Artificial intelligence in medical practice: current status and future perspectives : Cardiology Plus

Secondary Logo

Journal Logo


Artificial intelligence in medical practice: current status and future perspectives

Dai, Yuxiang1,2; Ge, Junbo1,2,*

Author Information
Cardiology Plus 8(1):p 1-3, January-March 2023. | DOI: 10.1097/CP9.0000000000000040
  • Open

Artificial intelligence (AI) is rapidly evolving due to fast-pace advances in computing power, big data, and machine learning algorithms[1,2]. ChatGPT is a representative of AI use and has attracted attention and enthusiastic discussion around the world[3,4]. AI has transformed the way of our life in many ways, and the impact is likely to continue growing in the coming years. In this article, we summarize the basic logics and elements of AI medicine and potential use of ChatGPT in research as well as in clinical practice.


In essence, AI is the use of computer science in solving real-world problems, and includes robotics, speech recognition, image recognition, natural language processing, and expert systems[5]. The key elements of AI are mathematics, statistics, and computer science. Machine learning, the core technology of AI, is based on a variety of statistical methods, for example probability theory, statistics, and approximation theory[6,7]. For AI medicine, commonly used statistical and mathematical methods include regression analysis, generalized linear models, iteration and approximation, intra-sample modeling, and external validation. The three elements that ultimately determine the success (of failure) of AI medicine are algorithms (eg, logic, functions, and mathematical models), computing power, and databases[2].

ChatGPT is a well-known representative product of AI. In essence, ChatGPT is an AI model for natural language processing, and features a number of functions, including chatting, creation, and code writing[8]. GPT, short for Generative Pretrained Transformer, is distinct from previous search engines and can generate new content on its own[8]. The term “pretrained” in the full name refers to the fact that it has been trained in some databases before and already has some self-logic and judgment. Transformer is a new algorithm architecture efficient in processing long texts and requires relatively short training time. ChatGPT is the product of an immense engineering project that integrates algorithm, computing power, and database resources[9].

AI medicine faces a unique set of challenges in China. Limited computing power due to reliance on imported computer chips is the first obstacle. Towards this end, China has made huge investment during the past few years to develop the ability to mass produce high-end computer chips. Lack of human resources with background in mathematical algorithms represent another barrier that requires long-term efforts at a strategic level. Limited number of high-quality databases is the third obstacle despite of recent progresses. Accessibility to information on website also represents an issue. Despite of these challenges, we believe AI medicine will profoundly change the landscape of medical care system in China in the very near future.


With the help of AI technology, medical professionals can process and analyze large amounts of data more accurately and efficiently, which in turn may translate into more accurate diagnosis and more effective treatment[10]. AI could not replace physicians for several reasons.

First, daily practice of physician is based on information that could be processed with AI, but often require judgment based on past experience. Physicians must take into account a variety of issues, including the personal needs, value system, and socioeconomic status, when making decisions in individual patients. Second, many of the clinical features (eg, pain, fatigue, itching, and drowsiness) have strong subjective component and are often difficult to quantify accurately. Third, clinical features and lab tests vary significantly even in the same patients. For example, the 6-minute walk test requires patients to walk as far as possible for 6 minutes. Patient performance vary significantly across multiple sessions due to distinct physical conditions in each session and difference in physician instruction. Ultrasound evaluation of left ventricular ejection fraction (LVEF) of the same patients could yield inconsistent results across different physicians. In addition, test accuracy is also affected by a variety of factors, including valve regurgitation, heart size, and imaging clarity. As a matter of fact, medicine is a fuzzy and probability science. In practice, even the simple surgeries are not 100% free of complications, and even the most skilled experts cannot ensure 100% accuracy in diagnosis[11]. Adding to the issue is the complex interaction between clinical features and lab tests, which is very challenging for AI.

Lack of large databases with high reliability is another key challenge in AI medicine. Many of the clinical variables are inherently ambiguous and subjective. Additionally, ethic issues must be considered, and patient privacy must be respected. Limited staffing and human resource also represent a major barrier in establishing high-quality database. Last but not least is the willingness to collaborate among different physicians and hospitals, which in turn requires fair and transparent operation mechanisms.

AI medicine represents a major advance in medicine and allows individual physicians to tackle into the collective experience and wisdom of the entire medical society to make more accurate diagnosis and more prudent treatment decisions with much higher efficiency[12]. By no means physicians will be replaced by “AI doctors.” The failure of some early pioneering efforts, such as IBM Watson and Google’s artificial intelligence Streams, is a testament.


AI has already being used in a huge collection of scenarios in medical practice. The following is a brief summary of current status and perspective for the near future.

The most exciting application of AI in medicine is the use of machine learning algorithms to analyze imaging data [X-rays, computed tomography (CT) scans, magnetic resonance imaging (MRIs), and pathological sections][13]. AI algorithms can collect and analyze imaging data in a way that is not possible by human beings[14]. The analysis of pathological sections by AI is helpful in identifying disease patterns and making treatment decisions[15]. Indeed, sufficiently trained AI models could, in certain circumstances, achieve superior accuracy in diagnosis than average physicians. Invasive diagnostic procedures, thus, could be avoided.

AI has been extensively used in drug discovery. Spatial structure interactions and correlations between different molecules based on structure databases have allowed efficient drug design[16]. Mechanism simulation and virtual binding of lead compounds to putative molecular targets have been used to predict in vitro activity, and thus greatly expedite the development process with lower cost[17]. The use of AlphaFold (DeepMind Inc, London, United Kingdom) in solving the protein spatial structure prediction problem is one of the many examples of success story[18–20].

AI can also be used as virtual organs to predict effectiveness of candidate drugs and medical devices in human subjects. Computer simulation of human trials is at the infancy stage, but holds great potential in reducing the time and cost of preliminary evaluations and mitigating ethic concerns in using human volunteers. Digital twin collects biological information from an individual patient to build a digital person that conforms to their health characteristics and predicts the treatment response in the real patient[21].

Furthermore, AI can also be used to improve the accuracy and efficiency of diagnosis and treatment through patient self-monitoring and personalized medicine. For example, AI-powered chatbots can help patients to self-diagnose and manage their conditions, whereas virtual assistants can help physicians to quickly find the needed information. Wearable medical devices collect vital signs, activity levels, and other metrics from individual patients for analysis using AI algorithms to identify patterns and make personalized treatment recommendations[22].


Despite of the recent advances, many obstacles must be overcome before eventual use of AI medicine in daily practice. The first obstacle is the lack of large, high-quality databases. A more important challenge is the lack of tools for standardization that allow consolidation and synthesis of the data from different sources. There are also many concerns in ethical issue and regulatory considerations. For example, measures that minimize disparity in access to healthcare must be built in AI algorithms. The legal ramification (eg, who is responsible for the decisions made by AI systems) is also a formidable barrier that requires the input from physicians, AI technology developers, regulators, as well as the general public.

In summary, AI medicine is currently in the very early stage, and lack sophistication in many ways. Nevertheless, significant changes have already been made, and we predict that AI medicine will radically transform the landscape of healthcare system worldwide. Significant challenges and formidable barriers exist, and it must be pointed out that AI could not replace the work of physicians.


Junbo Ge is the Editor-in-Chief of Cardiology Plus.


Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.


[1]. Rampton V. Artificial intelligence versus clinicians. BMJ 2020;369:m1326. doi:10.1136/bmj.m1326.
[2]. Zeitoun JD, Ravaud P. Artificial intelligence in health care: value for whom. Lancet Digit Health 2020;2:e338–e339. doi:10.1016/S2589-7500(20)30141-2.
[3]. Stokel-Walker C, Van Noorden R. What ChatGPT and generative AI mean for science. Nature 2023;614:214–216. doi:10.1038/d41586-023-00340-6.
[4]. van Dis E, Bollen J, Zuidema W, et al. ChatGPT: five priorities for research. Nature 2023;614:224–226. doi:10.1038/d41586-023-00288-7.
[5]. London AJ. Artificial intelligence in medicine: overcoming or recapitulating structural challenges to improving patient care. Cell Rep Med 2022;3:100622. doi:10.1016/j.xcrm.2022.100622.
[6]. Krittanawong C, Johnson KW, Rosenson RS, et al. Deep learning for cardiovascular medicine: a practical primer. Eur Heart J 2019;40:2058–2073. doi:10.1093/eurheartj/ehz056.
[7]. Al’Aref SJ, Anchouche K, Singh G, et al. Clinical applications of machine learning in cardiovascular disease and its relevance to cardiac imaging. Eur Heart J 2019;40:1975–1986. doi:10.1093/eurheartj/ehy404.
[8]. Doshi RH, Bajaj SS, Krumholz HM. ChatGPT: temptations of progress. Am J Bioeth 2023;1:1–3. doi:10.1080/15265161.2023.2180110.
[9]. Baumgartner C. The potential impact of ChatGPT in clinical and translational medicine. Clin Transl Med 2023;13:e1206. doi:10.1002/ctm2.1206.
[10]. Krittanawong C. The rise of artificial intelligence and the uncertain future for physicians. Eur J Intern Med 2018;48:e13–e14. doi:10.1016/j.ejim.2017.06.017.
[11]. Hashimoto DA, Rosman G, Rus D, et al. Artificial intelligence in surgery: promises and perils. Ann Surg 2018;268:70–76. doi:10.1097/SLA.0000000000002693.
[12]. Siontis KC, Noseworthy PA, Attia ZI, et al. Artificial intelligence-enhanced electrocardiography in cardiovascular disease management. Nat Rev Cardiol 2021;18:465–478. doi:10.1038/s41569-020-00503-2.
[13]. Sermesant M, Delingette H, Cochet H, et al. Applications of artificial intelligence in cardiovascular imaging. Nat Rev Cardiol 2021;18:600–609. doi:10.1038/s41569-021-00527-2.
[14]. Hosny A, Parmar C, Quackenbush J, et al. Artificial intelligence in radiology. Nat Rev Cancer 2018;18:500–510. doi:10.1038/s41568-018-0016-5.
[15]. Stenzinger A, Alber M, Allgäuer M, et al. Artificial intelligence and pathology: from principles to practice and future applications in histomorphology and molecular profiling. Semin Cancer Biol 2022;84:129–143. doi:10.1016/j.semcancer.2021.02.011.
[16]. Rogers MA, Aikawa E. Cardiovascular calcification: artificial intelligence and big data accelerate mechanistic discovery. Nat Rev Cardiol 2019;16:261–274. doi:10.1038/s41569-018-0123-8.
[17]. Díaz O, Dalton J, Giraldo J. Artificial intelligence: a novel approach for drug discovery. Trends Pharmacol Sci 2019;40:550–551. doi:10.1016/
[18]. Mullard A. What does AlphaFold mean for drug discovery. Nat Rev Drug Discov 2021;20:725–727. doi:10.1038/d41573-021-00161-0.
[19]. Callaway E. What’s next for AlphaFold and the AI protein-folding revolution. Nature 2022;604:234–238. doi:10.1038/d41586-022-00997-5.
[20]. Jumper J, Evans R, Pritzel A, et al. Highly accurate protein structure prediction with AlphaFold. Nature 2021;596:583–589. doi:10.1038/s41586-021-03819-2.
[21]. Corral-Acero J, Margara F, Marciniak M, et al. The “Digital Twin” to enable the vision of precision cardiology. Eur Heart J 2020;41:4556–4564. doi:10.1093/eurheartj/ehaa159.
[22]. Yu KH, Beam AL, Kohane IS. Artificial intelligence in healthcare. Nat Biomed Eng 2018;2:719–731. doi:10.1038/s41551-018-0305-z.
Copyright © 2023 China Heart House.