Sites Inria

Version française

OPIS Research team

OPtImization for large Scale biomedical data

Team presentation

The objective of the OPIS project is to design advanced optimization methods for the analysis and processing of large and complex data. Applications to inverse problems and machine learning tasks involving high-dimensional biomedical data, e.g. 3D CT, PET, ultrasound images, and MRI are targeted in this research project. The focus is put on optimization methods able to tackle data with both a large sample-size (“big N”) and/or many measurements (“big P”). The explored methodologies are grounded on nonsmooth functional analysis, fixed point theory, parallel/distributed strategies, and neural networks. The new optimization tools that are developed are set in the general framework of graph signal processing, encompassing both regular graphs (e.g., images) and non-regular graphs (e.g., gene regulatory networks).

Research themes

Three main research avenues are explored.

  1. Proposing novel algorithms able to encompass high-dimensional continuous optimization problems. Attention is paid to develop methods with established convergence guarantees, and that are well-suited to parallel implementation. Because of the versatility of the proposed approaches, a wide range of applications in image recovery can be considered. These include parallel MRI, breast tomosynthesis, coronary disease assessement, and two-photon microscopy

  2. Designing efficient optimization approaches for the resolution of graph signal processing and graph mining problems. In terms of applications, novel graph mining and learning methods are proposed for biological network analysis.

  3. Developing a new generation of deep learning strategies, characterized by robustness guarantees, faster training, and suitable account for prior information.  Proposing novel neuronal models is particularly interesting in settings such as the diagnosis or prevention of diseases from medical images, because they correspond to critical applications where correct decisions are frequently life saving and need to be interpretable. The use of deep learning techniques in such contexts is particularly challenging as there are lots of limitations that have to be addressed, e.g., limited data availability, lack of consistent textural or salient patterns, and the high dimensionality of medical data. Optimal ways are investigated for allowing deep learning architectures to successfully address these challenges.

International and industrial relations

A wide array of academic collaborations exist both at the national and international levels, in particular with

  • North Carolina State University
  • Sup'Com Tunis
  • IIIT Delhi
  • University Paris-Est
  • Institut Gustave Roussy

Industrial partnerships are also developed, in particular with

  • GE Heathcare
  • IFP Energies Nouvelles

Keywords: Optimisation Biomedical imaging Inverse problems Machine learning Neural networks Graphs