Key points from article :
AI can detect past atrial fibrillation in ECGs even when the heart is beating normally.
Uses a convolutional neural network to detect the ECG signature of AF present during normal rhythm.
Analysed ECGs of 181,000 patients between 1993 and 2017 who had had normal test results.
Correctly identified the subsequent diagnosis from the normal test results in 83% of cases.
Can lead to easier and earlier detection of atrial fibrillation.
Needs to be tested further on general population.
Research by Mayo Clinic, USA published in The Lancet.