Artificial Intelligence (AI) is poised to revolutionize healthcare like never before. It has the potential to bring about unprecedented innovations in medical care with accelerated progress. In the field of cardiac surgery, AI is being increasingly utilised to enhance surgical care through technological remodeling and cognitive augmentation. With more robust infrastructures and information technology wings, automation and machine learning are bound to metamorphose clinical care and healthcare administration by eliminating human errors, time constraints, manpower shortages, and financial hindrances. This article provides an overview of the various applications of AI in cardiac surgery and its astounding impact.
With a technological revolution engulfing the world in all domains, the healthcare sector has slowly, steadily, concretely and compellingly joined the bandwagon. Cardiac surgery is a high-risk, high-octane, resource-intensive, sophisticated speciality that requires exceptional surgical prowess and astute cognitive skills. AI has been effectively incorporated into cardiac surgery to facilitate dynamic clinical decisions through innovative risk stratification tools. AI has significantly impacted areas such as disease prevention/well-being, early detection, diagnosis, surgical decision-making, surgical techniques, prognosis, research, and chronic care management, contributing to enhanced and efficient patient care.
AI, as the name indicates, refers to the intellectual capacity that is artificially created for machines to learn from crowd sourced data sets that can be processed at incredibly rapid times to provide lightning-fast inferences. It is the science of making machines simulate human intelligence to enable problem-solving by recognising patterns for decision-making, thereby reducing the margin for error.
Heart failure (HF) is a deadly dangerous disease with high morbidity and mortality, making it one of Europe's leading causes of death. Preventive care is essential to help patients circumvent the onset of illness, slow disease progression, and reduce the chance of developing severe complications. AI is reshaping the way heart failure is anticipated and prevented. AI-driven preventive care has the potential to extend life expectancy and improve the well-being of patients with chronic conditions like diabetes and hypertension, which are significant risk factors for heart disease. AI algorithms improve clinicians' comprehension, judgment, and decision-making for each patient. Predictive and prescriptive analytics strengthen disease prevention and support positive outcomes.
AI has a vital application in the early detection of heart diseases. In the terminal stages of HF, when biventricular dysfunction sets in, a patient may quickly progress to kidney, liver, and sometimes neurological dysfunction due to low cardiac output, ultimately leading to multi-organ failure. AI is increasingly being employed by clinicians to diagnose HF early, enabling the prompt initiation of treatment. Abnormal heart sounds and murmurs are classical features of heart disease. However, identifying them is highly subjective and depends on the clinician’s level of experience and expertise. A novel HF screening framework based on the Gated Recurrent Unit (GRU) has been developed by Chinese researchers. The GRU model is a deep learning model that can acquire features of heart sounds directly, process signals, and generate inferences on HF. It is a promising tool for non-invasive screening of HF and is paving the way for early detection of coronary collapse. In another advancement, physician-scientists from Northwestern Medicine, have developed an algorithm feeding in signs that can detect early HF. the algorithm runs through the hospital-registered patient database to pick up these signs from their laboratory tests or physician notes during their visit, thus identifying potential HF patients. Based on the flags from this computer model, the patient is contacted and evaluated for HF. The hospital reports that out of 2500 patients that were recognised and reached out, 500 patients have been diagnosed with HF with the use of AI and initiated treatment. This model has helped identify candidates for heart transplantation and enrol patients in advanced therapies through clinical trials, thus improving their quality of life and longevity. Due to the potential sensitivity of AI tests in detecting disease early and providing deep phenotypes, it may appear to predict future diseases by creating a class of ‘previvors’ who have not yet experienced a disease.
An electrocardiogram (ECG) is a standard, ubiquitous test used in clinical workflows for a long time and is widely available for the diagnosis of heart disease. However, with the emergence of AI ECG, the diagnosis of heart disease has been revolutionized. With the help of massive labelled datasets, complex neural networks have been created to uncover subtle patterns in ECGs that even expert ECG interpreters may have missed. These neural networks bring new diagnostic power and value to the ECG to identify low ejection fraction of the heart, propensity for atrial fibrillation, hypertrophic cardiomyopathy, hyperkalemia, medical comorbidity/frailty, age, sex, markers of valvular heart disease, amyloidosis, etc. FDA approval has been granted to AI-based tools for rhythm detection. The ‘EAGLE’ trial, a large cluster design pragmatic trial, is currently underway at the Mayo Clinic to evaluate the effectiveness of 12-lead ECG in identifying left ventricular dysfunction, which is asymptomatically present in 3% to 9% of the general population. The AI ECG has also been embedded into a stethoscope form with embedded electrodes, by Texas Heart Institute clinicians, to record ECGs during routine clinical examination. Additionally, it can be used to monitor patients at risk of ventricular dysfunction, such as those undergoing chemotherapy or heart transplant patients, in the comfort of their homes at a lower cost.
The use of AI in Echocardiography, a crucial diagnostic tool for heart failure, has led to the development of convolutional neural networks that can determine ejection fraction and longitudinal strain. These networks can also detect pulmonary arterial hypertension, cardiac amyloidosis, and hypertrophic cardiomyopathy.
AI is incredibly helpful in surgical decision-making, providing explicit predictive models and risk assessment tools created by collating information from diverse sources about patient risk factors, anatomical features, and natural history of the disease, patient values, and costs. AI-enhanced risk assessment models have proved to be more accurate than any surgical risk score predictor.
Ex Vivo Lung Perfusion (EVLP) is a cutting-edge technology in lung transplantation that involves continuous machine perfusion of marginal donor lungs after procurement, to optimize and test functionality before implanting them into the recipient. Clinician-scientists at Toronto have used AI in developing an Extreme Gradient Boosting model (XGBoost), named InsighTX, which has been constructed from donor features along with biological, physiological, and biochemical parameters assessed during EVLP. The InsighTX model, with a machine learning approach, has proven to be accurate in predicting lungs unsuitable for transplantation after an EVLP run. This is an excellent aid in clinical decision-making for thoracic surgeons using the EVLP technology. Another centre has created a convolutional neural network that pools thousands of chest radiographs from EVLP lungs. For every neural network, a standard model was adapted to detect atypical findings on the chest radiograph such as atelectasis, infiltration, consolidation, interstitial lines, nodules etc. This automates EVLP radiograph image processing with accurate analysis.
Advancements in Magnetic Resonance Imaging (MRI) technology, powered by AI, are having a groundbreaking impact on surgical decision-making for complex cardiac conditions. By uploading MRI scan images into Virtual Reality (VR) software, precision VR images can be generated to help in unparalleled surgical planning. The MRI-VR technique provides surgeons with additional insight and enables a more effective game plan, including a 3-D trial surgery, before beginning a complex invasive procedure.
The impact of AI on surgical techniques has been transformative. Robotic heart surgery and robotic-assisted thoracic surgery are now commonplace. Surgeons use a computer-enhanced robotic system to perform surgery through minimal access. The surgeon typically sits in a console, similar to a video game, and controls surgical instruments attached to robotic arms that mimic the human hand, wrist, and finger movement. This increases the range and precision of movements. From coronary bypass surgery, valve repair and replacement, closure of holes in the heart to cardiac mass, lung cancer, and mediastinal tumour removal are being performed robotically. In such complex surgical procedures, AI leveraging machine learning, computer vision, and robotics can significantly help surgeons perform intricate surgeries with greater success, lesser complications, and upholding patient safety.
Interestingly, AI has also improved surgical training simulators that can assess the participant’s performance and offer customized feedback to enhance and refine surgical skills.
Prognosticating outcomes is an important aspect of cardio-thoracic surgery, to weigh the pros and cons, the risks and benefits for better judgement and informed decision making. End-stage HF patients may often decompensate while on the wait list for a suitable donor heart and may require a form of mechanical circulatory support to sustain life till the time an organ becomes available. There are various options available but the kind of support and its outcome is often unpredictable and depends on multiple factors. Researchers in Switzerland have effectively used AI in developing an interpretable machine learning model based on a large clinical database that can effectively predict one-year mortality in patients with temporary mechanical circulatory support as a bridge to cardiac transplantation. Another gradient-boosted classification algorithm model (GBM) developed by scientists at Mayo Clinic, helps in accurately predicting graft failure and five-year mortality after orthotropic heart transplantation. Machine learning has also been proven effective in projecting one-year mortality after heart and lung transplantation.
Another pioneering application of AI is in chronic care management, which plays a crucial role in determining the outcome and quality of life of the sick subset of patients undergoing heart and lung transplantation. One of the most common complications post-transplant is rejection. Physician-scientists in Boston successfully integrated AI into a novel deep-learning model to detect Acute Cellular Rejection in heart and lung transplant biopsies. With digital pathology and AI technology, the machine learning algorithm can differentiate the vascular component of rejection in transplant biopsies from normal tissue and thus can detect, classify and grade transplant rejections. AI has also been used to predict post-heart transplant graft function, re-hospitalisation, re-transplantation, graft survival, cardiac allograft vasculopathy, and model blood levels of immunosuppressive medications.
The care of heart failure patients needs a paradigm shift from reactive to predictive, preventive and personalized care.16 Active participation of patients in their care processes is mandatory to reduce the burden of HF on healthcare labour and costs. PASSION‐HF consortium (PAtient Self‐care uSing eHealth In chrONic Heart Failure) is a collaborative initiative by physician-researchers in the Netherlands and the UK. They use AI to develop a ‘virtual doctor’ at home for self-care using digital therapeutics. This virtual doctor will follow HF guidelines, consider patient co-morbidities, and ensure safe prescribing and handling of medications. It will include an AI-powered decision support engine, an interactive physician avatar interface, serious gaming tools, a self-learning feedback system, and patient coaching.
AI and machine learning also hold the promise of positively impacting research, drug discovery and development, as researched by Indian scientists. They can enable leveraging human datasets leading to a better understanding of target biology. AI can use real-world human data to generate new insights and translate them into potential therapeutics for the benefit of patients suffering from cardio-metabolic conditions.
A new robot named Moxi is unpretentiously traversing through the hallways of Northwestern Memorial Hospital! Moxi is designed to deliver items such as laboratory specimens and medications between the lab, pharmacy and floor. It also helps with the stocking of medication and supplies. The hospital believes that automating these tasks will help free up time for Laboratory services and Pharmacy team members to better focus on tasks that they are licensed to perform. Moxi uses an AI system to navigate its way through the hospital and also has a robotic arm to perform tasks such as pressing elevator buttons.
In conclusion, AI is undeniably reshaping the healthcare landscape by refining clinical practice for better patient outcomes. And it isn’t about the future, it’s the direction the healthcare is headed right now. However, there are numerous challenges in the implementation of AI in healthcare such as accessibility, equity, cost and resource allocation, validation, regulation, and ethical dilemmas. While AI can provide unprecedented levels of data analysis and pattern recognition, thus minimising errors, the depth, empathy and nuanced understanding that come with human interactions are irreplaceable. AI cannot replace but complement and enhance human intelligence to an extent never conceived before.
1. Kilic A. Artificial Intelligence and Machine Learning in Cardiovascular Health Care. Ann Thorac Surg 2020; 109:1323-9.
2. Gao S, Zheng Y, Guo X. Gated recurrent unit-based heart sound analysis for heart failure screening. Biomed Eng Online 2020; 19(1):3.
3. Cheema B, Mutharasan R, Sharma A, et al. Augmented Intelligence to Identify Patients With Advanced Heart Failure in an Integrated Health System. JACC Adv 2022;1(4):100-123.
4. Sionti KC, Noseworthy PA, Attia ZI, et al. Artificial intelligence-enhanced electrocardiography in cardiovascular disease management. Nat Rev Cardiol 2021; 18:465–478.
5. Xiaoxi Yao, Rozalina G. McCoy, et al. Noseworthy. ECG AI-Guided Screening for Low Ejection Fraction (EAGLE): Rationale and design of a pragmatic cluster randomized trial. American Heart Journal 2020; 219: 31-36.
6. Attia ZI, Dugan J, Rideout A, et al. Automated detection of low ejection fraction from a one-lead electrocardiogram: application of an AI algorithm to an electrocardiogram-enabled Digital Stethoscope. Eur Heart J Digit Health 2022; 3(3):373-379.
7. Coulter SA, Campos K. Artificial Intelligence in Echocardiography. Tex Heart Inst J. 2022 Mar 1;49(2):e217671
8. Sage AT, Shamandy AA, Mousavi SH, et al. InsighTx: A Machine-Learning Model That Accurately Predicts Transplant Outcomes During Ex Vivo Lung Perfusion. The Journal of Heart and Lung Transplantation 2022; 41(4): S15.
9. Sadeghi AH, Mathari S, Abjigitova D, et al. Current and Future Applications of Virtual, Augmented, and Mixed Reality in Cardiothoracic Surgery. Ann Thorac Surg 2022; 113:681-91.
10. Harky A, Hussain SMA. Robotic Cardiac Surgery: The Future Gold Standard or An Unnecessary Extravagance? Braz J Cardiovasc Surg. 2019; 34(4):XII-XIII.
11. Kim S, Bharat A. Video-assisted thoracoscopic surgery versus robotic-assisted surgery: Are we asking the right question? J Thorac Cardiovasc Surg 2020; 160 (5):1374-1375.
12. Shou BL, Chatterjee D, Russel JW, et al. Pre-Operative Machine Learning for Heart Transplant Patients Bridged with Temporary Mechanical Circulatory Support. J Cardiovasc Dev Dis 2022; 9:311.
13. Agasthi P, Smith SD, Murphy KM, et al. Artificial Intelligence Helps Predict 5-year Mortality and Graft Failure in Patients Undergoing Orthotopic Heart Transplantation. J Heart Lung Transplant, 39(4), S142.
14. Lipkova J, Chen TY, Lu MY, et al. Deep learning-enabled assessment of cardiac allograft rejection from endomyocardial biopsies. Nat Med. 2022 Mar;28(3):575-582.
15. Al-Ani MA, Bai C, Hashky A, et al. Artificial intelligence guidance of advanced heart failure therapies: A systematic scoping review. Front Cardiovasc Med 2023; 10:1127716.
16. Barrett M, Boyne J, Brandts J, et al. Artificial intelligence supported patient self-care in chronic heart failure: a paradigm shift from reactive to predictive, preventive and personalised care. EPMA J 2019; 10(4):445-464.
17. Ski CF, Thompson DR, Brunner-La Rocca HP. Putting AI at the centre of heart failure care. ESC Heart Fail 2020; 7(5):3257-3258.
18. Sarkar C, Das B, Rawat VS, Wahlang JB, et al. Artificial Intelligence and Machine Learning Technology Driven Modern Drug Discovery and Development. Int J Mol Sci. 2023; 24(3):2026.