Abstract
An academic letter by Rajesh N. and Sri Sai Srujana P. on ChatGPT (Chat-based Generative Pre-trained Transformer) left me pondering for days―Is Artificial Intelligence/Machine learning a boon for the cardiologists? These authors mentioned both sides of the new technology and cautioned us to go slow with it in academic cardiology. 1 In this digital era, we have to get conversant with a few terms like such as artificial intelligence (AI), machine learning (ML), and deep learning (DL).
Artificial Intelligence
In simple terms, it means the simulation of human intelligence processes by computer systems. This is developed using several programming algorithms like Python, R, Java, C++, and so on. Its impact on our lives is already a common experience like autocompletion of sentences as we start typing, web-search at lightning speed, virtual assistants, chatbots, Alexa, Siri, industrial robots, self-driving vehicles, and so on.
Machine Learning
It is an application of AI that includes parsing data, learning from the data, and then applying that when challenged to make informed decisions. ML can achieve several tasks; facilitate automation, risk stratification, prediction, quantification, and precision phenotyping. It also has a potential for false discovery and biases. These limitations can be minimized by active supervision during algorithm training and development.
Deep Learning
It is a type of ML but it structures algorithms in layers and creates an artificial neural network, which can help to make intelligent decisions on its own (Closer to human intelligence).
AI and ML can process large data and perform any task in less time than humans with fewer errors high consistency and endlessly without fatigue. They made their way into healthcare, education, business, transportation, banking, finance, law, media, entertainment, software coding, IT processes, security, manufacturing, and many other spheres. In the last few years, a lot of literature has been pouring in highlighting their surge in healthcare systems in general and into cardiovascular practice in particular. The wide use of electronic health records, wearable sensors, and modern imaging advances is facilitating the fast growth of AI in cardiology. 2
Diagnosis
AI made an entry into medical diagnostics through AI algorithms for reading chest X-rays, which the FDA had already approved. In cardiology, it invaded the many systems and machinery that have data from multiple sources and do complex calculations and positively impacted fields such as cardiovascular imaging, electrophysiology, heart failure, and interventional cardiology. Automation of reporting of ECGs in bulk and recorded Echo images can make life easier in high-volume centers and conquer fatigue in such jobs. We are going to see more AI-ECG embedded into stethoscopes. In echocardiogram section recognition, optimization, chamber volume, EF calculations, prognostication, and multiple other applications add to the accuracy of diagnosis. Image processing in nuclear cardiology using AI/ML-based algorithms has been validated. Similarly, CT and MRI reconstruction and image interpretations can be further refined in consistency and accuracy. Companies are keen to reduce the time takes to perform investigations and at lesser cost to the patient. IBM Watson is a successful healthcare technology that understands natural language and interacts when questions are asked. AI-based diagnosis can have an edge over traditional methods in case of rare diseases and atypical presentation of common diseases. 3 Pre-clinical diagnosis and primordial and primary prevention of deadly diseases may be made easy with this new technology.
Treatment
Robotic surgeries and interventional procedures are the brightest examples to understand the role of AI. Coronary angiogram image analysis, functional assessment of lesions, and decisions based on AI-ML algorithms fared as well as expert opinions and decisions. AI has the potential to predict stent size, and length, complex lesion characteristics, and likely future stent restenosis. 4 Detection and alerting capabilities in impending sudden cardiac death or exacerbation of HF have the potential to save more lives. AI-based algorithms are also validated in care for atrial fibrillation, detection of incident cardiomyopathy, electrolyte disturbances, drug levels in blood, and so on. Concepts of personalized treatment plans will be a reality with the advent of AI-ML.
Medical Education, Research, and Training
Chat-GPT performed with surprising accuracy in USMLE and other tests. Virtual training models make the procedures safe and more efficient. Data analysis, literature review, and manuscript writing have been made easy and a lot of time is saved. There is potential for biased results, misinformation, and distorted conclusions.
Medical Writing
The young medical writers have already begun using ChatGPT-based search, manuscript–writing and sending for peer review and publication. New challenges are coming up for the editors and publishers of scientific write-ups in the evaluation of submitted publications. In the future, we may have to declare if Chat-GPT was used. Is it necessary to cross-scrutinize such manuscripts manually by the reviewers? To what extent can we accept the AI language models?
Other Applications
It has potential applications in medical billing, hospital administration processes, prediction of pandemics, telemedicine, and remote monitoring. The quality of rural health care may show a sea change if AI technology is properly applied.
Concerns and Challenges
Machine learning bias is inherent in the current scenario. Can AI match human analytical abilities? Will this boom in AI cause healthcare personnel to lose jobs or make cardiologists superfluous? Future cardiologists may turn to managers and supervisors to cross-check the AI-based diagnosis and treatment decisions. It adds to the initial expense of upgrading any system. There is potential for misuse of AI applications towards unhealthy business gains.
Ethical Concerns
Ethical concerns like patient privacy, data security, liability for medical decisions and errors caused by AI, and such other ethical and legal issues are bound to surface as days go by. For patients looking for compassionate and caring physicians, acceptance of clinical decisions being made by AI-driven algorithms and machinery is questionable and unpredictable. Putting it simply is going to impact how we live, work, and play. Validation and prospective assessment of the greatly promising advantages is the need of the hour. As cardiologists, we have to keep our eyes and minds open to this new mind-blowing, and fast-growing development that is going to enter into our clinical practice in a big way.
