Postdoctoral fellow in Biophysical guided AI for (medical) imaging (M/F)

Over the past decade, deep learning-based image methods have emerged as a prominent tool in medical image processing. While they have shown impressive success in various computer vision tasks, their application in the medical imaging field requires additional controls and considerations due to the need for heightened control and the limitations associated with reduced data availability, especially in 3D imaging.
The goal of this postdoc is to focus on new AI approaches, that while exploiting deep auto-differentiation frameworks will also incorporate explicit physics-inspired additional modules to account for physical laws, constraints, or knowledge about the modality into the network structure or training process.

The postdoctoral fellow will notably investigate the following concepts:
- Deep unrolling (“learning to optimize”) iteratively applies a deep neural network to simulate the steps of optimization algorithms, allowing for an approximate solution to complex problems in e.g. reconstruction, deblurring, super-resolution, segmentation or quantification.
- Deep image prior methods that utilizes the structure and priors learned by a deep neural network to generate high-quality image reconstructions from incomplete or corrupted input data. The network's architecture itself serves as a prior, enabling the generation of plausible images without the need for extensive training data. Modality-specific considerations (CT, MRI, US, nuclear) as well as biophysical/biochemical modeling of biological tissues will be incorporated into the network design to improve performance and reliability.
- Physics-informed neural networks compute the derivative of the estimated outputs of a network to compute additional losses that corresponds to physical laws or priors. In this way, the network is able to optimize (or learn) a solution that must respect the underlying physics.

This methodological research will be applied to one of the laboratory's current cross-disciplinary projects:
- Multiple sclerosis and neuro-inflammation: from preclinical to clinical investigations (MUSIC)
- Radiomic for tumor characterization and treatment response (TUMOR-ID)
- Tissue Optical Imaging (TipTop)
- Multiparametric Multimodality Imaging of Musculo-skeletal & Myocardial muscles Damage (IDM4)
- Functional Imaging and Modelling of the Lungs (FILM)

The research activities of the CREATIS laboratory are within the field of health technologies and aim to contribute to predictive and personalized medicine through imaging. Interdisciplinary research brings together experts in image processing and analysis, computer science, physics, instrumentation and radiology. Ischemic heart disease, multiple sclerosis, cancers, stroke are among the pathologies addressed at CREATIS. The post-doctoral student will be involved in one of the laboratory's cross-team projects with plans to apply for a junior research position in the laboratory.