As I sit and write this article at a beautiful café in Kampala, I can’t help but be in awe of our era. It seems that the sci-fi movies and books I used to read about as a child are becoming a reality — or will soon become a reality. I am talking about the technological boom brought on by artificial intelligence.
I’m sure you have heard about or are using chat GPT which is, in my opinion, one of its most outstanding achievements. You might not have realized it yet, but AI surrounds you, and you’ve grown dependent on it. Let me paint a picture for you: say someone is talking about a topic you’re unfamiliar with, and as a ‘know-it-all scientist’, you want to chime in on the talk. What do you do? Google, right? As you type the topic of interest, you are surprised that with the first two or three letters you order, Google gives you a list of suggestions. Or, like most people, you would say, “Hey, Google!” and then ask your question – it will give answers because, my friend, it’s AI, and Google have been using it since the early 2010s’. If, instead, you wish to relax over a movie with friends and family at home and you open your Netflix account, there comes a list that suggests what you might enjoy; most of the time, it’s always spot-on. This is also one of the technological advancements in AI. As a researcher/scientist, you have probably already used chat GPT for review or draft writing and research. The futuristic era that books and movies used to discuss is now moving at a breakneck pace, augmenting or potentially replacing humans in the workplace.
But what is AI? And how has it developed over the years? The term “AI” is somewhat enigmatic, and some companies have used the term interchangeably with machine learning and deep learning. It was coined in 1956 by computer scientist John McCarthy. In simple words, artificial intelligence is a human intellect simulated by machines. The field of AI was founded on the concept that the human mind can be precisely described and that a device can mimic it. This has raised many questions and apprehensions on the ethics of pursuing AI, which we have seen in many science fiction movies such as “I Robot”.
From the above definition, you can understand that AI is an umbrella encompassing multiple technological advancements. Computer scientists have classified artificial intelligence into General Intelligence, which consists of building neural networks and simulating human/superhuman level intelligence -but hasn’t become feasible yet- and Narrow Intelligence; this is where we are witnessing the majority of the breakthroughs in the news lately and involves machines or software performing a single or straightforward task efficiently. “Machine learning” and “deep learning” refer to the process by which software or applications become accurate at predicting outcomes without being explicitly programmed to do so.
AI has been on a rollercoaster over the past few decades. At its inception, machines that played checkers solved mathematical problems, and spoke English were built which brought much promise in the field. Most scientists in this area promised Artificial General Intelligence (AGI), which could mimic all human functions within 20 years; however, they soon ran into a rut as many tasks to be considered were overlooked, and funding for the field dwindled by the end of the 1970s. This era became known as ‘the AI winter’. The area experienced a second, albeit short, boom in the 1980s with Lisp machines, and funding started coming through, but it soon dwindled due to the failure of these machines. Artificial Intelligence (AI) finally broke out in the 1990s when scientists realized that focusing on artificial narrow intelligence would be the right path before reaching general intelligence. Since then, this field’s contributions have been increasing steadily and are still booming.
- AI Benefits in Medicine
Artificial narrow intelligence has made tremendous strides in healthcare, connecting patients with appropriate physicians, diagnosing patient conditions, predicting prognosis, transcribing and translating languages, and sorting files and images. All of these technologies use similar patterns; first, the system is given a considerable amount of data called ‘the learning data’; on this data, machine learning algorithms are used to gain meaningful information, such as recognizing patterns, classifying images, and so on. This information is then utilized to find a meaningful solution to a defined medical problem. This breakthrough has been applied in numerous departments in the field of medicine; some examples include:
Augment Clinician diagnostic capabilities
AI has come so far in augmenting diagnoses in cases involving pictorial data. A study done in 2017 has shown that an AI-powered system was as accurate in diagnosing pathology slides of early breast cancer as 11 clinical pathologists. Another study, by Esteva et al. showed that AI was more accurate in diagnosing skin conditions when compared to 6 experienced dermatologists. A study at MIT by Mozannar et al. in 2020, has revealed that the assessment of radiologic cardiac findings had about 8% better results with the aid of AI. Numerous studies in colonoscopes and endoscopic assessments in the gastroenterology department have shown better outcomes with early detection of various pathologies, such as colonic polyps, atrophic gastritis, and IBD.
AI’s capability to process extensive data in a fraction of the time makes it superior in diagnosing diseases with close to no errors and an additional advantage of lack of burnout compared to human physicians. One startup tech, Thymia, has put this capability of AI into health use. The app assesses patients’ depression status by analyzing the type of video games they enjoy, comparing the facial reactions and voices of these patients with thousands of anonymous data, and providing essential results to mental health clinicians.
Another critical AI technology is using smartphones and smartwatches such as the Apple Watch 3. A study on atrial fibrillation has shown better detection than traditional in-office assessment. IBM’s AI, Watson, has made great strides in detecting hypoglycemic states in diabetic patients by monitoring blood sugar levels intermittently.
Precision Medicine applications
In an era in which healthcare has narrowed its focus to individual patients by studying their genetic makeup and making predictions on their responses to medications, prognoses of certain diseases they may have, as well as the risk factors associated with those diseases – this is known as the field of precision medicine. Artificial Intelligence (AI) has found multiple applications within this realm. AI will come with easy and fast processing of large datasets in population genomics, allowing a patient’s genetic makeup to be analyzed and compared against these datasets. This makes it possible to predict certain diseases, reactions to medications, and prognoses of diseases. Not only that but patients’ current history and clinical condition are considered and processed during therapy planning; moreover, social determinants of health are taken into account when using this data, making the outcome better focused on an individual patient.
Expedited Drug discovery and development
In the pre-AI times, drug development took years, sometimes decades. With the Advent of AI and its super computational capabilities, this lengthy process will be shortened considerably. Scientists used trial and error on some compounds as medication; however, that could easily be bypassed now with AI and its extensive data computational abilities. It can deduce and develop combinations that would be highly successful and cheap. Hence, pharmacists could prioritize such predicted compounds to test.
Seamless healthcare operations:
Patients’ complaints during their hospital visits are a lack of communication, as they are overwhelmed by their medical condition and the crowded, disorganized hospital environment. To bridge this gap, some healthcare facilities have leveraged AI; one such technology is Babylon. This AI takes in the patient’s symptoms from them, directing them to an appropriate physician and providing guidance on where to go for treatment.
Trilliant Health founded another patient aid AI, the Similarity Index; this AI aids patients seeking medical care by providing information on hospitals; it utilizes data such as hospital readmission rates, hospital-acquired diseases, and mortality rates to rate hospitals and when a patient offers a name for a hospital its rate opposed to other 2000 US hospitals is provided.
- Limitations and Ethical Consideration of AI Use in Medicine
Artificial intelligence has a lot of promise but has yet to be perfect. There are various limitations, drawbacks, and ethical issues associated with it. Here I will mention some of the most notable ones;
Data Related issues
AI is entirely dependent on data; it represents the given data. This fact opens the technology to several limitations, with the most important one relating to data being the design of the model. The model is given a set of training data and designed to fit this dataset, which poses an issue known as “overfitting” when incoming new data varies from learning data (retrospective) and fails to fit into its changing pattern. To account for this constant recalibration is necessary.
Another issue with data is its quality and bias. AI models will compute anything based on the patterns they find from their training data. So, if the data collected is skewed or biased, the resulting output will be incorrect. For example, in a study using AI that was assessing the health of communities based on healthcare costs, because so little budget is allocated to black communities despite having similar levels of health issues as white communities, the AI concluded, based on budget data, that black communities were healthier than white ones. This issue is fixed by making the training data representative of the population of interest.
The third issue is more of an ethical one; healthcare data is susceptible as it contains individual personal and private information. Accessing this data would pose a lot of difficulties and legal issues regarding data ownership. The consensus on this issue is to give data ownership to the owner, i.e. the patient, as this empowers patient engagement.
Expertise needed
Artificial Intelligence has recently started to be applied in healthcare. Consequently, there needs to be more knowledge about the detailed processing of AI among healthcare providers. This issue has been termed the “Black Box” problem – a lack of understanding in physicians when they can’t explain how AI models transform inputs into outputs. This limitation could have serious consequences if physicians cannot explain how an AI thinks – then they won’t recognize mistakes made by the model. Some medical schools are trying to train the upcoming physicians in AI engineering as part of their curriculums to narrow the gap and train “augmented physicians” to work with AI.
Limiting human contact
One of the fears of augmenting healthcare with AI is that it will lose the human touch. There are many signs that physicians pick up from their patients by their body/ verbal cues. An Accenture 2020 Digital Health Consumer Survey showed that 29% of patients were opposed to AI augmentation of healthcare because they were afraid it would lessen the human touch, which they believed was crucial for healthcare. A possible solution for this concern would be augmenting health with AI in diagnostics while doctors will be free to interact and focus on the patient’s treatment.
Wrong Diagnoses
An ethical problem arising with AI’s advent in healthcare is wrong diagnoses. Who is responsible when a physician has used AI models for patient diagnosis, and a mistake is made? The jury is still out on this issue, as multiple factors must be considered. Possible solutions to this issue include proper training on AI among physicians.
However, it is essential to note that multiple studies have shown that the margin of error among physicians and AI is similar. According to a global survey of primary care errors, 5 out of 100 outpatient cases will be misdiagnosed.
Fear of technology taking over
Doctors fear being replaced by the novel AI technology like every other individual. Recent studies comparing AI to physicians have fueled this fear. Most of these studies were flawed by design and lacked primary replication. The crude truth is that in its current form, AI cannot replace healthcare provided by physicians; however, it can augment it.
AI’s exact, error-free, and fast data processing capabilities can be utilized to aid physicians in processing healthcare documents and diagnoses. At the same time, doctors can focus on what initially drew them to the field of medicine: “patient care”. To address this issue, comparative studies should be conducted between physicians using AI and those without. This will help strengthen the relationship between AI and healthcare professionals by recognizing it as a complementary technology.
- Prospects of AI in Medicine
According to Tom Lawry, National Director of AI for Health & Life Sciences at Microsoft, there are two major applications of AI in healthcare: automation and augmentation. We are still a long way from healthcare being completely automated by AI. However, augmentation holds a lot of promise for the healthcare industry. The main focus of most health tech companies is for AI to be used as an assistant to physicians, helping them with diagnosing, risk-factor identification, and management planning. In contrast, the physician has overall responsibility when it comes to patient interaction and clinical decisions.
Contemporary healthcare systems have increased life expectancy for populations. The downside is that a large population now has chronic conditions requiring monitoring. Older patient data, with multiple health conditions, stretches previous systems to their limits, requiring extensive processing and sifting by human labor. Ever since the Covid pandemic, healthcare professionals have reported more burnout than before.
There is a generational difference in how healthcare should be carried out. According to a report from Forbes’ interview with Tom Lawry, millennials don’t care much about face-to-face contact. At the same time, boomers want constant primary care physicians for interaction on a face-to-face basis. AI would cater to each individual according to their terms as it will provide boomers with apps that attach them to physicians. And millennials would be provided virtual healthcare akin to the way they handle everything else in life.
Another prospect of AI is an augmentation of surgical care. Robotic surgeries have resulted in minimal scarring, less pain, and shorter hospital stays. These advantages will be further enhanced by using AI, which can use CT and MRI results of patients to create a 3D patient anatomy and aid the surgeon in handling critical anatomies. As AI takes in multiple data and learns about human anatomy, it is hoped that robotic surgery can be led with little to no human input through AI.
The Impact Factor of a journal can typically be found on the journal’s official website or through various mentioned academic databases in this post.
- Summary
Artificial Intelligence (AI) has witnessed astounding progress, ushering in a new era across various domains, including medicine. AI’s primary focus in healthcare is on narrow intelligence, enabling machines and software to excel at specific tasks with unparalleled efficiency. It has shown great success in transforming diagnosis, precision medicine, drug discovery, and the overall operations of healthcare systems.
Despite its undeniable advantages, AI in medicine has inherent limitations and ethical considerations. Challenges include data-related issues, including overfitting, data quality, and biases that can skew results. The opacity of AI algorithms to healthcare providers gives rise to the enigmatic “Black Box” problem, impeding comprehensive understanding. Concerns about the potential for wrong diagnoses have raised discussions about accountability. The fear of technology replacing physicians is common, but AI’s current role primarily revolves around augmentation rather than outright replacement.
The prospects of AI in medicine lie in both automation and augmentation. While fully automated healthcare remains a distant prospect, augmentation holds promise as a valuable assistant to physicians, such as aiding in diagnosis, risk-factor identification, management planning, and relieving them of mundane tasks. By alleviating burdensome responsibilities, AI empowers physicians to focus more on personalized patient care. In surgical care, AI’s integration can revolutionize procedures by enhancing robotic surgeries through advanced techniques like 3D anatomical mapping, potentially paving the way for autonomous surgical interventions.
Alexander Habte
Global Researcher Club Author
Assab Military Hospital, surgical department, Orotta School of Medicine, Eritrea
1. Anderson, Susan Leigh. “Asimov’s ‘Three Laws of Robotics’ and Machine Metaethics.” AI & SOCIETY 22, no. 4 (April 2008): 477–93. https://doi.org/10.1007/s00146-007-0094-5.
2. Beal, Jacob, and Patrick H. Winston. “Guest Editors’ Introduction: The New Frontier of Human-Level Artificial Intelligence.” IEEE Intelligent Systems 24, no. 4 (July 2009): 21–23. https://doi.org/10.1109/MIS.2009.75.
3. Berdahl, Carl Thomas, Lawrence Baker, Sean Mann, Osonde Osoba, and Federico Girosi. “Strategies to Improve the Impact of Artificial Intelligence on Health Equity: Scoping Review.” JMIR AI 2 (February 7, 2023): e42936. https://doi.org/10.2196/42936.
4. Bowling, Michael, Neil Burch, Michael Johanson, and Oskari Tammelin. “Heads-up Limit Hold’em Poker Is Solved.” Science 347, no. 6218 (January 9, 2015): 145–49. https://doi.org/10.1126/science.1259433.
5. Briganti, Giovanni, and Olivier Le Moine. “Artificial Intelligence in Medicine: Today and Tomorrow.” Frontiers in Medicine 7 (February 5, 2020): 27. https://doi.org/10.3389/fmed.2020.00027.
6. Buch, Varun H, Irfan Ahmed, and Mahiben Maruthappu. “Artificial Intelligence in Medicine: Current Trends and Future Possibilities.” The British Journal of General Practice 68, no. 668 (March 2018): 143–44. https://doi.org/10.3399/bjgp18X695213.
7. Buttazzo, G. “Artificial Consciousness: Utopia or Real Possibility?” Computer 34, no. 7 (July 2001): 24–30. https://doi.org/10.1109/2.933500.
8. Davenport, Thomas, and Ravi Kalakota. “The Potential for Artificial Intelligence in Healthcare.” Future Healthcare Journal 6, no. 2 (June 2019): 94–98. https://doi.org/10.7861/futurehosp.6-2-94.
9. “Digital Health Consumer Survey 2020 | Accenture.” Accessed September 1, 2023. https://www.accenture.com/us-en/insights/health/leaders-make-recent-digital-health-gains-last.
10. Enterprise AI. “What Is Artificial Intelligence and How Does AI Work? TechTarget.” Accessed September 1, 2023. https://www.techtarget.com/searchenterpriseai/definition/AI-Artificial-Intelligence.
11. “How AI And Machine Learning Will Impact The Future Of Healthcare.” Accessed September 1, 2023. https://www.forbes.com/sites/bernardmarr/2022/09/14/how-ai-and-machine-learning-will-impact-the-future-of-healthcare/?sh=4cccdf0647e5.
12. McGaughey, Ewan. “Will Robots Take Your Job? Automaton, Inequality, Full Employment and Law (Presentation Slides).” SSRN Electronic Journal, 2017. https://doi.org/10.2139/ssrn.3044448.
13. Mitchell, Sean. “How AI Will Power the Future of Healthcare.” MUO, January 29, 2023. https://www.makeuseof.com/ai-will-power-future-healthcare/.
14. Schmidhuber, Jürgen. “Deep Learning in Neural Networks: An Overview.” Neural Networks 61 (January 2015): 85–117. https://doi.org/10.1016/j.neunet.2014.09.003.
15. Schulz, Hannes, and Sven Behnke. “Deep Learning: Layer-Wise Learning of Feature Hierarchies.” KI – Künstliche Intelligenz 26, no. 4 (November 2012): 357–63. https://doi.org/10.1007/s13218-012-0198-z.
16. Slowik, Claudia. “Benefits of AI in Healthcare – Usage, Advantages.” Neoteric, November 29, 2021. https://neoteric.eu/blog/benefits-of-ai-in-healthcare/.