AI-assisted diagnosis for lung cancer screening
Google's AI model enhances radiologists' performance in early detection.
1Challenges in lung cancer diagnosis
Lung cancer is the leading cause of cancer-related deaths globally. Late diagnosis significantly reduces survival chances. While CT scans are crucial for early detection, false positives can lead to unnecessary procedures for patients.
2Google's AI solution
Google developed machine learning models for lung cancer screening. These models automatically detect and classify potential cancer signs, delivering results comparable to those of specialists.
3Model communication capability
Effectively communicating AI findings to radiologists is critical. Google developed a user-centric interface for this purpose. The interface analyzes CT images and highlights suspicious regions.
4Suspicion rating system
The system analyzes CT images and assigns a suspicion rating in four categories: no suspicion, probably benign, suspicious, highly suspicious. This rating supports radiologists' decision-making process.
5Tests in the US and Japan
Google conducted studies with radiologists in the US and Japan to evaluate the system's clinical performance. With model assistance, radiologists' specificity increased in both studies.
6Google Cloud integration
The system runs on Google Cloud using Google Kubernetes Engine. This enables the processing of CT images and writing results to DICOM stores, offering scalability and speed.
7Open-source code sharing
Google shared open-source code to process CT images and generate PACS-compatible images. This is a valuable resource for researchers aiming to accelerate similar studies.
8Future potential
AI-assisted systems can reduce radiologists' workload in lung cancer screening and improve early detection rates. This could be a significant step in the fight against cancer.
Original article: Google AI Blog — Akciğer kanseri taramasında yapay zeka destekli teşhis