Lunit INSIGHT DBT

Lunit INSIGHT DBT is an AI-powered software solution designed to enhance lesion detection in digital breast tomosynthesis (DBT) exams. Built for radiologists managing high case volumes, it analyzes multi-slice 3D mammography studies to highlight suspicious findings, reduce reading fatigue, and support faster, more confident interpretation of complex breast imaging cases.

The platform integrates seamlessly into PACS and existing imaging viewers, supporting both single and double-reading workflows. By automating slice navigation and providing clear AI-guided decision support, Lunit INSIGHT DBT helps radiology teams streamline their workflow while maintaining high diagnostic precision across diverse patient populations.

Core Detection and Classification Capabilities

• Suspicious lesions are automatically identified and assigned an abnormality score from 0 to 100, quantifying the likelihood of malignancy to support consistent, confident decision-making.
• Lesion locations are clearly marked using heatmaps or contours, providing visual guidance during image review.
• Lesions are classified by type, including masses, architectural distortions, asymmetries, and calcifications, enabling targeted interpretation.
• The optimal slice for lesion visualization is automatically flagged on the navigation bar, eliminating the need for manual scrolling through image stacks.
• AI-guided navigation allows radiologists to jump directly to key slices, significantly reducing review time.

Clinically Validated Performance

• Lunit INSIGHT DBT has been shown to improve breast cancer detection in both fatty and dense breast tissue, supporting more consistent interpretation across tissue types and helping identify subtle or early-stage cancers that may be missed in standard screening.
• Trained on diverse datasets from the United States, United Kingdom, and Korea, the AI delivers robust and consistent performance across racial and ethnic groups. A study of over 163,000 DBT exams conducted at Emory confirmed consistent results across patient demographics, supporting more equitable screening outcomes.
• Improved specificity helps reduce unnecessary recalls and false positives, lowering patient anxiety and enabling radiologists to interpret exams with fewer interruptions and clearer decision support.
• In a published study, Lunit INSIGHT DBT correctly localized nearly one-third of interval cancers, many of which were large or more advanced at the time of standard screening. Earlier identification of these cancers has the potential to improve outcomes and reduce breast cancer-related mortality and morbidity.
• Clinical evidence is supported by peer-reviewed publications, including studies in Radiology Artificial Intelligence (2024) and Radiology (2025).

AI Scoring and Interpretability

• Each lesion receives a numerical abnormality score (0–100) reflecting the AI's confidence that a malignant lesion is present, not the severity or prognosis of the finding.
• The scoring scale is non-linear; a higher score indicates greater AI confidence in malignancy, but scores should not be interpreted as proportional comparisons.
• The AI is designed to predict malignancy likelihood regardless of cancer subtype, meaning invasive and noninvasive cancers may receive varying scores based on the model's assessment.

Workflow Integration and Efficiency

• Designed for seamless integration with PACS and imaging viewers, minimizing disruption to existing radiology workflows.
• Supports both single and double-reading workflows, making it adaptable to different institutional screening protocols.
• Reduces overall interpretation time for DBT exams, as demonstrated in peer-reviewed research, helping radiology departments manage increasing case volumes.
• Simplifies slice selection through automated navigation, reducing cognitive load and reading fatigue for radiologists.

Lunit INSIGHT DBT is part of the broader Lunit INSIGHT Breast Suite, which also includes Lunit INSIGHT MMG for 2D mammography. The product is intended for use in clinical environments where PACS integration and viewer compatibility are essential. Availability may vary by country and region.