IIT-Indore Researchers Pioneer AI for Early Cancer Detection
In a significant advancement that could transform early cancer screening and improve accessibility, scientists at the Indian Institute of Technology, Indore have engineered sophisticated artificial intelligence algorithms capable of automatically identifying and pinpointing breast and cervical cancers within medical images with remarkably high precision.
AI System Designed to Assist Medical Professionals
The innovative AI framework is specifically crafted to support physicians by rapidly analyzing diagnostic images, highlighting suspicious areas, and substantially decreasing the likelihood of missed diagnoses. Research leaders emphasize that this technology holds particular promise for resource-limited environments where specialist medical personnel are scarce, facilitating the prioritization of high-risk patients and enabling prompt therapeutic interventions.
Research Team and Development Details
The groundbreaking project was spearheaded by Professor Kapil Ahuja alongside his dedicated team at the Mathematics of Data Science and Simulation Laboratory within the Department of Computer Science and Engineering. Key contributors include PhD scholar Saurabh Saini, former postdoctoral researcher Dr. Deepti Tamrakar, and former PhD student Dr. Aditya A. Shastri.
For breast cancer detection, the team engineered a histogram of oriented texture descriptor algorithm that meticulously examines subtle texture patterns in mammograms. These patterns exhibit irregularities when cancerous growth occurs, allowing the method to differentiate between healthy and malignant tissue even in cases involving dense breast composition.
For cervical cancer identification, researchers constructed a deep learning model named the block-sused attention-driven adaptively-pooled ResNet descriptor. This sophisticated system captures both granular features like color variations and edge definitions, along with abstract structural characteristics extracted from colposcopy images.
Exceptional Performance on International Datasets
The AI models underwent rigorous evaluation using four distinct international datasets, achieving accuracy rates consistently in the mid-to-late 90th percentile range. These results demonstrate substantial superiority over existing conventional techniques currently employed in medical imaging analysis.
Institutional Support and Future Directions
IIT-Indore Director Professor Suhas Joshi commented, "This research embodies our institute's dedication to creating technology-driven solutions for pressing national healthcare challenges. The team prioritized developing AI systems that provide transparent explanations for their diagnostic conclusions, enhancing physician comprehension and confidence in the outcomes."
The initiative to transition this technology from laboratory to clinical application receives backing from the DRISHTI Cyber Physical Systems Foundation at IIT-Indore.
Professor Ahuja elaborated on future plans, stating, "Our artificial intelligence algorithms were initially trained on global datasets primarily comprising patients of European ancestry. We are currently developing a prototype specifically trained on Indian patient data through collaboration with Dr. Renu Dubey Sharma of HCG Cancer Hospital in Indore. Furthermore, we intend to expand this methodological approach to address other prevalent cancers including thyroid, lung, oral, colorectal, and oesophageal malignancies."
