Robust Artificial Intelligence Tool for Atrial Fibrillation Diagnosis: Novel Development Approach Incorporating Both Atrial Electrograms and Surface ECG and Evaluation by Head-to-Head Comparison With Hospital-Based Physician ECG Readers.

BACKGROUND: Atrial fibrillation (AF) increases risk of embolic stroke, and in postoperative patients, increases cost of care. Consequently, ECG screening for AF in high-risk patients is important but labor-intensive. Artificial intelligence (AI) may reduce AF detection workload, but AI development presents challenges. METHODS AND RESULTS: We used a novel approach to AI development for AF detection using both surface ECG recordings and atrial epicardial electrograms obtained in postoperative cardiac patients. Atrial electrograms were used only to facilitate establishing true AF for AI development; this permitted the establishment of an AI-based tool for subsequent AF detection using ECG records alone. A total of 5 million 30-second epochs from 329 patients were annotated as AF or non-AF by expert ECG readers for AI training and validation, while 5 million 30-second epochs from 330 different patients were used for AI testing. AI performance was assessed at the epoch level as well as AF burden at the patient level. AI achieved an area under the receiver operating characteristic curve of 0.932 on validation and 0.953 on testing. At the epoch level, testing results showed means of AF detection sensitivity, specificity, negative predictive value, positive predictive value, and F1 (harmonic mean of positive predictive value and sensitivity) as 0.970, 0.814, 0.976, 0.776, and 0.862, respectively, while the intraclass correlation coefficient for AF burden detection was 0.952. At the patient level, AF burden sensitivity and positive predictivity were 96.2% and 94.5%, respectively. CONCLUSIONS: Use of both atrial electrograms and surface ECG permitted development of a robust AI-based approach to postoperative AF recognition and AF burden assessment. This novel tool may enhance detection and management of AF, particularly in patients following operative cardiac surgery.

Heart rate variability, sleep and sleep disorders.

Heart rate (HR) is modulated by the combined effects of the sympathetic and parasympathetic nervous systems. Therefore, measurement of changes in HR over time (heart rate variability or HRV) provides information about autonomic functioning. HRV has been used to identify high risk people, understand the autonomic components of different disorders and to evaluate the effect of different interventions, etc. Since the signal required to measure HRV is already being collected on the electrocardiogram (ECG) channel of the polysomnogram (PSG), collecting data for research on HRV and sleep is straightforward, but applications have been limited. As reviewed here, HRV has been applied to understand autonomic changes during different sleep stages. It has also been applied to understand the effect of sleep-disordered breathing, periodic limb movements and insomnia both during sleep and during the daytime. HRV has been successfully used to screen people for possible referral to a Sleep Lab. It has also been used to monitor the effects of continuous positive airway pressure (CPAP). A novel HRV measure, cardiopulmonary coupling (CPC) has been proposed for sleep quality. Evidence also suggests that HRV collected during a PSG can be used in risk stratification models, at least for older adults. Caveats for accurate interpretation of HRV are also presented.

Career development of biomedical engineers in medical device industry.

With concerns of the current health care system, biomedical engineers have expertise, opportunity and responsibility in developing innovations that may improve cost, coverage and quality of health care delivery. This paper reviews the product development process in the medical device industry, and the associated training and experience required for biomedical engineers involved at each stage of the process. This paper also provides personal perspectives of some of the differences between established device companies and start-ups in the product development process and career paths for biomedical engineers.

Analysis of interaction of acute atrial overdrive pacing with sleep-related breathing disorder.

Dose-related effects of atrial overdrive pacing (AOP) on sleep-related breathing disorder (SRBD) were studied. Fourteen patients with pacemakers with moderate to severe SRBD (mean screening apnea-hypopnea index [AHI] 35.2 +/- 21.9 events/hour) were randomized to 3 levels of pacing (50, 10, and 20 beats/min greater than the mean nocturnal heart rate) and studied by polysomnography, observing for changes in AHI. At the 2 AOP levels, no significant change was observed in the primary end point of reduction in AHI. Additionally, there was no observed impact on secondary end points of the study. Cyclic variation of heart rate was progressively abolished with higher levels of AOP without affecting AHI. Large variations were observed between the screening and control studies in SRBD indexes in a number of patients. In conclusion, AOP demonstrated no benefit to predominantly obstructive SRBD disorder of at least moderate severity.

Nocturnal cardio-respiratory indices - a novel screening tool for pediatric obstructive sleep disordered breathing.

Obstructive sleep apnea (OSA) is a common breathing disorder among children with a prevalence estimated at 2-4%. Electrocardiography (ECG) based signal processing, especially heart rate analysis, demonstrated its promising future as both an alternative to other expensive portable sleep study methods and additional approach to sleep cardio-respiratory autonomic interactions which the current clinical standard polysomnography (PSG) lacks. A novel strategy was applied on pediatric OSA detection using RR intervals. A sensitivity of 89%, specificity of 96%, positive predictive value of 89%, and negative predictive value of 96% was achieved in case-to-case detection using 37 clinical pediatric full PSG data.

Comparison of R-wave detection errors of four wireless heart rate belts in the presence of noise.

Four commercial wireless chest belts (Vetta, Nashbar, Polar and new Polar) were assessed for their susceptibility to noises in R-wave detection. A normal ECG signal was generated using a LionHeart simulator (Bio-tek Instruments, Inc.) with a fixed RR interval (750 ms) and R-wave amplitude (1 mV). Different levels of EMG and baseline wanderings (sinusoidal waves) were recorded from a healthy subject and a Quartec Model MFG-1 generator, respectively. They were added to the ECG signal in a BioPac system (BioPac systems Inc., Santa Barbara, CA) to simulate an ECG in physiological noise. The BioPac system applied the 'contaminated' ECG to the belts via a voltage divider. A PC-based Polar Precision Performance system was used to receive the detected R-wave pulses transmitted from the wireless belts and to calculate the RR intervals. Two types of detection errors were observed in the RR intervals: small time shifts, the potentially non-fixable small variance, and missed/false beats, the abnormally large and potentially fixable intervals. Results showed that small time shifts exist in all tests ranging from -10 ms to 10 ms and increase with the level of EMG before missed/false beats occur. Missed/false beats occur only when EMG level is beyond the threshold of 0.4 mV, 1.6 mV, 1.2 mV and 1.2 mV for Vetta, Nashbar, Polar and new Polar, respectively. The potential to detect and fix EMG introduced missed/false beats showed that this type of error could only be improved when the added EMG was below a certain value. Results also showed that no missed/false beats occur when the frequency and amplitude of sinusoidal waves were below 1 Hz and 5 mV.