Computer-aided diagnosis for lung cancer screening

A new study has demonstrated that integrating artificial intelligence (AI) into lung cancer screening processes can significantly improve diagnostic specificity, potentially reducing the number of unnecessary follow-up procedures and associated patient anxiety. The research, conducted by a team including experts from Google Health, DeepHealth, and Apollo Radiology International, involved a multi-case, multi-reader study with 12 radiologists using pre-existing, de-identified CT scans. The AI system was designed to assist radiologists in evaluating lung images for signs of cancer. In the study, each radiologist reviewed 627 challenging cases twice — once with and once without the AI assistance. The results showed that when using the AI-assisted system, radiologists' specificity improved by an absolute 5–7% compared to when they worked without assistance. This translates to potentially avoiding one unnecessary follow-up visit for every 15–20 patients screened. "This improvement in specificity could help reduce the burden on the healthcare system and significantly decrease patient anxiety associated with false positives," said Corbin Cunningham, one of the key contributors to the project. "It also supports the sustainability of lung cancer screening programs as more individuals become eligible for screening." The AI model was tested across different clinical settings and populations, showing its adaptability to various cancer scoring systems and patient demographics. The researchers also open-sourced code used in the study to support other researchers working on similar translational research in medical imaging. "We're excited to collaborate with partners like DeepHealth and Apollo Radiology International to bring this technology into real-world clinical practice," said Krish Eswaran, another lead contributor. "Our goal is to make AI tools that not only improve diagnostic accuracy but also enhance the overall patient experience and healthcare delivery." The findings were published in a blog post by Google Health and are expected to influence future developments in AI-assisted diagnostic tools for lung cancer screening. The research underscores the importance of integrating AI in clinical workflows in a way that supports, rather than replaces, the expertise of medical professionals.