RadScan AI Uses GCP and Vertex AI to Cut Radiology Scan Review Time by 6 Minutes
Developers built RadScan AI, an autonomous radiology triage tool, as an entry for the Google Cloud and Devpost All Things Agentic Hackathon. The system targets radiologist burnout, as clinicians currently spend 10 to 15 minutes manually reviewing each volumetric MRI study. RadScan AI combines a 2.5D CNN-BiGRU neural network trained on over 819,000 DICOM images with Google Cloud Run GPU microservices and Vertex AI Gemini models to detect 12 pathology types in under three seconds. The tool generates Grad-CAM visual heatmaps to highlight lesion locations and automatically drafts structured clinical reports, saving an estimated six minutes per scan. The backend is deployed on GCP Cloud Run with scale-to-zero support, keeping idle infrastructure costs at effectively zero.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in