About the Project:

In the case of acute ischemic stroke, an accurate and rapid diagnosis is crucial for determining the appropriate treatment. Brain MRI images play a key role in deciding on interventions such as thrombolysis or thrombectomy. However, the time-sensitive nature of stroke treatment often prevents in-depth evaluation, and the low resolution and noise in multimodal MRI images can complicate decision-making.

Building on prior research at the IBISC laboratory, this internship will focus on exploring innovative approaches to multimodal fusion aimed at improving the segmentation of MRI images during acute ischemic stroke. The goal is to integrate Bayesian deep learning to enhance the precision of segmentation and provide reliable uncertainty estimates, making it easier for doctors to make critical treatment decisions under time pressure.

Internship Objectives:

The intern will work on the following:

• Development of a multimodal fusion network: Using advanced deep learning techniques, the intern will design a model that combines different MRI modalities to improve segmentation.
• Incorporation of Bayesian deep learning: Implement approaches to estimate aleatoric (inherent noise in MRI images) and epistemic (model uncertainty due to limited training data) uncertainties, based on the research of A. Kendall and Y. Gal, "What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?"
• Training and validation: Train the model to identify redundant and unique information across multiple modalities, and use uncertainty maps to enhance the fusion process.
• Research and analysis: Explore how understanding and quantifying uncertainty in multimodal fusion can lead to better accuracy and confidence in MRI segmentation for stroke diagnosis.

Key Skills Required:

• Strong background in machine learning and deep learning, particularly in computer vision applications.
• Familiarity with Bayesian neural networks and uncertainty estimation.
• Experience with Python, PyTorch or TensorFlow.
• Knowledge of medical imaging, particularly MRI, is a plus.
• Ability to work independently and as part of a research team.

What We Offer:

• Hands-on experience with cutting-edge AI techniques in the field of medical imaging.
• Exposure to real-world applications of deep learning for critical healthcare problems.
• Mentorship from experienced researchers at IBISC laboratory.
• Potential to contribute to publications or conferences based on your work.

How to Apply:

Please send your CV, cover letter, and any relevant project or research work to vincent.vigneron@univ-evry.fr and congej@yahoo.fr by 30/10/2024 maximum.