AI in Mammography: Enhancing Early Breast Cancer Detection

During a six-month research period at the Kalam Institute of Health Technology in India, a multidisciplinary team worked alongside radiologists and software engineers to build a deep-learning tool for early breast-cancer detection. The system focuses on solving the high rate of false negatives associated with analog film degradation in regional screening centers.

The team used the RSNA mammogram dataset to create a convolutional neural network (CNN) classifier designed for use in clinics with limited resources. By employing custom compression and noise-reduction pre-processors, the algorithm achieves high sensitivity even on low-fidelity, desaturated, or low-contrast mammography scans typical of legacy equipment.

The system is currently undergoing clinical pilot testing to boost diagnostic accuracy in underserved areas where access to experienced radiologists is scarce. Our ongoing study aims to empower general practitioners to reliably triage urgent cases, decreasing the clinical bottleneck for specialized oncology departments.