We are thrilled to announce a major milestone for BCN-AIM! Our lab member, Lidia Garrucho Moras, has published in Scientific Data about the MAMA-MIA Breast Cancer DCE-MRI dataset, now publicly available. This dataset marks a significant step forward for AI research in breast cancer imaging.
Addressing a Critical Gap in Breast Cancer AI
Developing reliable AI models for breast cancer MRI has been challenging due to the scarcity of large, expert-annotated datasets. MAMA-MIA changes that by providing the largest publicly available breast DCE-MRI dataset with expert-validated 3D tumor segmentations. This enables groundbreaking AI research for cancer diagnosis and treatment.
Key Features of MAMA-MIA
1. Automatic Tumor Segmentation
- Includes 1,506 expert-validated 3D segmentations
- Supports the development of generalizable, robust AI models
- Pre-trained nnU-Net weights available as a baseline
2. Treatment Response & Survival Prediction
- Provides clinical and treatment outcomes for 1,506 cases
- Enables AI models to predict response to neoadjuvant chemotherapy (NAC)
3. Segmentation Quality Control
- Facilitates research into AI-driven quality control for clinical applications
4. Image Synthesis & Standardization
- Diverse dataset from multiple scanners, protocols, and institutions
- Supports image standardization, domain adaptation, and bias mitigation
5. Fine-Tuning Foundational AI Models
- Enables the adaptation of multimodal foundation models for breast MRI tasks
Usefull links:
📖 Read the full paper: here.
⬇️ Download the dataset: here.
We are excited to see how the research community will leverage MAMA-MIA to advance AI in breast cancer imaging!
