AI-based Algorithms That Predict Cognitive Decline
Smartphone-based applications may help detect atrial fibrillation, warn clinicians about worsening mental status, and improve the management of heart failure.
By John Halamka, M.D., president, Mayo Clinic Platform, and Paul Cerrato, senior research analyst and communications specialist, Mayo Clinic Platform
In a previous blog, we discussed the use of mobile devices that can help clinicians detect atrial fibrillation, which in turn increases the risk of stroke. New research suggests these smartphone-enabled devices, coupled with the right software, are also capable of predicting which patients are most likely to develop some form of cognitive decline.
Researchers from Mayo Clinic and AliveCor Inc. have used artificial intelligence (AI) to develop a mobile device that can identify certain patients at risk of several heart conditions, including atrial fibrillation (AF). The researchers determined that a smartphone-enabled mobile ECG device can rapidly and accurately determine a patient's QTc, thereby identifying patients at risk. In addition, the FDA allowed the use of KardiaMobile 6L to measure QTc in COVID-19 patients to monitor QT duration in patients receiving medications that can cause potentially life-threatening QT prolongation.
The European Journal of Internal Medicine has published a prospective, blinded evaluation of this AliveCor system for monitoring atrial fibrillation, comparing it to a standard 12-lead ECG and reported: “The AliveCor Kardia ECG monitor allows a highly accurate detection of atrial fibrillation by an interpreting electrophysiologist both in the standard lead I and a novel parasternal lead.” Similarly, a systematic review and meta-analysis of handheld ECG devices for atrial fibrillation screening concluded: “The pooled sensitivity and specificity of single-lead handheld ECG devices were high.”
More recently, Erika Weil, M.D., with the Department of Neurology at Mayo Clinic, and her colleagues, investigated the association between AI-based ECG readings and cognitive functioning. They looked at a data set of patients who had normal sinus rhythm readings on their AI-enabled ECGs and a subset of patients who had brain MRI testing. Their ECG readings were used to calculate an AI-ECG score between 0 and 1 to estimate their risk of AF.
Their analysis found a correlation between AI-ECG scores and lower baseline and cognitive scores, as well as a faster decline in these scores. It also suggested a link between AF and cerebral infarcts. The metrics used to assess patients’ cognitive skills, referred to as z scores, evaluated memory, attention, language, and visuospatial skills. The investigators concluded: “…most infarcts observed in our cohort were subcortical, suggesting that AI-ECG not only predicts AF but also detects other non-AF cardiac disease markers and correlates with small vessel cerebrovascular disease and cognitive decline.”
Weil et al. used a convolutional neural network (CNN) to power their AI-based ECG algorithm. Previous research documented its value in helping clinicians identify patients with AF during sinus rhythm. That earlier investigation examined over 180,000 patients with almost 640,000 normal sinus rhythm ECGs. Using single, AI-based ECGs, it was able to identify the arrhythmia with an area under the curve (AUC) rating of 0.87. (An AUC of 0.5 would indicate that the algorithm has no better than a 50/50 chance of identifying the existence of AF, an AUC of 1 indicates the model is 100% accurate).
The atrial fibrillation research is only one of many potentially useful applications of AI in cardiology. There is evidence to suggest that a single-lead smartphone platform may help detect acute coronary syndrome. The platform has been shown to assist in the rapid diagnosis of ST-segment elevation myocardial infarction (STEMI).
Francisco Lopez-Jimenez, MD, who leads AI at Mayo Clinic’s Department of Cardiovascular Medicine, and his colleagues, explain: “This technology could conceivably be widely disseminated and paired with ML interpretation to rapidly triage patients with STEMI. Expedited transfer to a percutaneous coronary intervention-capable institution could then be facilitated in a timelier manner…”
Similarly, AI based models may help improve the management of heart failure, playing a role in its prevention and reducing hospital readmissions.