Machine learning for ultrafast vascular and fonctionnal ultrasound imaging of the brain
Ultrafast ultrasound cerebral imaging is a technique recently introduced by the Physique for Medicine Paris laboratory to visualize and quantify blood flow in the brain with very high sensitivity.
The technique allows the detection of blood flow (vascular imaging mode called ultrasensitive Doppler, uDoppler), and their variations related to neuronal activity (ultrasound functional imaging mode, fUS).
This innovative technology makes it possible to monitor the hemodynamics of the brain in rodents for various applications such as: characterization of cerebrovascular pathologies, or behavioral studies of the awake animal, in motion, at rest or in different phases of sleep.
Initially implemented for 2D imaging, ultrafast ultrasound cerebral imaging is now being developed for 3D imaging (whole brain).
Each recording of hemodynamics over time and in the whole brain is extremely rich in information, and constitutes a very large amount of data to be analyzed. Thus powerful algorithms are required. At present, data are manually processed and there is no automated tool to analyze and compare data robustly. The main objective of this PhD is to develop powerful tools for automated data processing, based on machine learning algorithms.
During this PhD, the student will develop machine learning algorithms for:
- Improving the quality of vascular images of the brain
- Recalibrating 3D images spatially to compare data from different imaging sessions
- Identifying cerebrovascular abnormalities or different functional states
Such analytical tools will greatly enhance the potential of ultrafast ultrasound imaging of the brain by extracting relevant information for medical diagnosis, monitoring of therapies as well as identifying functional cerebral states in neuroscience.
- Physics for Medicine Paris – CNRS : National science research center and ESPCI and Inserm and Langevin Institute