Framework boosts breast cancer detection from breast MRI images
Feature Space Guidance for Breast Cancer Classification in DCE-MRI
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Breast MRI scans can be tricky to analyze automatically because the images are very detailed and have lots of different settings depending on where they come from. The authors created a system that adjusts itself to these differences by focusing on important parts of the images and the timing of how contrast appears. They trained their system on many examples to help it better recognize cancerous areas, and their approach works better than previous methods on tests with data from multiple hospitals. Their system also won a recent challenge for breast MRI analysis, showing it can be helpful in real clinical settings.
What this means in practice
- •For hospital imaging teams: Automate breast cancer detection from varied MRI machines by adapting to different scan protocols for better diagnosis support.
- •For medical device developers: Integrate this adaptive feature analysis framework to improve software tools that interpret dynamic breast MRI scans across multiple clinical sites.$Commercial implications: This enables selling advanced diagnostic imaging software that improves cancer detection accuracy in breast MRI across diverse clinical protocols.
Authors
Benjamin Hamm, Yannick Kirchhoff, Maximilian Rokuss, Moritz Langenberg, Constantin Ulrich, Tassilo Wald, Jeremias Traub, Karol Gotkowski, Klaus Maier-Hein
Abstract
Dynamic contrast enhanced breast MRI (DCE-MRI) is a powerful clinical tool for breast cancer detection, providing high resolution anatomical detail together with rich temporal contrast information. However, high dimensional 4D inputs, small lesions, and heterogeneous acquisition protocols across clinical sites hinder robust automated classification of healthy, benign, and malignant cases. To address these challenges, we propose a framework that dynamically analyzes latent representations to adapt to protocol-specific characteristics. Spatial variability is mitigated by reducing confounding background uptake and compensating for misalignment caused by deformable soft tissue. Additionally, relationships in the latent space across phases are leveraged to select the most informative temporal features, improving robustness to protocol-specific temporal variability. Finally, task specific discriminative features are promoted through large scale supervised lesion segmentation pretraining, which substantially enhances downstream finetuning. Evaluated under leave-one-center-out validation on the ODELIA dataset and the held-out AMBL cohort, the proposed framework substantially outperforms finetuned radiology foundation models and prior methods, improving mean AUROC by nearly 8 points and balanced accuracy by 4 points over the strongest baseline. Additionally, our method achieved first place in the MICCAI ODELIA Breast MRI Challenge 2025, further demonstrating its effectiveness for robust breast cancer classification. We publicly release our codebase under https://github.com/MIC-DKFZ/CURIAtor.