A silent killer now has a digital watchdog. The FDA has cleared the first AI algorithm designed to diagnose cardiac amyloidosis from routine ECG waveforms — a condition that routinely slips past human cardiologists until it’s already too late.
Developed by Anumana, a GE HealthCare spinout built on Mayo Clinic research, the ECG-AI algorithm achieved 78.9% sensitivity and 91.2% specificity across a validation study of more than 15,000 adults. It does something no human can: detect the subtle, combined pattern of ECG anomalies that betray a disease that masquerades as more common heart conditions.
Why This Matters More Than Benchmarks
Cardiac amyloidosis is caused by abnormal protein deposits in the heart. It’s fatal if missed, treatable if caught early — but it’s notoriously underdiagnosed because its symptoms overlap with far more common cardiac conditions. By the time most patients receive a correct diagnosis, significant heart damage has often already occurred.
The ECG has been sitting in front of doctors the entire time. The signals were always there. Humans just couldn’t reliably see them.
“Cardiac amyloidosis can be challenging to detect early, especially when its signs overlap with more common heart conditions,” said Dr. Martha Grogan, consultant in cardiovascular medicine at Mayo Clinic and co-principal investigator of the clinical study. “A tool that helps clinicians recognize suspicion of amyloidosis from a routine ECG could support earlier diagnosis and more timely next steps in care.”
This isn’t an AI suggesting a diagnosis. It’s an AI making one — from data already collected in standard clinical practice, with no additional tests required.
From Assistive to Primary
That distinction is critical. Most FDA-cleared AI in healthcare operates in an assistive role: flagging potential concerns, nudging doctors toward second looks, prioritizing caseloads. Anumana’s algorithm works differently. It detects patterns invisible to the human eye across ECG waveform data and outputs a diagnostic signal that directly informs treatment decisions.
“What makes this work especially meaningful is the rigor of the validation,” said Dr. Angela Dispenzieri, hematologist at Mayo Clinic and co-principal investigator. “This ECG-AI algorithm was validated in a large multicenter study that included both ATTR and AL cardiac amyloidosis at major referral centers with deep expertise in amyloidosis diagnosis.”
Both major types of cardiac amyloidosis — ATTR and AL — were represented in the validation data, which spanned major referral centers. This isn’t a single-site study with cherry-picked results.
The Pattern: AI Excels Where Human Perception Fails
This clearance follows a growing pattern in medical AI: the technology proves most valuable precisely where human perception is weakest.
Cardiac amyloidosis is defined by a constellation of subtle ECG features that no individual cardiologist can reliably combine into a diagnostic signal. The AI doesn’t have that limitation. It was trained on patterns across thousands of confirmed cases and can recognize the specific combination of waveform features that indicate the disease.
Anumana already has two other FDA-cleared ECG-AI algorithms — one for low ejection fraction and another for pulmonary hypertension. Both address conditions characterized by delayed or missed diagnoses. Each new clearance makes the routine ECG more valuable as a screening tool.
“The more conditions we can identify from a single ECG, the more valuable the test becomes in clinical practice,” said Anumana CEO Maulik Nanavaty. “That’s what Anumana is working toward with each new clearance.”
What Changes When AI Gets Regulatory Authority
The FDA clearance transforms this from an experiment into infrastructure. Hospitals can now integrate the algorithm into existing ECG workflows. No new equipment. No additional tests. No specialist referrals just to run the analysis. The AI reads the same ECG that’s already being taken and surfaces a signal that would otherwise remain invisible.
This is the trajectory medical AI has been building toward: not replacing doctors, but expanding the diagnostic range of routine tests beyond what human perception can achieve. When a standard 12-lead ECG becomes a screening tool for a disease that currently requires specialized imaging to confirm, the economics and accessibility of early detection change fundamentally.
For cardiac amyloidosis patients, the difference is measured in years of life. For the healthcare system, it’s measured in avoided late-stage interventions. For AI’s trajectory, it’s measured in regulatory precedent: the FDA is increasingly comfortable clearing algorithms that make diagnoses, not just suggest them.
SOURCES
- Inside Precision Medicine — FDA Clears First AI Algorithm to Diagnose Cardiac Amyloidosis
- Anumana — ECG-AI Algorithm Clinical Validation Study
- Mayo Clinic — Dr. Martha Grogan & Dr. Angela Dispenzieri, co-principal investigators
