Development of a breast imaging matching tool

The aim of this D-A-CH proposal is to establish a novel method for matching magnetic resonance imaging (MRI) and close-up stereotactic biopsy X-ray mammograms based on accurate simulation of the biomechanics of the female breast, which could improve the cost- effectiveness of the work-up of MRI-detected breast lesions up to 50%. We aim to transfer information of lesions only visible in MRI onto mammographic close-up projection views with sufficient registration accuracy to fit a mammogram-based lesion workup and thereby reduce the need for costly and time-consuming MRI biopsies. The challenges of such a registration are to model the substantial non-linear deformations of the breast undergoing mammographic compression, to cope with patient-specific variabilities of the compression process due to individual positioning and to match the MRI information to close-up mammography views performed during stereotactical biopsy, which do not cover the complete breast. For accurate prediction of a lesion position in these views in a clinical setting, it is essential to identify the required complexity of a biomechanical model overcoming the non-linear deformation while at the same time serving clinical needs in both registration accuracy and computation time. Moreover, it is important to identify and consider factors influencing matching accuracy like anatomical, clinical or biomechanical features. In this proposed project we will address these research questions by means of patient specific biomechanical models to simulate the deformation of the breast undergoing mammographic breast compression. The project will investigate the clinical feasibility of X- ray guided biopsy, which is planned based on MRI findings only by developing the necessary software and workflow while considering the clinical relevance of matching accuracies and workflow speed. The project will provide a final method which can directly be implemented into clinical practice. Our research team at KIT and MUW provides combined experience with respect to technical aspects of breast image registration and clinical needs of diagnostic breast imaging and intervention. The research in this proposal will provide new insights in the clinical application of biomechanically driven breast image registration, enable a cost reduction of up to 50%, address the shortage of facilities performing MRI-guided interventions and thus considerably facilitate and fasten work-up of patients with MRI-detected breast lesions.

This D-A-CH lead agency project involved teams from Austria and Germany and focused on developing a method to localize MRI-only breast lesions on mammographic biopsy images. The overall goal was to enable these lesions-typically requiring MRI-guided biopsy-to be targeted using more accessible X-ray-guided stereotactic systems. The project was initiated by the Austrian partner at the Medical University of Vienna, where the clinical need was identified. The team contributed a curated dataset of 57 patient cases, annotated by experienced breast radiologists. These cases included high-quality mammograms, contrast-enhanced MRIs, and stereotactic biopsy images, allowing for comprehensive multimodal analysis. The German partner at the Karlsruhe Institute of Technology developed two registration algorithms: a biomechanical model-based method for aligning 3D MRI data with full-field mammography, and an image-based approach to further map this onto stereotactic biopsy views. The combined method was evaluated retrospectively. A median target registration error of 35.6 mm was achieved. Simulation showed that in 11 to 14 out of 51 cases, a standard X-ray-guided biopsy would have successfully sampled the lesion based on the predicted location. Results were presented at international conferences and led to multiple joint publications. While further refinement is needed for clinical application, the project demonstrated the feasibility of the approach and provided a structured dataset and evaluation framework for future research. This work forms the basis for follow-up studies focused on improving accuracy and testing the method in a prospective clinical setting.