QRS Complex Detection Using Adaptive Lead Selection

The patent segments the QRS detection process into two distinct phases: an offline learning phase where normal QRS templates are acquired and stored, and an online detection phase where these templates are used for comparison. This segmentation allows the system to establish a reference baseline of normal QRS complexes free from stress-induced noise, which then serves as a filter during actual stress testing, improving detection accuracy despite the presence of muscle noise and artifacts.

The patent performs preliminary action by acquiring and storing normal QRS complex templates during a baseline period before stress testing begins. This preliminary template acquisition creates a reference database of what normal QRS complexes look like for each patient, which is then used during stress testing to identify and filter out false detections caused by muscle noise and artifacts, thereby improving detection accuracy.

The patent implements feedback by continuously comparing detected QRS complexes against the stored normal templates and using the results to adjust detection thresholds and parameters. The system analyzes detection outcomes and feeds this information back into the detection algorithm, allowing it to adapt to individual patient variations and stress conditions, thereby reducing false negatives and false positives while maintaining manageable complexity through automated adjustment rather than complex manual configuration.

The patent applies parameter changes by dynamically adjusting detection thresholds, time windows, and amplitude criteria based on the patient's baseline characteristics stored in the normal QRS templates. Instead of using fixed detection parameters, the system modifies these parameters according to the individual patient's normal ECG patterns, improving reliability by adapting to patient-specific variations while keeping the overall algorithm structure relatively simple.