Researchers introduce an open-source approach for evaluating AI anatomy segmentation models in medical imaging without requiring ground truth annotations.
Key Details
- 1Six open-source AI segmentation models evaluated on chest CT scans from the National Lung Screening Trial (NLST).
- 2All model outputs standardized using DICOM format and harmonized medical lexicons (SNOMED-CV) for fair comparison.
- 3Strong agreement found across models for lung segmentation, but inconsistencies and systematic errors for bones and cardiac structures.
- 4Interactive visualization tools built on OHIF Viewer and 3D Slicer plugin enabled detailed side-by-side model review.
- 5Framework and datasets are open source; enables identification of reliable models and flags problematic cases.
Why It Matters
This framework addresses a major barrier for the radiology AI field by enabling fair, evidence-based model comparison in settings lacking expert-annotated ground truth. Such approaches help researchers select appropriate AI tools for large-scale studies and highlight the variability between widely used segmentation models.
Source
EurekAlert
Related News
•EurekAlert
AI-Powered OCT Enables Rapid 'Optical Biopsy' for Early Endometrial Cancer Detection
A team at Washington University has developed a catheter-based 3D OCT system with AI to quickly and noninvasively detect early endometrial cancers.
•EurekAlert
AI Clinical Reasoning in Diagnostics and Digital Fatigue in Healthcare
Recent JMIR features explore large language models in clinical diagnostics and digital fatigue among healthcare professionals.
•EurekAlert
KAIST, MIT, Microsoft Develop Efficient AI Image Upsampling for Robotics
KAIST, MIT, and Microsoft have created 'Upsample Anything,' a training-free AI method to restore high-resolution visual data from compressed images with up to 16x improved GPU memory efficiency.