant challenges to maximize the potential clinical impact of innovative algorithmic contributions from participating teams.

Since its inception at MICCAI 2012, BraTS has advanced brain glioma segmentation by enabling and benchmarking algorithmic developments, providing high-quality annotated datasets. In 2023, BraTS expanded to include multiple benchmarks, quantifying tumors beyond glioma (brain metastases, meningiomas, pediatric brain tumors, and sub-Saharan patient populations), histology images, and image synthesis. This novel, innovative design creates a benchmarking ecosystem for the systematic comparison of algorithms across diverse tasks and clinical challenges.

The significance of BraTS 2025 lies in its focus on addressing actual clinical needs across brain tumors, spanning i) tumor entities (for which there is currently limited publicly available annotated data), ii) disease course (pre- and post-treatment, for reliable longitudinal assessment of tumor response), iii) domains (radiology & histopathology), & iv) computational tasks (e.g., segmentation, synthesis). Corroborating BraTS's goal to address clinical needs, authoritative and leading federal and clinical organizations have partnered with BraTS (such organizations include NIH, FDA, RANO, RSNA, ASNR, ASFNR, and CBTN).

In 2025, the BraTS Cluster of Challenges will encompass 12 tasks, designed to benchmark and advance the current state-of-the-art for addressing (Task 1) Pre- and Post-Treatment Adult Glioma, (Task 2) Pre-Treatment and (Task 3) Pre-RT intracranial Meningioma, (Task 4) Pre- and Post-Treatment Brain Metastases, (Task 5) Brain Glioma
in the underserved sub-Saharan African patient population, (Task 6) Pre-Treatment Pediatric Tumor Patients in partnership with multiple related societies, (Task 7) Generalizability of Segmentation Methods Across Tumors, (Task 8) Evaluation of Augmentation Techniques, in partnership with FDA, (Task 9) MRI Synthesis, (Task 10) MRI Inpainting, (Task 11) Assessing the Heterogeneous Histologic Landscape of Glioma, as well as (Task 12) Predicting the Tumor Response During Therapy. Detailed descriptions of each task are provided in the following sections.

BraTS 2025 participants can obtain the training and validation data of the challenge at any point from the Synapse platform. These datasets will be used to develop, containerize, and finally evaluate their algorithms in unseen validation data until August 2025, when the organizers will stop accepting new submissions and evaluate the
submitted algorithms in the hidden testing data. To ensure accurate performance evaluation, expert neuroradiologists, and neuropathologists create and approve ground truth annotations for all datasets for each subject in the training, validation, and testing datasets. Notably, in 2025, multiple raters will independently annotate a subset of test cases for several Tasks to compare algorithmic performance against human expert inter-rater variability.

The final version of this document including further revisions will be released soon.