Interview with Dimitrios Karampinos

Interview, Top |

Prof. Dimitrios Karampinos, principal investigator of the Experimental Magnetic Resonance Imaging professorship, describes his projects related to novel MRI technologies, experiences in data science and his collaboration with other MDSI members.

picture of dimitrios karampinos

What cutting-edge project are you working on right now?

Our research group develops novel techniques for Magnetic Resonance Imaging (MRI). An example is the development of a novel MRI methodology to depict small calcifications in the breast.  Such small calcifications (microcalcifications) are considered a precursor of cancer and are a primary diagnostic target when using X-ray-based mammography. We aim to depict these microcalcifications with MRI which is non-invasive and would decrease the radiation exposure in young high-risk-for-breast cancer patients receiving annual screening.

What was the key experience that made you want to do research in data science and what fascinates you about working with MDSI?

We are particularly interested to encode additional dimensions beyond space during an MRI experiment, that would enable us to quantify tissue properties to improve diagnostics and therapy response prediction. However, MRI measurements are typically associated with long acquisition times and the encoding of such additional dimensions makes the MRI examination even slower. The incorporation of machine learning approaches within the image formation process opens new ways to tackle the critical question on how to accelerate an MRI examination. MDSI offers the perfect platform to build collaborations with data science experts, introducing machine learning not only in the processing of the MR images but also within the MR image formation process. It is exactly at the intersection of imaging physics principles and state-of-the-art machine learning algorithms where the breakthrough can occur.

What paradigm shift do you expect within the next ten years, triggered in particular by the Institute's interdisciplinary approach?

Our ongoing collaborations with the Chairs of Daniel Rückert and Julia Schnabel provide the current foundation to introduce machine learning algorithms within the MR image formation process, eventually affecting the MRI measurement itself. Our long-term goal would be to build the tools that would replace the currently patient-agnostic, motion-sensitive qualitative, segmented MR exams into a patient-specific, motion-robust, quantitative single-scan MR exam.