Lunit INSIGHT DBT

Lunit INSIGHT DBT is an artificial intelligence-powered software designed to enhance breast cancer detection by analyzing digital breast tomosynthesis images and providing radiologists with quantitative scores and visual cues for suspicious lesions. Clinical studies demonstrate that its concurrent use improves radiologists' diagnostic accuracy in identifying breast cancer, thereby aiming to improve patient outcomes in screening mammography.

Lunit INSIGHT DBT is a software as a medical device designed to assist physicians in interpreting digital breast tomosynthesis (DBT) images for the detection and diagnosis of breast cancer. Utilizing artificial intelligence technology trained through deep learning, the software automatically analyzes DBT slices to identify suspicious lesions. It provides information on the lesion type, location, an abnormality score indicating the likelihood of malignancy, and an outline of the suspected cancerous regions. This information is intended to be used concurrently by radiologists to augment their workflow and improve the accuracy of breast cancer detection and characterization in screening mammograms for the female population. The performance of Lunit INSIGHT DBT was evaluated through non-clinical standalone testing and a clinical assessment. In standalone tests using 2,202 DBT exams, the software achieved an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.928, surpassing the predicate device's performance target. A retrospective multi-reader multi-case clinical study involving 15 radiologists and 258 DBT exams demonstrated that using Lunit INSIGHT DBT improved diagnostic performance. The radiologists' AUROC increased from 0.897 without the software to 0.915 with its assistance, a statistically significant improvement (p=0.0498), confirming its effectiveness in aiding the detection and diagnosis of breast cancer.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the performance of Lunit INSIGHT DBT vary across different breast density categories, and are there specific BI-RADS categories where its assistance is most impactful?; What is the computational time required for Lunit INSIGHT DBT to analyze a typical DBT case, and how does this integrate into a high-volume screening workflow?; Were there specific types of breast cancer (e.g., invasive ductal, invasive lobular, DCIS) for which the software showed higher or lower detection sensitivity in the standalone or clinical studies?; How does the software's lesion localization accuracy (outline of suspected regions) compare to radiologist annotations, and what is the clinical significance of any discrepancies?; What were the characteristics of the false positives and false negatives generated by Lunit INSIGHT DBT in the clinical study, and how might these inform future algorithm refinement or radiologist training?