Understanding nephrotic syndrome using deep learning
Nephrotic syndrome is a kidney disorder caused by damage to the kidney filtering units, named glomeruli. This condition is associated with many health complications
that can result in total kidney failure and the need for dialysis. As of today, the causes and mechanisms of this syndrome are not fully understood, which complicates
the development of better treatments.
In collaboration with the ForMe registry project (ForMe Register), we are developing deep-learning methods to automatically segment and classify glomeruli in microscopic images of kidney tissue. In this context, we have implemented a new architecture for glomeruli instance segmentation (GlomNet). We are now developing a method to classify glomerular damage using self-supervised and interpretability methods to identify tissue morphology and cell populations predictive of damage.
Further analysis of cellular morphologies and organization could reveal new features predictive of kidney damage, potentially leading to a better understanding of kidney disorders and the development of new therapies.
Date:
05/09/2024
