ZICQ
中 Log in / Sign up
Newsroom LLMs & Foundation Models #Google Research #AI in Healthcare #Lung Cancer Screening #Machine Learning #Medical Imaging

Google Research Releases Assistive AI System to Enhance Lung Cancer Screening Efficiency and Accuracy

Avatar of Mr.Xu

By Mr.Xu Compiled & Reviewed by Editorial

Published:

中文阅读 (Chinese) English Version
Co-reading Select any sentence in the article to highlight, annotate, or copy a quote card. Join discussion ↓

Summary:Google Research has released an AI-powered assistive system designed to optimize lung cancer screening processes. The system takes CT images as input and outputs a cancer suspicion rating along with the localization of suspicious regions. Clinical studies conducted in the US and Japan have demonstrated that the system significantly improves radiologists' specificity in identifying cancer-negative cases, potentially reducing unnecessary follow-up procedures, lowering patient anxiety, and decreasi


Background and Challenges

Lung cancer is the leading cause of cancer-related deaths globally, with 1.8 million deaths reported in 2020. Late diagnosis significantly reduces the chances of survival. Screening via computed tomography (CT) has been shown to reduce mortality in high-risk populations by at least 20% by detecting potential signs of cancer earlier. However, false positives can cause patient anxiety and lead to unnecessary procedures, increasing healthcare costs. Additionally, the efficiency of screening large numbers of individuals is constrained by healthcare infrastructure and radiologist availability.

AI-Assisted Screening System

Google Research has developed machine learning models that have demonstrated high performance in lung cancer detection, comparable to that of specialists. To further enhance AI's performance in realistic environments, the team has developed an assistive AI system. The system takes CT images as input and outputs a cancer suspicion rating (no suspicion, probably benign, suspicious, highly suspicious) along with the localization of suspicious regions.

System Architecture and Functionality

The system comprises 13 models with an architecture similar to the end-to-end system used in prior work. The models coordinate to first segment the lungs, perform an overall assessment, locate three suspicious regions, and then assign a suspicion rating to each region. The system is deployed on Google Cloud using Google Kubernetes Engine (GKE) and connects directly to servers storing DICOM images.

Clinical Study Results

The research team conducted multi-center clinical studies in the US and Japan, involving 627 challenging cases and 12 radiologists. The results showed that with model assistance, radiologists' specificity in identifying cancer-negative cases increased by 5-7%, potentially reducing unnecessary follow-up procedures, lowering patient anxiety, and decreasing healthcare costs, while maintaining sensitivity for cancer-positive cases.

Future Directions and Partnerships

To accelerate the application of AI in medical imaging, the team has open-sourced code for processing CT images and generating images compatible with PACS systems. Additionally, Google Research is collaborating with DeepHealth and Apollo Radiology International to explore the integration of this system into future products. The team also plans to support other researchers by open-sourcing the code used in the reader study and incorporating the insights described in this blog.

Conclusion

The assistive AI system demonstrates the significant potential of AI in lung cancer screening, particularly in improving screening efficiency, reducing healthcare costs, and alleviating patient anxiety.


Source: Google AI Research (2024-03-20)

— END —

Tags: #Google Research #AI in Healthcare #Lung Cancer Screening #Machine Learning #Medical Imaging

Editorial & Fact-Checking Note: This article is compiled from primary research, official release documentation, and source papers by the ZICQ Newsroom pipeline with automated entity verification and human editorial review. If you notice any technical inaccuracy, please submit a correction via our corrections policy or email our editorial desk directly.

What's your take on this story?
🚀 Explore & Submit Works

Community Comments

Loading live comments and annotations…