Photo Credit: Wildpixel
Researchers have developed an AI model using surface ECG recording and atrial electrograms to detect AF.
Researchers developed an AI model using surface ECG recording and atrial electrograms to detect atrial fibrillation (AF) in patients recovering from heart surgery, according to findings published in the Journal of the American Heart Association.
“The use of epicardial atrial electrograms and surface ECGs in combination offer a novel means for determining a true atrial fibrillation diagnosis in postoperative cardiac patients, thereby permitting the creation of a robust AI tool for detection of AF during surface ECG monitoring following cardiac surgery,” Yuji Zhang, MD, and colleagues wrote. “AF detection in postoperative patients is important to prevent AF‐related complications but is currently labor intensive; artificial intelligence offers the potential to detect new postoperative AF more efficiently and at lower cost.”
Dr. Zhang and colleagues used atrial electrograms to identify “true” AF and to develop an AI model for detecting AF, using data from adults who were indicated for surgery but otherwise had “routine” operations. Trained expert ECG readers reviewed 5,000,000 epochs of 30 seconds apiece from 329 patients, annotating the readings as showing AF or the absence of AF. They used atrial electrograms to establish whether the given epochs showed true AF and trained the AI model to identify AF using only ECG records. Dr. Zhang and colleagues also used another 5,000,000 30-second epochs from 330 patients to test the AI model. The researchers tested the model’s performance at both the epoch and patient levels to determine AF burden.
The model appeared robust, attaining an area under the receiver operating characteristic curve of 0.932 upon validation. When tested, it achieved an area under the receiver operating characteristic curve of 0.952, the researchers report. Dr. Zhang and colleagues report that the model also performed well at the epoch level in terms of specificity, sensitivity, positive and negative predictive value, and harmonic mean of positive predictive value and sensitivity.
At the patient level, AF burden positive predictivity was calculated at 94.5%, whereas sensitivity was 96.2%.
Interpreting the Data
“Screening for AF in high‐risk populations may improve health care and reduce costs of care. However, detecting AF is currently labor‐intensive, requiring review of prolonged ECG monitoring,” the researchers conclude. “AI may facilitate AF detection but currently presents unresolved challenges in terms of diagnostic accuracy. In this study, a novel method employing atrial electrograms in conjunction with ECG recordings was used to develop a robust AI tool for AF recognition and AF burden determination. Such a method, while designed for postoperative cardiac patients, has the potential to be of diagnostic utility in large populations of community‐dwelling at‐risk individuals such as post-surgery patients, individuals with structural heart disease, and the elderly.”
Future Testing and Use of AI Model in AF
Dr. Zhang and colleagues plan to have the model applied to multiple treatment centers, where they hope to ascertain the tool’s value in diagnosing and treating AF.
“Inevitably, clinicians will need reliable computer‐aided ECG diagnostic analysis to not only identify AF but also quantify durations of continuous AF occurrences and overall AF burden for patients,” Dr. Zhang and colleagues wrote. “AI systems may prove helpful in this regard, but their development demands a well‐curated and structured AF database for AI‐based tool training and testing.”