Hyderabad researchers digitise brain tumour biopsies to train AI for cancer prediction

This groundbreaking initiative has resulted in the development of the first datasets of digital brain tumour biopsies, which provide crucial training data for machine learning AI models, enabling them to predict cancer risks among patients

Hyderabad: Can Artificial Intelligence and Machine Learning (AI/ML) based models predict brain tumours by analysing digital slides? In the fast emerging field of AI/ML technology, such breakthroughs are increasingly feasible.

In a first in Asia, researchers and pathologists from the International Institute of Information Technology (IIIT)-Hyderabad, and Nizam’s Institute of Medical Sciences (NIMS), have spearheaded the digitisation of brain tumour patient biopsies. This groundbreaking initiative has resulted in the development of the first datasets of digital brain tumour biopsies, which provide crucial training data for machine learning AI models, enabling them to predict cancer risks among patients.

“This marks a significant milestone in brain tumour research,” the researchers stated in a paper published in the prestigious peer-reviewed journal ‘Nature’ (December 2024). “Machine learning models trained on this dataset can not only explore regional and ethnic disease variations but also significantly enhance diagnostic precision by identifying cancer subtypes.”

The study titled ‘India Pathology Dataset (IPD)’ project comprised digitised biopsies of 547 high-resolution slides samples from 367 patients making it one of the largest in Asia.

In addition to brain cancers, the NIMS and IIIT researchers are also collaborating to compile a digital dataset of lupus, the kidney ailment, which is highly prevalent in Telangana, especially among women.

The AI-based algorithms and systems can analyse such datasets that are related to lupus at a rapid pace and guide caregivers to take appropriate measures, especially in situations where there is a severe shortage of speciality nephrologists who can make a quick diagnosis.