Atrial fibrillation (AF) is part of a type of heart disease characterized by a rhythmic irregular heartbeat. AF conditions that occur continuously can potentially cause a stroke for sufferers. The method of reading and detecting the possibility of AF is needed to prevent the risk of stroke due to AF. In this research, the Recurrent Neural Network (RNN) method is used in classifying electrocardiogram readings to obtain accuracy in the assessment of AF. The data information used in the study was obtained from physicians who were the bases of ECG result image data, and data information was also obtained by implementing directly through a simple and low-cost ECG using Arduino AD8232 to test user information directly related to AF conditions at the user’s heart. RNN method that is tested can obtain more accurate accuracy values in detecting AF heart rate abnormalities, and the Arduino AD8232 module can be a good ECG in reading low-cost but high-accuracy heart records.

The medical world, especially those related to diseases and management of the heart uses ECG as a measurement tool. ECG has important points determined based on predetermined characteristics. The point is PQRST, where three of them are used as research objects in this paper. AD8232 is used as a research medium where the RST points must be determined in the AD8232 plot results by first determining the R points based on the highest peak. The results obtained were satisfactory wherein from 10 ECG graphic samples, 9 of them obtained RST point measurements which tended to be similar to conventional ECG measurements using millimeter paper as plotting media. Accuracy values reaching more than 90% indicate the reliability of the implementation results.