Body mass index and the relationship between QRS amplitude and left ventricular mass in hypertensive patients and patients with metabolic syndrome

Abstract

minutes). A statistical classifier was developed to detect AF based on RR interval irregularity and inconsistency in P-wave location and morphology. The algorithm was trained using 633 ambulatory Holter recordings. The 48 prospective recordings were used only for testing. The algorithm determined if the patient had any AF (primary diagnostic) and quantified the amount of AF (monitoring). A patient was assumed to have AF burden if more than 10 minutes per day of the ECG was AF. For ECGs not 24 hours long, this is equivalent to being in AF for more than 0.69% of the record length. Results: Atrial fibrillation burden was detected on all 10 patient records (sensitivity of 100%) with 9 of 10 AF burden values within 10% of the true value. Average duration sensitivity and specificity, which represent the ability to measure true AF and reject non-AF periods, were 98.8% and 99.7%, respectively. Conclusion: We have described a highly sensitive and accurate method for the assessment of AF burden using a miniaturized prototype ECG recorder with a novel AF detection algorithm. Using this device potentially increases patient comfort and compliance for long-term AF monitoring compared to Holter systems. It may be possible to enhance the algorithm further by finding the optimum location for the device placement and using devicespecific data to train the algorithm.