bmri-ai-prep: a preprocessing pipeline for breast MRI data for AI applications

Abstract Magnetic Resonance Imaging (MRI) is a versatile imaging technique extensively used in clinical and research environments for its ability to capture detailed anatomical and functional information across multiple sequences and contrasts. Despite its strengths, MRI’s complex data structure and variability pose significant challenges to deep-learning applications, which are predominantly optimized for natural images. In particular, existing preprocessing tools are tailored mainly for brain MRI, limiting their effectiveness for other anatomical regions, such as the breast. To address the unique challenges of preprocessing breast MRI data, we introduce bmri-ai-prep, a streamlined and modular preprocessing pipeline specifically designed to enhance the integration of breast MRI into AI and deep-learning workflows. The bmri-ai-prep pipeline emphasizes essential preprocessing steps, including precise cross-modal registration and anatomy isolation, tailored to breast MRI’s unique imaging characteristics. The pipeline assumes data formatted in the proposed Breast Imaging Data Structure (BRIDS), an adaptation of the Brain Imaging Data Structure (BIDS), to standardize and streamline data handling across diverse imaging protocols. Default preprocessing parameters were carefully optimized for breast MRI, balancing robustness and computational efficiency. The bmri-ai-prep pipeline offers a streamlined approach to preprocessing breast MRI data, reducing some of the complexity and time associated with adapting general-purpose neuroimaging pipelines or writing custom data parsers. Its registration and segmentation procedures are designed to improve data consistency and facilitate comparisons between sequences with different contrasts, supporting the training of AI models on the data. The modular structure allows for customization and future expansion, making it adaptable to evolving research requirements. By focusing specifically on breast MRI, bmri-ai-prep addresses a gap in available preprocessing tools. While the extent of its impact will depend on the context of application and further validation, the pipeline has the potential to enhance the efficiency of AI-driven research and contribute to the development of more reliable deep-learning models in breast MRI.