Lunit INSIGHT MMG
K211678Lunit Inc. · cleared 2021-11-17 · product code QDQ · Radiology
Premarket evidence — what FDA accepted
source quote (p.5)
“Lunit INSIGHT MMG is a radiological Computer-Assisted Detection and Diagnosis (CADe/x) software for aiding interpreting physicians with the detection, localization, and characterization of suspicious areas for breast cancer on mammograms from compatible FFDM (full-field digital mammography) systems. The software applies an artificial intelligence algorithm for recognition of suspicious lesions, which are trained with large databases of biopsy proven examples of breast cancer, benign lesions and normal tissues. and Software-only device”
source quote (p.5)
“The software applies an artificial intelligence algorithm for recognition of suspicious lesions, which are trained with large databases of biopsy proven examples of breast cancer, benign lesions and normal tissues. and In Lunit INSIGHT MMG, a range of medical image processing and machine learning techniques are implemented. The system includes 'deep learning' algorithm applied to images for recognition of suspicious lesions for breast cancer. The machine learning components are trained to detect suspicious lesions for breast cancer with large databases of biopsy-proven cases of breast cancer, benign lesions and normal tissues.”
source quote (p.5)
“The software applies an artificial intelligence algorithm for recognition of suspicious lesions, which are trained with large databases of biopsy proven examples of breast cancer, benign lesions and normal tissues.”
Validation studies (2)
Standalone
n=2,412 images
endpoints: ROC AUC; LROC AUC; sensitivity; specificity
Reader study (MRMC)
n=240 cases
endpoints: diagnostic ability of radiologist with CAD assistance is superior to without CAD assistance; FBR (Forced BI-RADS) ROC AUC between reader's interpretation and device-assisted interpretation for the detection of malignant lesions; inter-test difference in LCM (level of confidence of malignancy) LROC AUC; LCM ROC AUC; recall rate in cancer and non-cancer group (sensitivity and 1-specificity)
Reported performance (12 observations)
source quote (p.8)
“Sensitivity (%) [95% CI] 85.74 [82.95, 88.53]”
source quote (p.8)
“Specificity (%) [95% CI] 75.62 [73.64, 77.60]”
source quote (p.8)
“ROC AUC [95% CI] 0.903 [0.889, 0.917]”
source quote (p.8)
“Type I LROC AUC [95% CI] 0.781 [0.751, 0.812]”
source quote (p.8)
“Type II LROC AUC [95% CI] 0.792 [0.763, 0.822]”
source quote (p.9)
“ROC AUC (95%CI) 0.754 (0.702, 0.807)”
source quote (p.9)
“ROC AUC (95%CI) 0.805 (0.759, 0.852)”
source quote (p.9)
“Test2-Test1 0.051 (0.027, 0.075)”
source quote (p.9)
“LCM ROC AUC 0.052 (95% CI: 0.026 - 0.079)”
source quote (p.9)
“recall rate in cancer group (sensitivity) 5.97(95% CI: 2.48 - 9.46)”
source quote (p.9)
“recall rate in non-cancer group (1-specificity) -1.46 (95% CI: -3.41 - 0.05)”
source quote (p.9)
“ROC AUC in the standalone performance analysis is 0.863 (95% CI: 0.818 - 0.909)”
Each value carries its own analysis unit and task — never compare or pool across devices. Source: 510(k) summary PDF.
Predicate network
Postmarket — what happened after clearance
Recall and MAUDE counts are product-code-level (reports aren't reliably attributable to one device). Signals are descriptive observables with sources — never a judgment that the device is unsafe or drifting. Snapshot 2026-07-08.
Reimbursement — how devices like this got paid
Not yet tracked — no payment pathway indexed for this clearance (the reimbursement corpus is a growing seed set).