Bunkerhill ECG-EF

Bunkerhill ECG-EF is a machine learning-based software that analyzes 12-lead ECG data to aid clinicians in screening for left ventricular ejection fraction (LVEF) ≤40% in adults at risk for heart failure. It is not intended as a stand-alone diagnostic device and requires clinician judgment combined with additional diagnostic testing such as echocardiography for final diagnosis.

The device uses deep learning algorithms to process 10-second digital ECG waveforms at 500Hz sampling rate and outputs risk classification (Low EF Screen Positive, Low EF Screen Negative, or Error). It is software-only, delivered as a Docker container module integrated with third-party EMR or ECG management systems, with no proprietary graphical user interface. The algorithm was trained using transthoracic echocardiogram as ground truth and is restricted to validated ECG acquisition systems using standard Ag-AgCl electrodes.

Not stated in this summary. The document references FDA guidance documents on "Content of Premarket Submissions for Device Software Functions" and "Cybersecurity in medical devices" but does not cite specific ISO, IEC, or ASTM standards.

Bunkerhill ECG-EF is substantially equivalent because it shares the same intended use (screening for LVEF ≤40%), operational principles (machine learning analysis of 12-lead ECG), patient population (adults at risk), and performance characteristics as the predicate Low Ejection Fraction AI-ECG Algorithm (K232699). Validation on 15,994 diverse patient records demonstrated sensitivity of 82.66%, specificity of 83.20%, PPV of 37.20%, and NPV of 97.54%—meeting all acceptance criteria (Se/Sp ≥80%, PPV ≥25%, NPV ≥95%). Minor differences in labeling precision do not alter the diagnostic function or raise different safety and effectiveness questions.