Objective
Human fetal development is still largely described from sections: tissue cut, stained and read slice by slice, with the three-dimensional architecture reconstructed in the reader’s head. DREAMLAND images whole fetal organs in three dimensions without cutting them, and does it with two modalities that see different things — synchrotron phase-contrast tomography for cellular-scale architecture, MRI for anatomy and tissue contrast over the whole organ.
A first imaging campaign covered about a dozen individuals and some 180 organs; new samples arrive at a steady rate through CHU Grenoble Alpes. At that volume, the bottleneck is no longer the acquisition but the analysis.
Approach
Samples are imaged by synchrotron phase-contrast tomography and by MRI, and the two modalities are registered so that the microscopic architecture can be read against the macroscopic anatomy. A postdoctoral position, co-funded between 3T at IAB and the GIN, develops the deep-learning segmentation pipeline; it is written to be reused across the whole consortium rather than for one study.