Radiomic analysis of DCE breast MRI data sets for improved diagnosis of breast cancer - a multi-institutional evaluation

• Applicant: Professor Dr. Daniel Truhn
• Subject Area: Radiology ; Medical Informatics and Medical Bioinformatics
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 515639690

Breast cancer remains one of the most important causes of cancer death. There is therefore a great need for improved methods of early detection. With our study we want to investigate to what extent machine learning methods based on analyses of dynamic contrast enhanced (DCE) breast MRI data sets are suitable to support the clinical diagnosis of breast cancer. DCE-MRT of the breast offers the highest performance in detecting breast cancer of all available diagnostic procedures. One limitation of the method is the low specificity/PPV, especially when the examinations are evaluated by non-experts. We will develop an "Augmented Intelligence Expert Level Diagnostic System" (ELDS). For this purpose we will use a large database of existing multiparametric breast MRIs to refine our algorithms for the detection and classification of enhancing lesions. The ELDS is intended to enable an autonomous diagnostic evaluation of breast MRIs (independent detection of lesions and evaluation with respect to malignancy). We will initially establish the ELDS for use in regular, multiparametric, breast MRI data sets, then for "abbreviated MRI" and finally evaluate it in a prospective study with 500 patients in our own clinic. Since a major problem of machine learning procedures is that they work well on data sets on which they have been trained, but less well, on other data sets, we will subsequently test, improve and evaluate the ELDS in a multi-institutional study in cooperation with the university hospitals of Cologne, Düsseldorf and Bonn.

• DFG Programme: Research Grants