Computer-aided detection (CAD) is the ancestor of today's breast AI. First-generation systems, launched in the 1990s, marked mammograms with circles and stars to draw attention to suspicious regions. Early results were mixed — many were criticised for high false-positive rates.
Modern AI-based CAD is different. Deep-learning models trained on hundreds of thousands of images reach accuracy that matches or exceeds expert radiologists, and they integrate into the reading workflow as a quiet second reader. Today's systems are the practical deployment of decades of research.
The evolution of CAD
- 1990s: rule-based systems that over-flagged
- 2010s: deep learning transforms accuracy
- Today: AI-based CAD works as a quiet second reader
- Tomorrow: risk-aware, personalised screening