IIT Hyderabad develops AI tool RadGLO for non-invasive brain tumour diagnosis

IIT Hyderabad researchers have created RadGLO, an AI-powered platform that analyzes MRI scans to diagnose brain tumours non-invasively. Coupled with RaSPr, it predicts tumour aggressiveness, links MRI features to genes, and enables personalized treatment planning for glioma patients.

Hyderabad: Researchers from Indian Institute of Technology (IIT) Hyderabad have developed an AI-based tool that enables cancer specialists to diagnose and analyze brain tumours without relying on invasive procedures like a surgical biopsy.

The groundbreaking computational platform, named RadGLO, extracts crucial genetic-level insights directly from a normal MRI scan, paving the way for personalized, non-invasive treatment planning.

---

In the study, published in the prestigious Precision Oncology (October, 2025) journal, the IIT Hyderabad researchers said that the AI tool has the potential to revolutionize brain tumour diagnosis, especially gliomas (common primary brain tumours).

At present, oncologists rely on standard brain MRI to locate a tumor. Subsequently, they conduct a surgical biopsy to understand the complexity and aggressiveness of the tumour. While a biopsy is informative, it carries risks because of its invasive nature. The procedure required the surgeon to drill a hole in the skull to collect a small part of the suspected tumour for lab testing.

The IIT-H team leveraged radiomics, an emerging field that treats medical images as high-dimensional data. This involves using advanced computational technology to extract thousands of quantitative features such as the texture, shape, and intensity patterns from the tumor on an MRI scan.

These radiomic features are proving to offer insights comparable to those obtained from full genetic analyses, helping to identify the tumor’s grade (aggressiveness), predict how a patient will respond to treatment, and forecast survival outcomes, researchers said.

As part of the study, the IIT-H researchers have developed RaSPr (Radiomic Survival Predictor) and RadGLO ((Glioma Radiomics Analysis Platform).

The RaSPr uses the radiomic features to sort patients into high-risk (poor survival) and low-risk (better survival) groups, allowing doctors to make informed decisions.

The RadGLO is an interactive web platform that allows doctors and researchers to find radiomic markers that can distinguish between dangerous high-grade and less aggressive low-grade gliomas. Furthermore, RadGLO enables linking each subtle MRI feature to nearly 20,000 genes, helping scientists understand the tumor’s genetic basis for progression, the researchers in the study said.

The platform allows doctors to upload the patient’s MRI scan to receive automated risk predictions and detailed molecular/genetic correlations. The technology empowers the medical community to incorporate non-invasive techniques in cancer prognosis and paving the way for personalized treatment planning for brain tumor patients, research team Kavita Kundal, Dr K Divya Rani, Vinodini D, Dr Neeraj Kumar, and Dr Rahul Kumar, said.