Background: 12-lead ECG machine analyzes provides low accuracy level. To overcome this limitation algorithms were designed recently, but they are mainly specialized for rhythm and not morphology conditions, analyze only digital signal and can detect only one known disease per ECG. The goal of the 2020 Challenge is to identify clinical diagnoses from 12-lead ECG recordings and tackle this problem. Objective: To introduce a general, hybrid approach to an automated system for standard 12-lead ECG disease identification. Methods: Deep neural network trained using digital signals (CNN-dig). An open-source dataset provided by “PhysioNet/Computing in Cardiology Challenge 2020” was divided to 41,830 classified standard ECGs recordings (2.5 seconds of each lead and 10 seconds of lead II); 83% of this dataset was used to train CNN-dig to identify the eight most-common cardiac conditions or sinus rhythm, while 17% was used to test the result. We created 8 binary deep nets, one for each disease. If none of them gives a possibility above threshold, the record is classified as “Normal”. Results: On hidden test set, the algorithm reached and accuracy of F2 = 0.6 and G2= 0.415, which gives a geometric mean of 0.492. Conclusion: Automatic detection of cardiac conditions in standard digital 12-lead ECG signal is feasible and may allow screening of the general population and second opinions following manual interpretations of cardiologists. Key words: Artificial intelligence, Arrhythmia, Electrocardiography, Deep learning, Physionet, Challenge