The differential diagnosis of wide QRS premature contractions is often inaccurate in most ECG machines with automatic computer-assisted diagnosis. The purpose of the present study was to improve the accuracy of the automated differential diagnosis between premature ventricular contractions (PVC) and supraventricular contractions with aberrant conduction (A-PSC). The study investigated 180 consecutive electrocardiograms (ECGs) with wide QRS premature contractions picked up from 3,723 in the ECG database. A new algorithm, Detection of Wide QRS Complex --> Second Derivative --> Absolute Value --> Smoothing --> T Wave Subtraction --> P' Detection, was compared with a conventional QRS morphology algorithm and P' algorithm without T wave subtraction. The rate of false positives was reduced step by step (22.3% in the conventional algorithm, 7.8% in the P' without T subtraction algorithm and 3.0% in the P' with T subtraction algorithm), resulting in a marked increase in diagnostic accuracy for A-PSC from 77.2% to 90.6% and 95.0%, respectively. In a general population with external samples, the newest algorithm showed 77.8% sensitivity, 99% specificity, and 98.9% accuracy for diagnosis of A-PSC. The new algorithm for differential diagnosis of wide QRS complexes is simple, reliable, and easy to apply to most 12-lead ECG machines with computer-assisted automatic diagnosis.