Amélioration d’image, super résolution et classification d'images par l'IA : Application à la microscopie holographique numérique

Digital holographic microscopy (DHM) has emerged as a powerful and transformative imaging technique, offering label-free, non-invasive visualisation of cellular structures and dynamic processes with exceptional sensitivity. Such a technology offers great promise for neuronal studies because it is label-free and quantitative. While DHM provides highly sensitive quantitative phase images and is therefore widely developed worldwide, coherent noise remains a challenge, limiting its potential as a robust diagnostic tool. Polychromatic digital holographic microscopy (P-DHM), a cutting-edge approach to quantitative phase imaging, has recently been shown to be highly efficient for label-free imaging of live cells. It provides quasi-coherent-noise-free optical path difference (OPD) images that serve as accurate reference images, free from the effects of coherent noise. However, such polychromatic microscopes remain complex in design and expensive.
To address these limitations and to unlock the full diagnostic capabilities of DHM, particularly in the detection of mental diseases, this PhD project aims to explore the integration of Artificial Intelligence (AI) into DHM, with a focus on using Polychromatic Digital Holographic Microscopy (P-DHM) as a reference image to enhance DHM images. The AI algorithms will be carefully designed to denoise images, reduce coherent noise and improve spatial resolution using super-resolution methods on neural cell images acquired with different microscope optic objectives and magnification. AI trained on duos of experimental P-DHM/ DHM images will act as a powerful tool to overcome standard limitations , enabling clearer visualisation of minute cellular structures, including neuronal processes, which are essential for understanding neural circuits and cellular dynamics.