ECG
Mayo Clinic, in a randomized trial of 22,641 patients: AI analysis of a routine ECG gave 32% more diagnoses of low ejection fraction, 5 extra diagnoses per 1,000 patients examined (2021). The Yaroslavl Region put an AI telecardiograph into ambulances and more than 130 rural first-aid posts: more than 145,000 ECGs in 2023, and 848 patients with acute coronary syndrome sent to emergency care. The Volgograd Region: AI analysis of ECGs in mobile units, with accuracy up to 90%, in all 33 districts.
Risk and prognosis
The Tyumen Region checked the records of more than 735,000 patients and found about 300,000 at very high risk, with no doctor involved. The Almazov Center predicts complications in hospital patients with 89% accuracy and calculates the SYNTAX score from a coronary angiogram in seconds instead of 20–40 minutes. Oxford: AI analysis of coronary artery inflammation on a routine CT predicts events up to 10 years ahead; in a pilot with 744 patients, doctors changed the treatment plan in up to 45% of cases. A hospital chain in India built its own risk score on data from more than 400,000 patients.
What a clinic can take from this
AI ECG and risk models are medical devices: check their registration. Background AI brings the doctor a list of patients on its own, with no button to press: this is how the Tyumen system works. Reminders about follow-up visits by the dates the cardiologist set can be rolled out without registration.