1. Introduction to Artificial Intelligence in Healthcare

Artificial Intelligence (AI) technologies such as machine learning, natural language processing, and computer vision have become ubiquitous. However, for the layperson, such technologies can seem enigmatic and beyond comprehension. These technologies have been developed over decades, dating back to the seminal work of pioneers like Alan Turing and John McCarthy in the 1950s on logical reasoning and symbol manipulation. The early years of AI with research in robotics, natural language understanding, and game-play, set the stage for failure in the late 1970s and were characterized by the “AI winters,” when research funding was cut back (Wójcik et al., 2023). In the mid-90s and early 2000s, the advances of connectionist methods, known as neural networks, brought renewed enthusiasm to AI. At the same time, the rise of the internet expanded human capability to record and store data on an unprecedented scale. “Big data” and an increased ability of computers to perform vector and matrix computations allowed new learning algorithms to train highly complex models, containing millions to billions of parameters. Today’s most prominent AI technologies, such as speech recognition, online translation, image recognition, and gameplaying, are based on deep learning (DL), a connectionist method.

Ultrasound technologies have been an area of extensive research for decades and in many applications, from health and biomedical devices to aerospace engineering. The recent advances in ultrasonic imaging modalities and new AI-based algorithms for image reconstruction and processing have allowed a large leap in imaging capabilities with new potential applications. AI encompasses diverse technologies that emulate “human intelligence.” Including machine and deep learning techniques these technologies can be used in applications such as predicting outcomes from patients using historical records, creating pictures using trained datasets, and understanding complex language text (Grezenko et al., 2023).

1.1. Overview of AI Technologies

Artificial Intelligence encompasses an array of actuarial modelling methods. These methods can be broadly classified into five main categories: symbolic reasoning, artificial neural networks, computer-based systems with explicit content, intelligent agents, and hybrid systems that combine elements from two or more of these categories (Grezenko et al., 2023). Symbolic reasoning systems typically take the form of rule-based expert systems, which represent domain knowledge in the form of content and decision rules, providing a high level of interpretability, control, and explainability, valuable for conditions of uncertainty (e.g., patient population characteristics or choice of treatments). Conversely, Artificial Neural Networks, a category derived from neurobiology, are mathematical models of the human brain consisting of neurones, interconnections, and weights. In this category, knowledge representation is more implicit, which can be a problem for application in crucial settings, such as Healthcare Delivery, where understanding decisions is paramount.

Computer-based systems with explicit content are not of great interest for the field of Healthcare Delivery, as they consist of mathematics software tools (e.g., Matlab) or general-purpose programming languages (e.g., C, Visual Basic). Intelligent agents can be defined as a system that senses and acts on its environment in pursuit of goals. These agents can perform actions either automatically or with a human decision, which is far from ideal in healthcare systems where an action or treatment could result in the death of the patient. Finally, hybrid systems, which combine different methods from the categories above, are a field of active research and development, as integrating the strengths of various individual methodologies may compensate for their weaknesses.

1.2. Historical Context and Evolution

Artificial Intelligence (AI) is an interdisciplinary scientific domain dedicated to understanding what constitutes “intelligence.” Intelligence refers to cognitive faculties such as reasoning, knowledge, learning, problem solving, perception, planning, ability to manipulate and change the environment, social intelligence, creativity, emotional intelligence, and communication ability, both in effectiveness and language . Human intelligence has been heightened by evolution and is currently kept active by studying the world, learning, and applying newly acquired knowledge to enhance survival. Recent advances in computational resources and data availability have led to applications of AI in diverse fields such as agriculture, transportation, construction, music, finance, and business, among others. The biomedical field has a high discovery potential with AI due to the rapid accumulation of medical data such as scientific publications, experimentation platforms, biospecimens, and clinical data. However, challenges specific to biomedicine, such as technical complexity and high expectations, need to be addressed for AI to be adopted widely by health organizations and care providers (Vandenberk et al., 2023). A comprehensive review describing how AI will change healthcare timeframes and scenarios, with an emphasis on surgery, is needed.

The healthcare revolution is a timely complementation of the 4th industrial revolution, particularly in regard to the ongoing landscape of technological advancement. Notably, AI has gained attention as a key element of this technological advancement . AI has the potential to liberate health from a focus on sickness and care from governments and organizations to patients and communities. Technologies such as personalized medicine and longer lifespans are expected to become available. However, meticulous implementation coupled with the development of handles against ensuing risks is also pivotal. Without these conditions, there is a threat of increased separation into individual capabilities. Current integrated systems may cease to work sustainably.

2. Applications of AI in Healthcare Delivery

Diagnostic imaging and radiology have seen the most significant application in health care delivery. Every year, millions of patients undergo X-rays, CT scans, MRIs, or ultra-sonographical tests. Diagnostic imaging comprises complex signal processing algorithms that reconstruct an image highlighting abnormal anatomical features. Automated analysis of medical images can be broadly categorized into a classification of images with normal or abnormal anatomy, the segmentation of images highlighting the anatomical region of interest, and the analysis of images with multiple anatomical abnormalities. Previous studies have reported this issue of test-shortage and its far-reaching consequences in low-income and middle-income group countries. While many replaceable solutions exist to automate this process using artificial intelligence (AI), they are still in their infancy, under development, or have efficacy and uphold concerns. Research is currently being conducted on an AI algorithm that can learn from medical images similar to human clinicians. The prospective AI algorithm limits both the number of cases and the preciseness of the positive cases from the routine screening.

There is an immense amount of Executive Health check-up report data collected in hospitals, of an entire cohort of patients undergoing these tests every year. AI algorithms analyze the reports of this entire cohort of patients in the time series tossing when all these tests were performed. This leverages the utmost power of ML, as the power of ML lies in analyzing exhaustive data. Recent studies trained AI algorithms on a considerable amount (∼7000 images) of mammogram data, and the yield of learning improved more so. These newly trained AI algorithms in combination with the positive cases from the routine screening predicted cases in an entire cohort of patients (∼130,000 images) in the US better than the human clinicians. Further, the examination of medical images using these algorithms led to a ten-fold reduction in the false positives and increased the identification of malignancies more so (Aamir et al., 2024). The literature accentuated that the integration of AI algorithms in medical imaging also curtailed the cost of examinations substantially. The gradual introduction of these AI models in radiology selected hospitals effortlessly reduces the reliance on expensive diagnostic tests, avoiding unnecessary procedures that would test the full spectrum of a disease. Reports of AI diagnostics in breast cancer changed the paradigm of addressing AI linear to the introduction of new technology in clinical medicine.

Predictive analytics, a subset of AI that scrutinizes past facts to get patterns in hopes of determining future results, is becoming increasingly commonplace in healthcare settings. It has the prospective to revolutionize traditional approaches to early disease detection and criteria of treatment of malignancies. Screening, patient triage, drug selection, bone marrow transplant compatibility determination, outlier detection, and survival prediction of patients are some of the issues in broader avenues that can be addressed through predictive analytics. These predictive models give genuine and actionable insights like how at present, the Judiciary system utilizes predictive models of AI for risk assessment in trials. The methodology and process of big data predictive models creating superstars in baseball (Money-ball concept) can be reproduced in Medicine. Healthcare requires meticulous record-keeping, as one year of treatment can result in hundreds of laboratory tests and imaging procedures for any individual. There is a consistent and simple format to record Laboratory procedures of patients and the laboratory databases by vendors are quite standardized. Similarly, there is a simple and consistent format to record imaging procedures. Also, imaging procedures are considered more objective than the laboratory procedures, as both imaging and laboratory information are used to score a patient in majority ANNs if there is a need for a surgical procedure. The initial prototype models have been trained on the past two years of hospital records of the patients and the performance is promising surpassing the achievements made in MRI in an AI model or CT scans reading. Further modeling is in progress to consider multimodal medicine in the Data base. Prior malignancy detection has emerged to be one of the frontiers where the need and promise of simplistic and accurate tests not reliant on expertise are desired (Grezenko et al., 2023). AI’s promising role in revolutionizing personalized medicine-‘prevention rather than cure’ is indeed enigmatic and can only be compared to the effect of fast moving automobiles on the invention of steam trains.

2.1. Diagnostic Imaging and Radiology

Artificial Intelligence (AI) is revolutionizing diagnostic imaging and radiology by providing advanced systems and methodologies that allow faster, more accurate, and reliable interpretation of medical images than traditional approaches (Sindhu et al., 2024). Computer-aided diagnosis (CAD) refers to computer systems and methodologies that assist the decision-making process from an automatic or semi-automatic evaluation of medical images. Radiology is one of the main medical disciplines in which CAD is applied successfully. Early CAD systems in the 1980s were based on the expert system and traditional approaches. However, with the evolution of machine learning in the 1990s and deep learning in the 2010s, new paradigms for CAD systems and methodologies were enabled. Deep learning approaches successfully automated many in-house and ad-hoc pre-processing and analysis steps in CAD and allowed the creation of end-to-end systems applicable to different imaging modalities – Computed Tomography (CT), X-ray, Digital Mammography (DM), and Magnetic Resonance Imaging (MRI). These advancements provided automatic, robust, and scalable solutions that in many cases surpassed the performance of earlier models based on the expert system (Gurgitano et al., 2021).

Digital pathology is considered as a new evolving research field that aims for a systematic digitization of pathology services similar to radiology. CAD in digital pathology is not yet adopted, and the current systems are still in clinical trials, addressing the very challenging task of tissue classification and lesion assessments in heterogeneous histological images. With the introduction of Generative AI (G-AI), the CAD field opens fresh avenues for innovation, enabling the development of new advanced systems and methodologies that are not limited to supervised learning. G-AI brings automated data annotation and creation of synthetic and augmented datasets to simulate rare pathologies, artifacts, and different imaging conditions. More interpretable CAD systems modeled on the functioning of human vision can be developed with G-AI by integrating visual representations from computational architectures with deep learning models. Finally, beyond enhancing the interpretability of Deep Neural Networks (DNN) and AI classifiers, G-AI provides fresh methodologies for designing CAD systems mimicking human cognition, potentially leading to the transformation of modern medicine.

2.2. Predictive Analytics and Personalized Medicine

AI has the potential to revolutionize the field of predictive analytics and personalized medicine in the following ways. Patients can receive better insights into their health, supported by wearable and embedded devices that continuously read an individual’s biometrics, such as heart rate, blood pressure, and hydration levels. These devices analyze the biodata they read, building a better picture of their health. Abnormal values trigger alerts and recommended actions on apps. In time, AI systems will manage the health and lifestyle of individuals by continuously learning about them and advising them on how to maintain good health. AI will also allow predictions of diseases for populations, with analyses of medical records, alternative data, such as on buying patterns, social networks, gambling activity, and online behavior. If abnormalities in the records or predictions are found, the individual patient is monitored more closely and contacted personally (Ahmed et al., 2020). It is possible to detect and deal with the accelerating chronic lifestyle-related diseases affecting populations in developed countries.

AI will integrate patient data better, gathering and analyzing a vastly enriched data offer to get a holistic view of the whole patient background. Building on existing models, they will better define patient subgroups in heterogeneous populations and allow companies to target personalized interventions better to optimize treatment. Although diversity-based analysis is ten years away, AI will transfer effective interventions from one population to another, helping to adjust population-level models to converge to new environments and find effective policies. AI will therefore transform medical interventions from ill-defined “one-size-fits-all” strategies to fully personalized and clearly defined ones, maximizing intervention effectiveness while minimizing risks and costs.

AI will better scrutinize safety, quickly detecting unwanted side-effects for interventions with very many treated individuals. Analyzing treatment data, AI will detect abnormal patterns associated with treatment implementation and alert regulators if there are sufficient reasons for concern about safety. AI will therefore optimize the quality and safety of food, drugs, therapies, and commercial products globally.

AI will transform the medical profession, minimizing the need for doctors or other agents between patients and the system (Dixon et al., 2024). The combination of highly reliable AI health sensors with personal AI systems allowing direct contact between patients and diagnosis & prescription systems will render most of healthcare fully automated. Health agents will have access to instant and detailed health information on everyone. The incomplete implementation of AI in healthcare will favor these agents in the near future, allowing them to more effectively investigate and manipulate the health of the population to fit their interests. Law enforcement agencies or unethical corporations could exploit AI to profit from inaccurate health information or manipulate interventions and health-related social dynamics. The future perspective analyzed here makes desirable the immediate action to democratize AI in healthcare.

3. Challenges and Ethical Considerations

The integration of AI in healthcare delivery comes with numerous challenges and ethical considerations that need to be addressed in order to ensure the safe and effective use of AI technologies. One of the most critical challenges is data privacy and security, as the use of AI in healthcare often involves the collection and analysis of sensitive personal health information. Healthcare organizations must implement robust data privacy and security measures to safeguard this information and comply with regulatory frameworks such as the Health Insurance Portability and Accountability Act (HIPAA) (Mennella et al., 2024). AI algorithms are often trained on large datasets, which can lead to unintended bias if the data is not representative or if certain demographic groups are underrepresented. This bias can result in unfair treatment recommendations and health disparities. Healthcare organizations must ensure that AI algorithms are designed and trained in a way that minimizes bias and ensures fairness in treatment recommendations.

Another ethical consideration is the transparency and interpretability of AI algorithms. Many AI technologies used in healthcare are complex and difficult to interpret, which can make it challenging for healthcare professionals to understand how treatment recommendations are made. This lack of transparency can erode trust in AI technologies and lead to reluctance in its adoption (Pasricha, 2022). Healthcare organizations must prioritize the development and use of interpretable AI algorithms and provide training to healthcare professionals to understand and effectively use these technologies. Lastly, AI technologies have the potential to replace or augment the role of healthcare professionals in certain tasks, leading to concerns about job displacement and the devaluation of human expertise. Healthcare organizations must ensure that the use of AI technologies is accompanied by clear guidelines on the role of healthcare professionals in treatment decisions and that these technologies are designed to enhance rather than replace human expertise.

3.1. Data Privacy and Security

The growing complexity of data privacy and security issues in healthcare fuels the uncertainty about the best approach to preserving these standards in AI applications. With the rapid development of health-focused technologies, the history of data breaches in the healthcare industry raises concerns about patient data and sensitive medical information. With the inevitable drive toward digitization and automation, protecting both design and personal data becomes an even more demanding challenge (Murdoch, 2021).

Advances in healthcare artificial intelligence (AI) are occurring rapidly, and many technologies in the space are approaching feasibility. A few AI technologies are close to being integrated into the healthcare system. But there are serious issues regarding data privacy, trust, and accountability raised by commercial healthcare AI. The design and implementation of healthcare AI systems will necessitate placing sensitive health data under the control of for-profit corporations, and there are significant privacy considerations. The rapid expansion of digital health data, in terms of quantity, granularity, and diversity, from electronic health records, wearable and implantable devices, genomes, and other sources has created significant opportunities for the use of the technology in healthcare. In parallel, advances in computational capability and growing ecosystem infrastructures (e.g., cloud computing, open-source software, AI-as-a-service) have improved the feasibility to build and operate healthcare AI systems. At the same time, growing public distrust in technology companies and concerns about the ethical use of digital health data compound these challenges (Bak et al., 2022).

3.2. Bias and Fairness in AI Algorithms

AI algorithms must undergo critical examination, by clinicians, programmers, and bioethicists, before implementation in patient care settings. In addition to patient safety, biased AI algorithms raise issues regarding compliance with State, Federal, and international anti-discrimination statutes (Cerrato et al., 2022). One of the strongest societal mandates for public health is at least minimal healthcare for all. Biased algorithms—those that allocate healthcare resources in a manner that favors one demographic group over another—would violate this mandate to treat similarly situated individuals alike, raising a host of legal and ethical concerns. Biased algorithms would invariably also adversely affect social determinants of health leading to bias cascade later in the care continuum and influencing decisions about triage and timely receipt of transplant, in-patient admission, and administering life-saving therapies. Such cascading bias could lead to health outcomes inequitable on a group level and persisting differences between groups (Chen et al., 2023).

Healthcare is a domain imbued with bias: innate biases in patient selection, diagnosis, and treatment decisions based on demographic factors, health systems biases relating to access and availability of services, and a host of variables unique to health and fitness (both mental and physical). The biases present in the healthcare system are compounded by algorithms that have been shown to accept the bias imbued in the training data on which they were developed. These biases can have an enormous negative impact on the health and well-being of individuals and continue to widen the gap between the privileged and the disadvantaged.

4. Future Directions and Opportunities

One of the most promising opportunities introduced by AI is the enhanced integration of technology and healthcare service delivery through telemedicine. Telemedicine can facilitate remote consultations and monitoring of patients away from the practitioner, improving accessibility to healthcare services. Combining telemedicine with AI’s capabilities in diagnostics and monitoring can expand both access to quality healthcare services and the proportion of patient care that can be performed remotely (Arora, 2020).

The introduction of diagnostic and monitoring AI frameworks paired with telemedicine can provide healthcare benefits for rural and underprivileged areas where healthcare facilities are scarce. In addition, patients that do not have the resources, whether it be time, transportation, finances, or parental assistance, to visit healthcare facilities can benefit greatly from at-home monitoring technologies paired with telemedicine systems. Telemedicine has the potential to be considered self-evident in the delivery of healthcare service. However, to truly make large-scale telemedicine widely implementable, a combination of further innovation in the medical use of AI with telemedicine service delivery is required (Grezenko et al., 2023).

Currently ongoing research and development into innovative AI medical frameworks in diagnostics, monitoring, prescription management, radiology, and surgical assistance can pave the way for imagining the healthcare trajectory AI would enable in the future. The implementation of AI systems as a healthcare service in daily use by hospitals and practitioners would undoubtedly drive even wider innovation and development mission within the field of AI.

4.1. Integration of AI with Telemedicine

Telemedicine is one of the most important applications supported by Artificial Intelligence technologies, technologies that can help to improve the quality of telemedicine at the same time that they democratize access to specialists. Medical consultations and medical care remotely have been provided for years, essentially since portable medical devices and mobile phones become a reality. The explosive growth of the Internet has pushed medicine into cyberspace through telemedicine practices. However, telemedicine practices are generally only viable in private medical systems, as in social medical systems they are generally perceived as too expensive and of questionable added value. AI technologies can improve telemedicine practices, finding applications in preliminary diagnosis systems, data compression, and intelligent systems that model and understand the patients’ mood using text understanding techniques. The goal is not to replace doctors but to be intelligent front-ends that provide preliminary analysis of medical cases and then reinforce medical analysis and treatment suggestions (Grezenko et al., 2023).

4.2. Research and Development Innovations

Innovations within the healthcare landscape can be approached from a plethora of aspects. However, one of the most impactful advancements within the medical field with AI facilitation includes the opportunities for the advancement of research and discoveries of drugs, medical devices, and treatment modalities. It is through such innovative advancements that the field of healthcare can continuously be developed and improved upon (Ioana Visan and Negut, 2024). By ensuring efficacy and effectiveness alongside improved safety mechanisms, healthcare can take on a broader aspect than just wellness as it embarks upon the sense of precaution.

Traditionally, the discovery, testing, and production of drugs and devices remained at hands of highly skilled professionals in laboratories. Considered both time and labor intensive, decades would go by prior to the mass production of a product if passed trials (Han and Tao, 2024). With the introduction of AI, the ability to build computer generated models designed to replicate human anatomy has brought about simulated training capabilities that could replicate real complexity. In approach to pharmacology, variables in anatomy and health considerations can be input alongside the genetic structure of the molecule in question to evaluate the compatibility of the drug with the patient in question. With promising results, AI can go a step further by assisting in the early monitoring of the administered drugs through computerized imaging methods. Furthermore, AI can design drugs that upon latent complex modelling can recreate molecules counteracting the factors of a disease based on pre-stated variables.

Additional prospects of AI can allow biologics to have digital twins within the systems of patients who have incorporated a newer medication to the market. These digital representations can later be used to monitor the exposition of drug particles biologically across time allowing for unforeseeable drug behavior to be early detected. Lastly, AI is able to triage patients based on biomarkers identified within imaging methods, lab tests, or biofluids parameters allowing for a more refined selection criteria in drug testing trials.

5. Conclusion

The convergence of artificial intelligence (AI) and healthcare holds great potential to transform health services, delivery, and outcomes. AI amplifies human expertise, enabling unparalleled insights and optimizing decision-making based on trained algorithms and data. Although past technologies primarily enhanced established processes, AI’s unprecedented capacity permits entirely new workflows and systems as a system tool (Grezenko et al., 2023). The transition of AI activities in medicine, particularly to surgical and therapeutic applications, ushers in a multifaceted revolution in patient care’s execution and functionality.

AI’s fast-moving applications in surgical and therapeutic medicine reshape the practice of medicine holistically. These technologies’ active engagement in multiple medical decision-making processes warrants vigilance and responsibility by relevant stakeholders. AI in surgical medicine encompasses a diverse toolbox, deploying novel devices and procedures while leveraging various algorithms on diverging technical grounds. AI technologies require speciation based on input types—input-computational AI sculpting capabilities, knowledge-based AI, and output AI-technology classification for actuation. Further nuance highlights data-level distinction, operational niche, and the consideration of medical roles and structural involvement.

References:

Wójcik, S., Rulkiewicz, A., Pruszczyk, P., Lisik, W., Poboży, M., and Domienik-Karłowicz, J. “Beyond ChatGPT: What does GPT-4 add to healthcare? The dawn of a new era.” (2023). ncbi.nlm.nih.gov

Grezenko, H., Alsadoun, L., Farrukh, A., Rehman, A., Shehryar, A., Nathaniel, E., Affaf, M., Khaleel I KH Almadhoun, M., and Quinn, M. “From Nanobots to Neural Networks: Multifaceted Revolution of Artificial Intelligence in Surgical Medicine and Therapeutics.” (2023). ncbi.nlm.nih.gov

Vandenberk, B., S. Chew, D., Prasana, D., Gupta, S., and V. Exner, D. “Successes and challenges of artificial intelligence in cardiology.” (2023). ncbi.nlm.nih.gov

Aamir, A., Iqbal, A., Jawed, F., Ashfaque, F., Hafsa, H., Anas, Z., Olatunde Oduoye, M., Basit, A., Ahmed, S., Abdul Rauf, S., Khan, M., and Mansoor, T. “Exploring the current and prospective role of artificial intelligence in disease diagnosis.” (2024). ncbi.nlm.nih.gov

Sindhu, A., Jadhav, U., Ghewade, B., Bhanushali, J., and Yadav, P. “Revolutionizing Pulmonary Diagnostics: A Narrative Review of Artificial Intelligence Applications in Lung Imaging.” (2024). ncbi.nlm.nih.gov

Gurgitano, M., Alessio Angileri, S., Maria Rodà, G., Liguori, A., Pandolfi, M., Maria Ierardi, A., J. Wood, B., and Carrafiello, G. “Interventional Radiology ex-machina: impact of Artificial Intelligence on practice.” (2021). ncbi.nlm.nih.gov

Ahmed, Z., Mohamed, K., Zeeshan, S., and Dong, X. Q. “Artificial intelligence with multi-functional machine learning platform development for better healthcare and precision medicine.” (2020). ncbi.nlm.nih.gov

Dixon, D., Sattar, H., Moros, N., Reddy Kesireddy, S., Ahsan, H., Lakkimsetti, M., Fatima, M., Doshi, D., Sadhu, K., and Junaid Hassan, M. “Unveiling the Influence of AI Predictive Analytics on Patient Outcomes: A Comprehensive Narrative Review.” (2024). ncbi.nlm.nih.gov

Mennella, C., Maniscalco, U., De Pietro, G., and Esposito, M. “Ethical and regulatory challenges of AI technologies in healthcare: A narrative review.” (2024). ncbi.nlm.nih.gov

Pasricha, S. “AI Ethics in Smart Healthcare.” (2022). [PDF]

Murdoch, B. “Privacy and artificial intelligence: challenges for protecting health information in a new era.” (2021). ncbi.nlm.nih.gov

Bak, M., Istvan Madai, V., Fritzsche, M. C., Th. Mayrhofer, M., and McLennan, S. “You Can’t Have AI Both Ways: Balancing Health Data Privacy and Access Fairly.” (2022). ncbi.nlm.nih.gov

Cerrato, P., Halamka, J., and Pencina, M. “A proposal for developing a platform that evaluates algorithmic equity and accuracy.” (2022). ncbi.nlm.nih.gov

Chen, F., Wang, L., Hong, J., Jiang, J., and Zhou, L. “Unmasking Bias in AI: A Systematic Review of Bias Detection and Mitigation Strategies in Electronic Health Record-based Models.” (2023). [PDF]

Arora, A. “Conceptualising Artificial Intelligence as a Digital Healthcare Innovation: An Introductory Review.” (2020). ncbi.nlm.nih.gov

Ioana Visan, A. and Negut, I. “Integrating Artificial Intelligence for Drug Discovery in the Context of Revolutionizing Drug Delivery.” (2024). ncbi.nlm.nih.gov

Han, Y. and Tao, J. “Revolutionizing Pharma: Unveiling the AI and LLM Trends in the Pharmaceutical Industry.” (2024). [PDF]