In-silico drug delivery in the human brain

Reduced order modeling, combined with deep learning techniques, can significantly enhance the computational models for fluid flow and solute transport models in the human brain, allowing for fast personalized predictions of drug delivery.

Fluid flow and solute transport in the brain are critical components in brain clearance mechanisms and neurological diseases treatments. Despite extensive research, the dosing of intrathecal drugs is still not tailored at the individual level. Data-driven biomedical computational models can be employed to perform personalized predictions of drug concentration, but such models remain computationally demanding due to the complexity and multi-scale aspect of the brain. Reduced Order Modeling (ROM) techniques are designed to handle such parametrized models, providing fast simulations. The combination of such methods with Deep Learning strategies (DL-ROM) in recent years has shown good performances and accuracy. We aim to apply DL-ROM techniques on fluid flow and solute transport models in the human brain, allowing for fast personalized simulations and advanced uncertainty quantification.

Goal

The project aims to enhance the performances of our current models predicting fluid flow and drug delivery in the human brain, using reduced order modeling and deep learning.

Learning outcome

• Knowledge of reduced order modeling and deep learning.
• Experience with finite element software, such as FEniCSx.

Qualifications

• A background in applied mathematics and/or computer science.
• Knowledge of partial differential equations and numerical methods.
• Experience with Python programming.
• Experience with ML frameworks is a plus.

Supervisors

• Cécile Daversin-Catty
• Henrik Nicolay Finsberg
• Miroslav Kuchta

Collaborations

• The master's thesis is associated with the K.G Jebsen Centre for Brain Fluid Research