FEATURE | Applications of Artificial Intelligence to Global Healthcare in the Past and Present

With the release of ChatGPT to the general public in November 2022, discussions of artificial intelligence became mainstream, and the implications of AI for many facets of human society became apparent. Students could use large language models (LLMs) to complete their homework in record time, customer support agents could employ artificial neural networks (ANNs) to analyze queries, and linear regression machine learning (ML) models could be used for price prediction in financial markets. However, the initial development of AI began long before ChatGPT and other generative AI models emerged. 

In 1950, English mathematician and computer scientist Alan Turing proposed the Turing Test, a method of determining whether a computer can “think” and perhaps the first notion of artificial intelligence in the twentieth century. Shortly after, at the 1956 Summer Dartmouth Conference on AI, the term “artificial intelligence” was officially coined by American computer scientist John McCarthy. 

Picture of a few organizers and participants from the 1956 Dartmouth workshop on AI (in front of Dartmouth Hall)

During the 1970s, AI began to enter the healthcare sector, beginning with the Stanford University Medical Experimental Artificial Intelligence in Medicine (SUMEX-AIM) project. This project, which was funded by the NIH, had two primary objectives: to introduce the applications of AI to medical sciences and to demonstrate the significance of network-based collaboration with computer systems. A key product of this project was the MYCIN system, a knowledge-based AI program used to address issues in “the domain of bacteremia therapy.” This implementation of AI inevitably had profound implications for many facets of healthcare innovation. 

At the turn of the twentieth century, AI technologies evolved from their primitive creations and were rapidly constructed to replicate human expert decision-making abilities in medicine. Initial prototypes of AI systems were designed to analyze electronic health records (EHRs) and medical imaging. In the 2010s, developers improved upon these programs to encompass certain diagnoses, with a focus on imaging studies such as MRI and CT scans, and to perform skilled predictive analytics. For instance, Korotkov et al. conducted a computerized analysis of pigmented skin lesions (PSLs) in 2012, and Ghose et al. reviewed applications of AI to cervical cancer radiation therapy treatment planning in 2015.

In the current day, pharmaceutical conglomerates apply AI to their endeavors on a large scale. For instance, Johnson & Johnson, the multinational biotechnology and pharmaceutical company, developed the Carto 3 System, an advanced 3D heart mapping technology that integrates AI with algorithmic models. Their VirtuGuide software uses AI to analyze patient anatomy and reduce complexity for bunion surgery. Nicole Turner, Senior Director of Global Development at Johnson & Johnson, remarked, “Our goal is to leverage the power of AI to bring clinical trials to more patients, rather than waiting for patients to come to us.” 

Smaller commercial startups and healthcare organizations have also joined the revolution of AI advancements in medicine. Butterfly Network’s iQ3 is a point-of-care ultrasound (POCUS) device and was honored as the best medical technology at the 2024 Prix Galien USA Awards. Baseline AI, a Y-Combinator-backed startup, uses AI to automate the writing of research protocols, data transformation, and analysis creation. 

Butterfly Network’s iQ3 POCUS device

Although the direct application of these technologies lies in the healthcare industry, computer programmers use their extensive knowledge to design these algorithms. Languages including Python, R, and Flutter have become instrumental platforms in the creation of secure AI models for healthcare. Medical imaging and disease classification systems employ convolutional neural networks (CNNs). Moreover, generative adversarial networks (GANs) are used for data augmentation and the creation of synthetic medical images. 

Various drawbacks make these AI technologies in healthcare disadvantageous in certain regards. Models can be susceptible to hackers via data breaches, leaving confidential electronic health records highly vulnerable. Algorithms make unimportant associations between data sets through a process called overfitting, which can make AI models unsuitable for application to other data sets. Lastly, some individuals view the prospect of AI in healthcare with ethical concerns about the “black-box” nature of most systems and the absence of standard guidelines for moral use. 

Nevertheless, with every passing year, AI technology is becoming more prevalent in healthcare networks across the globe. These advancements support millions of people and mark distinct steps in the progression of human innovation. AI in healthcare has the potential to expand into a highly lucrative field of the future and offer great occupational opportunities for high school students of the current generation. As Dr. Robert Pearl stated in his book ChatGPT, MD stated, “AI will become as common in healthcare as the stethoscope,” a statement encapsulating the vast, untapped potential of AI in the field of medicine.