egnite Obtains Patent for NLP Mitral Regurgitation Classifier

egnite, Inc. announced issuance of U.S. Patent No. 12,580,057 on April 6, 2026, for a natural-language-processing system that classifies mitral regurgitation mechanism from echocardiographic reports. The company positions the patented method as an operational solution to a pervasive clinical documentation problem: only 4% of echo reports explicitly document MR etiology.

The algorithm uses a hierarchical, rules-based NLP framework informed by ACC/AHA guideline logic and curated by cardiovascular specialists. Rather than relying solely on statistical or deep-learning classifiers, egnite encodes clinical decision rules to map clinician-authored report text to three MR mechanism categories: primary degenerative, secondary functional, and mixed. This design trades some model flexibility for interpretability and guideline alignment — useful properties for clinical decision support and regulatory/operational adoption.

egnite validated the system on a subset of its corpus of 5.5 million deidentified echocardiographic reports, reporting 97.3% accuracy in a maximal configuration intended for population capture and 99.0% accuracy in a minimal, high-confidence configuration. The company reports deployment across over 600 U.S. healthcare facilities. Clinically, correct MR mechanism classification directs pathway decisions: primary degenerative MR can indicate candidacy for surgical/transcatheter repair, while secondary functional MR flags patients for transcatheter edge-to-edge repair (TEER).

This patented approach addresses a practical bottleneck in retrospective and prospective patient-finding workflows for structural heart disease. For health systems and ML teams integrating clinical NLP, the combination of guideline-driven rules, high reported accuracy, and broad deployment signals a production-hardened solution focused on operational throughput and explainability — attributes that matter for clinician trust, validation, and downstream referral workflows.

Independent peer-reviewed validation and prospective outcome studies; details on error modes and annotation standards; how the algorithm handles linguistic variability across institutions; and whether the company exposes APIs, performance logs, or model cards for external audit.