IIT Madras and CMC Vellore researchers built AI tools to predict chronic kidney disease risk and sort kidney CT scans into four types.

Three tools

The first is a machine learning model that uses clinical and lab information to predict a person's risk of chronic kidney disease. The team built it into a simple prototype screen and worked to make its results clearer for doctors.

The second is a deep learning system that sorts kidney CT scans into four groups: normal kidney, kidney cyst, kidney stone and kidney tumour.

The third rebuilds the kidney in 3D from CT scans. It measures tumour volume and how much of the kidney the tumour affects. It uses open-source software, which the team describes as a low-cost, repeatable way to measure tumours.

Who did the work

Prof G L Samuel of the Department of Mechanical Engineering at IIT Madras and research scholar Jennifer Delighta led the research. Prof Santosh Varughese of the Department of Nephrology at CMC Vellore worked with them.

The project was funded by IIT Madras and the Scheme for Promotion of Academic and Research Collaboration (SPARC). Delighta said, "Early detection is of paramount importance when dealing with kidney diseases."

What happens next

The team plans to test the models on more patient data and to partner with hospitals for real-world use. It will also look at linking the tools with wearable sensors and digital twin platforms, which use a virtual copy of a patient's organ to track health and plan treatment.

IIT Madras did not publish accuracy figures, the model designs or the journal where the work appeared.