A team of researchers has developed a breakthrough AI that successfully diagnoses breast cancer more accurately than a trained radiologist.
Artificial intelligence technology has been used to successfully automate some pretty complex tasks in recent years, and some the most promising breakthroughs have been in the medical technology field. Computer vision systems are often used to analyse medical scans for possible signs of cancer, but ultimately it's just a tool to help radiologists interpret scans and make their diagnosis.
A new AI has been demonstrated that actually beats the doctors at accurately diagnosing brast cancer from mammogram images. The system was built by an international team composed of researchers from Google and Imperial College London, and once complete it outperformed six expert radiologists at the task.
The model was trained using images from almost 29,000 scans and ended up with 2.7% fewer missed cancer cases than a real radiologist and 1.2% fewer false positives. The AI also had less information to go on as it only had access to scan images while the radiologists had the patients' medical histories and family histories too.
This breakthrough is important because the NHS currently has a shortage of qualified radiologists to read scans, while the AI can work continuously. It's hoped that it could help catch breast cancer earlier and lead to better survival rates for those affected. The research was published in the science journal Nature this month.
Source: BBC News
