To realize the grand challenges of precision medicine in the next decade we will have to lift medical imaging to a new era of physics-driven functional deep imaging. By including the physics and the related mathematical inverse problems we utilize all data and we circumvent the present black box approach, where data is number-crunched with maybe the incorrect biological and physiological information. Thus, the proposed approach will elevate deep learning for medical imaging to a higher level, with increased confidence for clinical decision support.
Developing this unique continuous pipeline from physics to deep learning and back, will also unravel unexplored terrains in fundamental research in physiology, biology and physics. For example, we have first evidence for unexplored tumor cell types via back-propagation from deep learning to cell biophysics. Similarly, we expect fundamental observations for unchartered nonlinear flow phenomena and in quantitative imaging in photoacoustics. This back-propagation from data to fundamental research is the key novelty of our consortium.