Computer-aided diagnosis for lung cancer screening
Google's lung cancer AI cuts false positives by 5-7%. Every 15-20 patients screened, one avoids unnecessary procedures.
Why it matters
Google deployed a production ML system for lung cancer screening that improves radiologist accuracy in real clinical settings across two countries, with partnerships underway to integrate into commercial health platforms. This demonstrates practical translation of AI research into healthcare infrastructure.
The key facts
8 to knowSpecificity improved 5-7% with ML assistance in both US and Japan reader studies
System reduces unnecessary follow-up procedures for 1 in 15-20 screened patients
Lung cancer causes 1.8 million deaths globally annually; 20% mortality reduction achievable via early screening
US Preventive Services Task Force expanded screening recommendations by ~80%
System deployed on Google Kubernetes Engine with PACS integration for radiologist workstations
Published in Radiology AI journal with multinational validation (US and Japan)
Partnerships with DeepHealth and Apollo Radiology International for commercial deployment
Open-sourced code for CT image processing and reader study frameworks
Go to the source
Google Research Blogblog.research.google
Publisher excerpt: Posted by Atilla Kiraly, Software Engineer, and Rory Pilgrim, Product Manager, Google Research Lung cancer is the leading cause of cancer-related deaths globally with 1.8 million deaths reported in 2020. Late diagnosis dramatically reduces the chances of survival. Lung cancer screening via computed…