Our Take
A patent is not a product: AccurKardia has cleared the IP hurdle but still faces multi-site validation, FDA clearance, and the harder problem of proving that earlier detection actually improves outcomes and justifies screening costs.
Why it matters
Cardiac amyloidosis is systematically underdiagnosed despite newer therapies that halt progression if caught early. An ECG-based screening tool could flag at-risk patients during routine care without additional procedures, but only if the real-world accuracy and false-positive rate hold up beyond the development dataset.
Do this week
Cardiologists: request AccurKardia's validation timeline and planned trial design before integrating any ECG screening algorithm into HFpEF or LVH workups, so you can evaluate the evidence against your own diagnostic thresholds.
Patent covers three amyloidosis subtypes, but product still investigational
AccurKardia has received U.S. Patent No. 12,620,488 for a machine learning system designed to detect cardiac amyloidosis from standard 12-lead electrocardiograms. The patent covers AL amyloidosis and both wild-type and hereditary transthyretin (ATTR) amyloidosis, the major disease subtypes. The company's algorithm is currently for research use only and has not received FDA clearance.
Cardiac amyloidosis occurs when abnormal proteins accumulate in heart tissue, stiffening the muscle and causing potential heart failure. Studies estimate 13–15% of heart failure patients actually carry the diagnosis, often undetected for years (per company statement). Newer therapies can halt disease progression but cannot reverse existing cardiac damage, making earlier detection clinically significant.
AccurKardia's CEO Juan Jimenez emphasized that the ECG-based approach "works inside care that is already happening." Most patients eventually diagnosed with amyloidosis have accumulated numerous ECGs before anyone suspected the disease. The system uses explainable, feature-based machine learning rather than a black-box model, intended to support clinical interpretability and regulatory transparency.
The patent adds to AccurKardia's broader portfolio. Its AccurECG 2.0 platform is FDA-cleared; investigational applications for aortic stenosis and hyperkalemia have received FDA Breakthrough Device designations.
Real-world validation gap remains wide
The company has completed large single-site validation studies and plans multi-site retrospective validation using diverse datasets. But Jimenez acknowledged the validation requirement is rigorous: "we would want to see strong, consistent performance across patient subgroups, not just in aggregate, with sensitivity and specificity that hold up beyond the data the model was developed on."
Dr. Michelle Kittleson, professor of medicine at Cedars-Sinai, outlined the clinical questions that must be answered before adoption. Is the tool accurate across real-world populations, and in which ones? Does it prove cost-effective? Critically, what is the downstream effect on testing volume: how many diagnostic tests are triggered versus how many cases of amyloidosis are actually confirmed?
Kittleson noted that clinicians need evidence that screening improves diagnosis and treatment timing without generating excessive false positives. "Among patients with HFpEF or unexplained LVH," she said, "we would want to see AI-ECG screening identify cardiac amyloidosis earlier than usual care, with high sensitivity and an acceptable false-positive rate, leading to faster diagnosis and treatment initiation in a cost-effective fashion."
The FDA approval and evidence development required for reimbursement remain ahead. Jimenez stated the company is committed to the evidence work needed to support adoption, but no timeline or trial size has been disclosed.
Three questions before integration
Cardiologists considering ECG-based amyloidosis screening should demand answers to three questions before AccurKardia's tool reaches clinical practice.
First, performance consistency: does sensitivity and specificity hold steady across patient subgroups (age, geography, ejection fraction phenotype) in prospective or large independent retrospective cohorts? Single-site studies are a starting point, not endpoint evidence.
Second, downstream harm: what fraction of screened patients require confirmatory testing (cardiac MRI, biopsy, genetic testing) versus how many are diagnosed with amyloidosis? A low precision ratio creates unnecessary cost and anxiety.
Third, clinical impact: does earlier detection via ECG screening actually accelerate treatment initiation and improve outcomes compared to standard diagnostic pathways? Patent and FDA clearance prove the algorithm works; clinical efficacy proves it matters.