This article evaluates QRS detection algorithms in ECG recordings, focusing on the agreement between different methods and the audited annotation present in the database. Using the MIT-BIH dataset as a reference, the authors implemented and compared six publicly available algorithms. The process involved generating reference annotations using QRS detectors, followed by an automatic beat-by-beat comparison of the annotations from the ones available in the database. The study used an event window of 100 ms to determine the true positives in the fusion of methods, to mitigate errors and divergences in the algorithms’ detections. This process allowed for a comprehensive evaluation of the performance of QRS detection algorithms along with method fusion. Our analysis presented the performance for each ECG lead available in the database, and pointed out a trend of dependence on the ECG leads, with the V1 lead standing out as the one with the worst sensibility and specificity. Method fusion demonstrated its ability to reduce false positives, allowing for an improvement in data classification capacity. The fused gqrs and wavedet methods showed superior metrics in the 5 channels analyzed, being among the top 3 methods in all channels.