PowerLook Tomo Detection V2 Software

PowerLook® Tomo Detection V2 is a computer-assisted detection software that utilizes deep learning algorithms to identify suspicious soft tissue densities and calcifications in digital breast tomosynthesis images, providing quantitative scores to aid radiologists. Clinical studies demonstrate that its use improves diagnostic accuracy, specifically enhancing sensitivity and area under the ROC curve, while significantly reducing image interpretation time for physicians.

PowerLook® Tomo Detection V2 is a computer-assisted detection and diagnosis (CAD) software designed to aid physicians in interpreting digital breast tomosynthesis (DBT) examinations. Its primary function is to detect soft tissue densities, such as masses, architectural distortions, and asymmetries, as well as calcifications within 3D DBT slices. The software employs deep learning technology and pattern recognition algorithms to identify suspicious breast lesions. For each detected region, it provides a visual mark on the tomosynthesis image and assigns a "Certainty of Finding Score." Additionally, a "Case Score" is assigned to the entire examination. These scores, on a 0-100% scale, reflect the algorithm's confidence in the malignancy of a finding or case, based on a 50% cancer prevalence model, and are intended as relative indicators to guide physicians rather than absolute probabilities of malignancy. Standalone performance testing on Hologic and GE DBT systems demonstrated that the software met design specifications for sensitivity, false positive rates, and specificity. A pivotal multi-reader, multi-case clinical study involving 24 radiologists and 260 Hologic DBT cases showed significant benefits. Radiologist performance, measured by the area under the ROC curve (AUC), was superior when using the CAD software with DBT (AUC 0.852) compared to DBT alone (AUC 0.795). Furthermore, the software significantly reduced radiologist reading time by 52.7%. The study also confirmed superior case-level and lesion-level sensitivity with CAD, alongside non-inferior specificity and recall rates in non-cancer cases, indicating improved diagnostic accuracy and efficiency.

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 "Certainty of Finding Score," calibrated on a 50% cancer prevalence population, translate to clinical utility in a typical screening population with much lower cancer prevalence, and what steps are recommended for physicians to interpret these scores appropriately?; Given the observed 52.7% reduction in reading time, what is the impact on reader fatigue and overall diagnostic throughput in a high-volume clinical setting, and does this efficiency gain maintain or improve detection rates for subtle or early-stage cancers over prolonged use?; The standalone performance was tested on Hologic and GE DBT systems. How does the algorithm's performance vary with other DBT manufacturers or imaging protocols not included in the study, and what is the process for validating compatibility with new systems?; While case-level and lesion-level sensitivity improved with CAD, how did the software specifically impact the detection rates of different subtypes of lesions (e.g., spiculated masses vs. circumscribed masses, or microcalcification morphology types) beyond the broad categories of soft tissue densities and calcifications?; The pivotal study used an "enriched sample" of DBT cases. How might the reported performance metrics (AUC, sensitivity, specificity) differ in a purely sequential, unenriched screening population, and are there plans for post-market surveillance to monitor real-world performance?