iMerit, a leader in software delivered AI data solutions, has announced the launch of ANCOR, its Annotation Copilot for Radiology. ANCOR, integrated with iMerit’s Ango Hub product, is an AI-driven tool designed to enhance efficiency and accuracy of radiology AI development.
ANCOR, released in limited beta for select customers at the annual conference of the Radiological Society of North America (RSNA), acts as an intelligent assistant that helps radiologists and AI professionals develop training data for AI models while maintaining patient safety. It automates repetitive tasks, provides real-time expert guidance, and ensures adherence to complex project guidelines. It also allows customization for various medical image data annotation workflows, from mammography to cardiology.
“The Annotation Copilot aims to benefit radiologists by reducing workloads and increasing productivity. Developers also gain by having access to high-quality annotations for better model development.” said Dr. Sina Bari, AVP, Healthcare & Life Sciences AI, iMerit.
In customer tests, ANCOR reduced research time from an hour to 15 minutes. Complex guidelines running to 90 pages become instantly accessible via queries for the radiologist. On the annotation side, ANCOR reduced lookup time, resulting in 2x output speeds with 38% better accuracy.
Also Read: Viz.ai Collaborates with Microsoft to Advance AI-powered Clinical Workflows
Key features of the Annotation Co-Pilot include:
- Sub-Domain Expertise: Fine-tuned models for specialized areas of radiology
- Customization: Tailored taxonomy and training aligned with specific project needs
- Context-Aware Guidance: Interprets project instructions to provide context-driven support
- Real-Time Annotation Support: Instant suggestions and highlights of areas of interest
- Multimodal Capabilities: Combining text and image analysis for comprehensive understanding
- Three Modes of Operation:
- Workflow: Automatically generates initial annotations
- Interactive: Allows annotators to adjust, verify, and query annotations in real time
- Validation: Cross-checks annotations against guidelines and ground truth data
“Customers have been asking for our insights from hundreds of projects to be applied to more optimized medical image data annotation workflows.” said Radha Basu, CEO. “Their models are tackling increasingly complex problems, so we need to provide increasingly nuanced and intricate decisions. Our experts are driving the design of the copilot in order to reduce their cognitive load. It fits into our overall Ango Hub workflow for medical imaging.”
“We expect ANCOR to progressively become more powerful.” added Dr. Bari. “Future versions can act as an autonomous agent to revise guidelines, communicate with the client, and give the team feedback.”
Source: PRNewswire
