In a major breakthrough aimed at improving early cancer detection, researchers at Indian Institute of Technology Indore have developed advanced artificial intelligence algorithms that automatically detect and localise breast and cervical cancers in medical images with very high accuracy. Through this innovation, the team seeks to make cancer screening faster, more reliable, and more accessible.
Enhancing Screening Accuracy and Accessibility
The AI system assists doctors by rapidly analysing medical images, identifying suspicious regions, and reducing the risk of missed diagnoses. Importantly, the technology can significantly benefit resource-constrained settings where specialist availability remains limited. By prioritising high-risk cases, the system enables timely intervention and improves patient outcomes.
Research Leadership and Team
Prof. Kapil Ahuja led the research at the Math of Data Science and Simulation Lab in the Department of Computer Science and Engineering at IIT-Indore. His team included PhD scholar Saurabh Saini, former postdoctoral researcher Dr. Deepti Tamrakar, and former PhD student Dr. Aditya A. Shastri. Together, they combined expertise in artificial intelligence, medical imaging, and data science to build clinically relevant models.
Advanced Algorithm for Breast Cancer Detection
For breast cancer screening, the team developed a Histogram of Oriented Texture Descriptor algorithm. This method analyses subtle texture patterns in mammograms, which tend to become irregular when cancer develops. Notably, the algorithm effectively distinguishes between healthy and cancer-affected tissue, even in dense breast cases where conventional screening often faces limitations.
Deep Learning Model for Cervical Cancer
In parallel, the researchers created a deep learning model called the Block-Sused Attention-Driven Adaptively-Pooled ResNet Descriptor for cervical cancer detection. This model extracts both detailed visual features—such as colour and edges—and higher-level structural patterns from colposcopy images. As a result, it provides comprehensive image analysis that enhances diagnostic precision.
High Accuracy Across International Datasets
As reported by TOI, the team validated both models using four international datasets. The algorithms achieved accuracy levels in the mid-to-late 90 percent range, significantly outperforming several existing techniques. These results demonstrate the robustness and global applicability of the system.
Institutional Support and Explainable AI
Prof. Suhas Joshi, Director of IIT-Indore, stated that the research reflects the institute’s commitment to developing technology-driven solutions for critical national healthcare challenges. He further emphasized that the team incorporated explainable AI mechanisms into the system. Consequently, the algorithms clearly demonstrate how they reach their conclusions, thereby strengthening clinicians’ trust in the results.
The lab-to-market initiative currently receives support from the DRISHTI Cyber Physical Systems Foundation, which aims to accelerate the translation of research into real-world healthcare applications.
Towards India-Specific AI Models and Broader Cancer Applications
Looking ahead, Prof. Ahuja explained that the current AI models were trained on global datasets predominantly comprising patients of European ancestry. Therefore, the team has initiated the development of a prototype trained on Indian patient data in collaboration with Dr. Renu Dubey Sharma of HCG Cancer Hospital.
Furthermore, the researchers plan to expand this AI-driven approach to detect other major cancers, including thyroid, lung, oral, colorectal, and oesophageal cancers. Through these efforts, IIT-Indore aims to strengthen early cancer detection capabilities and contribute meaningfully to India’s evolving digital health ecosystem.