IIIT-H, NIMS launch India’s first pathology dataset

Hyderabad:IIIT Hyderabad (IIIT-H) and Nizam’s Institute of Medical Sciences (NIMS) have launched India’s first publicly available digital pathology dataset, a move poised to improve disease diagnosis and treatment in the country.

This initiative, known as the India Pathology Dataset (IPD), marks a significant step towards harnessing AI for improved clinical outcomes in brain cancer and kidney disease (lupus nephritis).

As part of the IPD project, supported by the Technological Innovation Hub for Data Banks, Data Services, and Data Analytics (TiH-Data), IIIT-H has installed a whole slide digital scanner at NIMS. This scanner digitises tissue biopsy slides, enabling collaborative diagnosis and reducing the risk of slide damage.

“Traditionally, tissue samples are examined under a microscope, but digitisation allows for enhanced visualisation and remote access for pathologists,” explained Prof. P.K. Vinod, who leads the dataset curation.

One of the first datasets released, IPD-Brain, is a comprehensive collection of 547 high-resolution slides from 367 brain cancer patients, focusing on Indian demographics. Published in the prestigious journal Nature Scientific Data, this dataset is pivotal for developing AI models that can enhance diagnostic precision and explore regional disease variations.

Dr Megha Uppin of NIMS said, “AI can help diagnose molecular abnormalities in brain tumours, bridging the gap in neuropathology expertise across India.”

In addition to cancer, the project includes a dataset on lupus nephritis, an autoimmune kidney disease prevalent among Indian women. With a high incidence in Telangana, the dataset aims to assist nephropathologists in interpreting biopsy slides and classifying the disease.

“AI can mitigate interobserver variations in diagnosing lupus nephritis, ensuring more accurate treatment plans,” Dr Uppin explained.

Beyond traditional subtyping and grading, the project explores using AI to predict molecular changes from tissue morphology, a task typically reserved for genetic labs. Prof. Vinod said the team’s efforts to predict IDH mutations from slide images are crucial for brain tumour prognosis.

The IPD’s open-source nature provides a valuable resource for researchers and students, fostering advancements in medical AI. Prof. Vinod said, “This is one of the first instances of open-source medical data from India, tailored for our population, unlike previous reliance on datasets like the TCGA from the US.”

Going forward, the project aims to expand its datasets to include other prevalent cancers, such as breast and lung cancer, further improving India’s position in the global digital pathology landscape.