From Nephron to Digital Twin: AI and Physics for Multiscale Kidney Modelling
Faculty: Prof. Gaurav Tomar (ME) and Prof. Aditya Konduri (CDS)
The kidney is an extraordinarily complex transport system. Within a microscopic structure called the nephron, blood is filtered, water and ions are selectively transported, concentration gradients are generated and maintained, and the final composition of urine is regulated. What makes the problem particularly fascinating from a mathematical and engineering perspective is that these processes occur across many length and time scales. At the glomerulus, blood flows through a complex branching capillary network and filtration takes place across a multilayered porous structure. Further downstream, the Loop of Henle functions as a sophisticated counter-current transport system, where water, ions, osmotic pressure and membrane permeability interact to generate concentration gradients. The surrounding interstitial region and blood vessels introduce additional coupled processes involving the transport of water, ions and oxygen. Each of these problems connects naturally to ideas from fluid mechanics, porous media, mass transport, membrane transport, nonlinear dynamics and numerical modelling. The interesting challenge is to understand how these individual physical processes are coupled and how their behaviour changes with the geometry and properties of the underlying biological system.
A particularly ambitious question is whether we can build a physics-based digital twin of the kidney, starting from the behaviour of an individual nephron and ultimately connecting it to the collective behaviour of 2-10 million nephrons. How does variability in nephron geometry affect kidney function? How do processes occurring at the scale of a few microns influence the behaviour of the whole organ? How can patient-specific information be incorporated into such a model? Addressing these questions opens up problems involving computational fluid dynamics, porous-media transport, network models, nonlinear differential equations, reduced order modelling, high-performance computing and machine learning. The attraction of the problem, in many ways, is that it allows us to look at the kidney not just as a biological organ, but as a remarkable physical system and ask the fundamental engineering questions: What are the governing equations? What are the relevant scales? What controls the transport? And how does microscopic behaviour give rise to the functioning of an entire organ?