​Artificial Intelligence (AI) has changed various healthcare fields, with cardiovascular medicine benefiting as well. In this domain, AI helps tackle once-persistent challenges such as diagnosis accuracy and surgical safety and outcomes, as well as brings key improvements in different areas.

AI has streamlined cardiovascular disease and stroke imaging, largely circumventing past image processing and interpretation challenges. AI-powered tools make diagnosing and prognosis of cardiovascular conditions easier by automating abnormality detection, measuring plaque and blood flow, and predicting sudden cardiac death or stroke risk. AI also offers quick automated measurements for procedures, cutting down treatment time. Still, institutions using AI for medical imaging should be cognizant of the challenges tied to acquiring data and organizing and labeling it, as well as address ethics, bias, and data quality concerns.

Electrocardiography (ECG), a medical test tracking heart rhythm, has also changed with AI. With too few experts to interpret the growing number of ECGs, AI offers a solution by automating this task. Current ECG interpretation devices are rule-based and have limits. However, AI and machine learning (ML) can read ECGs more like human experts, leading to better care decisions. Sometimes, AI can read ECGs better than specialists, spotting hidden heart problems and finding conditions like aortic stenosis (narrowing of the heart's aortic valve opening), something traditional methods often miss.

When paired with photoplethysmography (PPG), а light-based method to track blood flow in the skin, AI can spot atrial fibrillation (AF), an irregular heartbeat. Smartwatches and other wearables with PPG can catch AF in real time, even outside clinical settings. By analyzing the timing and shape of the PPG signal, AI algorithms can differentiate between AF and a normal heart rhythm. This setup is more efficient than older methods like regression models and manual feature extraction, which lack precision. Complex AI models can also automatically find critical patterns in the PPG signal.

In hospitals, AI and machine learning boost bedside monitoring. For years, bedside monitoring has relied on pre-set rules to trigger alarms when a patient's vital signs exceed a specific range. But these systems often give false alarms and look at each vital sign in isolation, often missing the connections between them. AI-based tools can simultaneously watch all the data from monitors and see subtle patterns across many vital signs. This approach gives a fuller picture of patient health and reduces false alarm detection.

Research highlights ways AI improves hospital monitoring beyond sorting true from false alerts. For instance, AI models can spot small physiological changes indicating a patient is getting sepsis, where the body overreacts to infection. AI tools also help doctors act quickly by predicting when a patient might experience cardiac arrest in the hospital. Moreover, they assess surgery risks by weighing various patient factors before, during, and after procedures, predicting complications and mortality, and guiding anesthetic drug dosing.

AI helps researchers and doctors study heart disease genes to create better treatments. For example, genome-wide association studies (GWAS) use AI to scan the entire genome, spotting patterns and links between genes and specific heart conditions. AI and machine learning methods such as gradient boosting sharpen polygenic risk scores (PRS), which gauge a person's genetic odds of developing cardiovascular disease. Training AI on diverse, multi-ethnic genomic data makes these models work better for different groups. AI also connects observable traits, or phenotypes, to genes by analyzing facial images to detect genetic syndromes tied to heart issues.

Teamed with intelligent robots, AI boosts precision in complex surgeries while keeping them less invasive. The da Vinci Surgical System, for instance, enables robot-assisted surgeries, reducing risk and shortening hospital stays. For heart procedures like coronary interventions, AI robots let doctors work from a safe distance, avoiding radiation exposure.

​Computational Medicine in the Heart (CHIP) is a T32 Program of the National Institutes of Health (NIH) National Cancer Institute (NCI). The descriptor T32 applies to all NIH Ruth L. Kirschstein National Research Service Awards, which support doctoral research training in various areas.

CHIP T32 is an integrated program of the Stanford Cardiovascular Institute (CVI) of Stanford Medicine. The program supports training for post-doctoral fellows interested in using computational medicine to address unmet patient needs in cardiovascular (CV) diseases. CHIP T32 provides two years of training for six fellows who choose their mentors from 38 participating faculty members in the School of Medicine, School of Engineering, or School of Humanities and Sciences.

In addition to interdisciplinary research and coursework, CHIP T32 training comprises professional development and team-based work. Beyond the CVI and the specific schools listed above, content for CHIP T32 training comes from Stanford’s Institute for Computational and Mathematical Engineering and Artificial Intelligence in Medicine and Imaging.

​The healthcare industry continues to see development. Technology, such as machine learning (ML), provides useful ways to leverage patient data and help medical professionals deliver effective care.

Machine learning is a specific variant of artificial intelligence (AI) that enables systems to learn from data and detect patterns with minimal human intervention. AI helps computers and machines reason, learn, and act like human intelligence or sift through large data sets. Therefore, machine learning allows systems to analyze data, recognize patterns, and make decisions with minimal human input.

In healthcare, machine learning helps medical professionals extract meaning from complex medical information and patient data. Medical professionals have applied it in diagnostics as algorithms analyze medical images, such as X-rays, MRI scans, and CT scans, by detecting patterns that indicate specific diseases or conditions. These algorithms can support doctors by providing faster, more accurate diagnoses, ultimately leading to better patient outcomes.

Moreover, machine learning has aided drug discovery and development. ML models can identify previously unknown side effects or drug interactions, enhancing drug safety and efficacy. Pharmaceutical companies and healthcare organizations also use it to accelerate drug development and improve treatment effectiveness.

Lastly, machine learning can potentially reduce costs, streamline operations, improve data security and privacy, and enhance overall patient care. By automating routine tasks, optimizing workflows, and analyzing large datasets, healthcare providers can offer more personalized and efficient patient care.

​Recent years have witnessed notable shifts in healthcare delivery, with cardiovascular care significantly impacted by the increased use of digital health technologies. Digital health tools are now commonly employed in arrhythmia management, including diagnostics, long-term monitoring, patient education, shared decision-making, medication adherence, and research.

Despite these advances, integrating digital health technologies into cardiovascular care presents several challenges. Some of these challenges are patient usability, privacy, system interoperability, physician liability, analysis, and incorporation of a large amount of real-time information from wearables. Clear objectives and significant changes to existing workflows and responsibilities are necessary to successfully implement digital health tech in heart medicine.

Technological advances and improved manufacturing processes have facilitated the development of wearable and implantable monitors in cardiac care. These devices address an extensive scope of cardiovascular use cases, including heart rate and arrhythmia monitoring, heart failure and hemodynamic monitoring, physical activity, and general health and well-being.

Moreover, smartphone technologies have enabled varying levels of invasiveness in cardiac monitoring devices. Some are noninvasive and worn externally, while others are minimally invasive, requiring subcutaneous or intravascular implantation. Devices like wireless ECG watches, patch monitors, and cardiac defibrillators offer diverse options for monitoring heart health.

Heart rate and rhythm monitors are among the most widely used tools, providing multiple options for both patients and clinicians. A clear understanding of the features and limitations of the technology can enable the selection of appropriate tools based on patient symptoms and clinical needs. For instance, implantable loop recorders are useful for detecting subclinical atrial fibrillation (AF) to prevent cardioembolic stroke and other subclinical high-risk arrhythmias that may cause devastating complications.

​The utilization of artificial intelligence (AI) in the medical field is not a recent development. In the 1960s, early forms of AI, such as automatic risk calculators and signal annotation, were already being used. However, advances in computing power and increased access to digital data have significantly expanded the capabilities of AI in medical test interpretation.

Since early AI models were mostly image-based, it seemed natural to apply them in radiology. They were used for the automatic detection of pneumonia on a chest X-ray, organ segmentation in magnetic resonance scans and computed tomography, and tumor detection in images acquired from various sources.

In cardiology, machine-read electrocardiograms (ECGs) have been used for many years, but earlier algorithms were limited in accuracy and predictive capability. Regardless, these systems enabled researchers to collect copious volumes of digital data, ultimately facilitating the use of machine learning (ML) for ECG analysis.

Introducing AI-enhanced ECGs (or AI-ECGs) has improved automated rhythm reading and enabled the detection of a low ejection fraction using a 10-second ECG, detection of paroxysmal AF (atrial fibrillation) from an ECG recorded during sinus rhythm, and the determination of patient age and sex. In other cardiology fields, such as echocardiography and nuclear cardiology, AI is still in the early stages of application due to the complexity of the data and the reliance on human input during data acquisition.

​A team of investigators from the Smidt Heart Institute at Cedars-Sinai have developed an artificial intelligence (AI) algorithm that detects abnormal heart rhythms in individuals who do not show any symptoms. Atrial fibrillation (AFib) is an irregular and rapid heart rate that can increase the risk of stroke and other heart-related complications. According to the Centers for Disease Control and Prevention, close to 12.1 million people in the United States have AFib. A new study published in the journal npj Digital Medicine shows that the AI program/algorithm can detect abnormal heart rhythms that may go unnoticed during medical check-ups.

The study’s findings suggest AI can analyze images from an echocardiogram, a common imaging test that uses sound waves to capture images of the heart. In atrial fibrillation, electrical signals in the heart can malfunction. This causes blood in the upper chambers to accumulate and may form blood clots or travel to the brain causing an ischemic stroke. To create the algorithm, investigators designed an AI tool to study patterns from electrocardiogram readings. In an electrocardiogram, a medical professional places electrodes on the patient's body to monitor the heart’s electrical activity.

The AI tool analyzed electrocardiogram readings from patients visiting two Veterans Affairs health networks between January 1, 1987, and December 31, 2022. The algorithm, tested on nearly a million electrocardiograms, accurately predicted the likelihood of patients having AFib within 31 days. The AI model applied to patient medical records at Cedars-Sinai accurately predicted AFib cases within 31 days. According to David Ouyang, MD, a cardiologist in the Department of Cardiology at the Smidt Heart Institute at Cedars-Sinai, the research provided an effective way of identifying hidden heart conditions.

​Using artificial intelligence, scientists have found more effective ways to detect irregular heart rhythms using atrial fibrillation (AFib) mapping. The process has proven successful at generating optical maps of localized organized AFib sites in both animals and human beings. After identifying these sites, physicians can effectively treat them with minimally invasive ablation. Introduced into the body through a small incision, an ablation catheter can safely cause scarring on the inside of the heart, interrupting the electrical signals that lead to the irregular heartbeat.

Traditionally, surgeons have used both an ablation catheter and an AFib mapping catheter for this kind of AFib procedure. For left atrial access, this approach requires a double transseptal puncture, which creates two small surgical passages through the atrial septum of the heart. In the 2020s, surgeons began using a single transseptal puncture method that created a single surgical passage to include both an ablation catheter and a mapping catheter.

A recent study has shown that both the double and single transseptal puncture approach have favorable long-term patient outcomes. The single transseptal puncture approach offers several potential advantages, including less time under fluoroscopy x-rays and a reduced risk of acute complications.

​Machine learning algorithms are capable of analyzing electrocardiogram (ECG) data to identify arrhythmias and categorize heart disease. A new algorithm combining two-event-related moving averages (TERMA) and fractional Fourier transforms (FrFT) is outperforming state of the art algorithms in detecting heart rhythm peaks.

These two methods combined have achieved accurate detection of P, QRS, and T waves. In an electrocardiogram (ECG or EKG), the PQRST waves represent the heart's electrical activity, with the P wave triggering the beat, the QRS complex causing the ventricles to contract, and the T wave allowing the ventricles to recover and prepare for the next beat. The features of ECG signals can be used to train machine learning models to classify heart disease. The algorithm, with the help of different databases, detects the peaks of heart rhythms and classifies heart diseases.

The algorithm under consideration utilizes TERMA and FrFT to identify heart rhythm peaks effectively. TERMA identifies specific regions of interest to find the desired peaks, whereas FrFT rotates the ECG signals to reveal the peaks' positions. This fusion improves the algorithm by allowing it to detect P, QRS, and T waves with ease. The detection performance of the algorithm surpasses that of existing algorithms, proving the efficiency of the method used for ECG signal analysis.

The proposed algorithm extracts features such as PR and RT durations from the detected peak locations. Age, sex, and other features are used to train machine learning models such as MLP (multilayer perceptron, a type of neural network) and SVM (support vector machine, used for classification and regression analysis). These trained models can predict the presence of heart disease with high accuracy, demonstrating the applicability of machine learning to cardiovascular disease diagnosis and treatment.

​Atrial fibrillation (AFib), also known as irregular heartbeat, is a type of cardiac arrhythmia which causes both the lower and upper chambers of the heart to beat fast, erratically, and out of sync. Persistent atrial fibrillation is one of the main types of AFib, which occurs when an abnormal heart rhythm continues for more than one week. The heart rhythm can either return to normal on its own, or treatment may be required to regulate the heartbeat. Persistent AFib usually begins as short-term AFib or paroxysmal AFib. Some of the major risk factors associated with persistent AFib include old age, high blood pressure, and chronic pulmonary obstructive disease (COPD). Others include smoking, coronary heart disease, and heart valve disease.

Persistent AFib can be treated using medication to stabilize the heart’s rhythm or a procedure known as cardioversion. Cardioversion is performed to restore a normal heartbeat, using either electrical cardioversion or chemical cardioversion. In electrical cardioversion, a defibrillator using handheld paddles is placed on the chest and back to deliver shock to the heart in order to stabilize its rhythm. Chemical cardioversion involves the use of medication whereby drugs are either administered orally or through an IV to control the heart rhythm. The medication can take minutes, hours, or days to take effect.

​Electrocardiogram (ECG) deep learning models have proven to be capable of predicting heart failure in patients. The technology offers a potential solution for early intervention and improved patient outcomes. However, studies have shown disparities in the accuracy of these models based on the backgrounds of the patients. As this deep learning model machine becomes integrated into clinical practice, it is vital to address these disparities, which are the result of differences in patients’ age, sex, and race.

The primary model declined with increased patient age and performed significantly worse in Black patients from birth to 40 years of age when compared to other racial groups, with it being most pronounced among young Black women.

Despite efforts to address disparities in model performance by integrating race, sex, and age, the disparities persisted. However, using a customized probability threshold tailored to an individual's race, sex, and age led to an enhanced F1 score.

According to a report on the Stanford Medicine website, Dr. Sanjiv Narayan and other medical practitioners concluded that the results of this study highlight the risk of increasing existing health disparities through the development of these models for heart failure prognosis. To minimize this risk, practitioners can explore the use of personalized probability thresholds based on individual patient factors, leading to more inclusive and efficient healthcare solutions.