Introduction

Imagine putting a patient’s history and examination findings into a chatbot and getting instant information on the diagnosis and plan of management. This is one of the many ways in which Artificial Intelligence (AI) is expected to change the patient care landscape in the coming years. AI is the ability of a computer or computer-controlled robots to perform tasks that are usually associated with intelligent human beings.¹

With the advent of AI, it is possible to develop systems that can reason and learn as humans do. AI is transforming healthcare through enhanced diagnosis, targeted therapy and disease prevention strategies. Thus, the integration of AI into healthcare is expected to improve patient experiences and health outcomes. In this article, we will consider the applications of AI in healthcare, the benefits and the challenges with its use.

Applications of AI in healthcare

AI has various practical applications in the realm of healthcare. 

Diagnostics

Artificial intelligence approaches, including Machine Learning (ML) and Deep Learning (DL) algorithms, are widely employed in the prediction and diagnosis of numerous diseases, especially those that are diagnosed using imaging.⁴

Machine learning (ML) is the process by which computer programs utilise data to identify patterns within the data and use these patterns to predict a relationship between the input and output. Deep learning (DL) is a subset of machine learning that allows the analysis of massive data. 

Images generated through ultrasound, X-rays, computed tomography, and magnetic resonance imaging can all be auto-analysed using AI. AI has also been integrated into Digital Pathology to enable the review of pathology slides. The Food and Drug Administration (FDA) has recently approved an automated whole slide imaging (WSI) scanner for primary diagnosis. This can generate high-resolution images from pathology slides. 

The use of AI in pathology provides an objective assessment of cell characteristics and helps streamline clinical pathology workflows. Thus, AI provides support to pathologists by speeding up reporting times.⁵ AI enables pathologists to diagnose rare diseases by providing a repository of histopathology images for comparison Content-based image retrieval (CBIR) entails searching an image database to find similar visual content images to a given query image.

Disease prediction

AI models facilitate the analysis of large datasets for disease prediction. AI is helpful in the early detection of chronic diseases like diabetes and cardiovascular diseases. It is also quite helpful in the detection of cancers. ² CancerSEEK is an AI-powered model that can detect eight common cancer types through analysis of cell-free DNA.

In addition, the integration of multiple data sources such as genomic sequences, environmental factors, and social behaviour patterns helps in infectious disease prediction.³ In the event of a pandemic, advanced machine learning algorithms could analyse trends in infection and mortality rates, providing accurate predictions about the spread and severity of the pandemic. In the management of the COVID-19 pandemic, AI found its usefulness mostly in the analysis of lung images, especially in complex cases. Many AI-enabled models are being developed to help tackle future pandemics.

Personalised medicine

Personalised medicine aims to generate precise treatments and prevention strategies and challenges the concept of “one-size-fits-all”. AI can be used to comprehensively analyse multiple disease-causing factors such as genetics, environment and lifestyle to deliver bespoke treatment options specifically tailored to individual needs. This enables medications and dosages to be tailored to the patient's genotype. This optimises treatment success and improves patient satisfaction.⁶

AI in precision medicine can guide clinicians in treatment decisions by predicting responses to therapy. In the management of cancers, AI can be used to predict response to treatment based on findings on histopathologic analysis. An AI-based serum proteomics test model has been developed to predict response to immunotherapy in patients with metastatic melanoma.

Optimising treatment protocols

By leveraging AI algorithms, healthcare providers can optimize medication dosages tailored to individual patients and predict potential adverse drug events, thereby reducing risks and improving patient care. This is very useful in drugs with low therapeutic index.⁷ 

Patient monitoring

Combining wearable technology with AI allows for real-time data collection for monitoring patient health. 

• AI has been integrated into remote patient monitoring systems (RPM). AI systems possess the ability to analyse data like vital signs and more. This is collected through patient-reported information, wearable devices and sensors. Personal baselines are established from these data and deviations from the norm are easily determined. It reduces the need for in-patient visits, reducing the burden on the patients and the healthcare system. The use, however, remains limited due to the limited availability in resource-poor nations⁸
• Beyond detecting deviations from the norm, remote patient monitoring systems also predict potential illnesses based on available historical data
• It also provides reminders and support to patients to enhance medication adherence

Administrative efficiency

AI increases administrative efficiency through the automation of administrative tasks in the following ways:

• Aids in healthcare recruitment by scanning resumes, conducting the initial assessment and shortlisting candidates
• AI can be used in the scheduling of duties thereby reducing staff friction and improving staff satisfaction
• Generative AI brings efficiency to clinical operations by aiding healthcare providers in administrative tasks like form-filling and note-taking⁶
• Enhances electronic medical recording by aiding the assessment and update of patient information
• Simplifies billing and claim processing
• Improves customer relations with the use of chatbots to respond to patients' enquiries
• Streamlines supply chain logistics by forecasting demand which helps to reduce out-of-stock and wastage

Benefits of AI in patient care

Many advantages of AI have been declared in literature, though the evidence is mainly from observational studies.⁹ It is expected that this evidence will be validated over time by randomised controlled studies. The benefits include:

Improved accuracy and speed in diagnosis

• AI improves the accuracy of medical diagnosis. This is very useful in cancer care. In breast cancer care, grading algorithms can be more objective, consistent, and clear about the prognosis, complementing human expertise. Human diagnosis may vary due to the complexity of cases, and AI helps to standardise and support these critical decisions
• Because of the ability to analyse large volumes of data, AI empowers healthcare providers to make data-driven decisions

Enhanced patient outcomes

• By predicting the risk of chronic and infectious diseases
• By making faster and more accurate diagnosis
• Personalised treatment plans empower patients to actively participate in their healthcare by providing insights into how their choices impact their health outcomes⁶ 
• Reducing hospital stays due to increased efficiency which results from improved workflow and leads to a better treatment approach
• Improving the monitoring of chronic diseases
• Streamlining of processes at the healthcare facilities, improving access to healthcare
• Reducing staff burnout and improved healthcare delivery

Increased access to healthcare

• AI improves access to healthcare by enhancing efficiency which improves the availability of health professionals to attend to more patients
• It increases the ability to provide diagnosis and treatment plans for complex cases
• AI detects early signs of disease thereby improving access to early treatment for a great number of people

Cost reduction and efficiency

• AI streamlines tasks, improving efficiency and reducing costs
• The automation of processes at health facilities reduces errors leading to cost savings
• AI risk prediction and early treatment of chronic diseases reduce the overall treatment cost and burden
• The use of AI reduced hospital stays, resulting in a reduction in healthcare expenditure

Challenges and limitations of AI use in healthcare

Before integrating artificial intelligence within the healthcare system, policymakers, practitioners and healthcare stakeholders should take into consideration the medical ethical principles of autonomy, beneficence, non-maleficence, and justice.¹⁰ 

Unfortunately, current ethical guidelines do not take into consideration the integration of AI in healthcare and this poses some ethical dilemmas with the use of AI. The challenges of the use of AI in medicine include:

Data quality and integration

The quality of AI data output depends on the quality of the data input. It is not uncommon to find disaggregated and incomplete data in health institutions. Accurate and comprehensive data are essential for AI to generate meaningful treatment recommendations. Therefore ensuring data accuracy and compatibility across various systems is crucial.⁶

Data privacy and security concerns

Machine learning and deep learning require the use of large datasets. Hence, large-scale electronic patient data are used in training AI models. The volume of data used has raised privacy concerns. The argument in favour is that data is de-identified and poses no harm to the patient. The challenge, however, is the ability of some AI models to put information together to re-identify a person. This can cause a breach of confidentiality.

Bias in AI algorithms

Compared to human systems that have well-explained processes, there is often difficulty in explaining how AI algorithms arrive at their conclusions. This is termed the “black box problem”. Therefore, AI algorithms need rigorous, multi-institutional validation before they can be clinically implemented.⁵ This would delay the full integration into the healthcare delivery system.

Informed consent and transparency

AI involves the pooling of large data from multiple sources. This involves data sharing and the movement of electronic medical records between providers, sometimes without individual patient consent. This is a source of ethical concern because of a breach of patient rights. 

Lack of accountability

One of the major problems with the use of AI in healthcare is a lack of accountability. The healthcare system requires that someone be responsible for management decisions. This is difficult to determine with AI. Who is to be held responsible when something goes wrong with the use of AI?

Difficulty of Integration with existing systems

The existing rules on medical ethics and practice did not take into consideration the development of AI, which thus results in areas of conflict. There is a need to revise the existing laws to accommodate the use of AI. The expedited development of standard guidelines for the ethical use of AI is imperative. In addition, health workers need to be trained in the use of AI to enable proper integration.

High implementation costs

Integration of AI into the healthcare system is capital-intensive. This may not be possible to do for small providers, especially in low-income countries.

Job losses

There are fears of possible job losses in the health sector due to the use of AI. This is not entirely true as responsible use of AI requires a human interface. However, over time, healthcare workers who are knowledgeable in the use of AI may replace those who are not.

Examples of successful implementation of AI in healthcare

• Tidalhealth Peninsula Regional Hospital, Maryland has implemented a clinical decision support software IBM Micromedex with Watson. This AI solution combines clinical decision support and AI with patients’ electronic health records. This makes it easier to access useful information thereby cutting down time spent on patient review
• The Mayo Clinic in conjunction with Google Clouds has developed an AI solution that helps in tracking changes due to chronic diseases like polycystic kidney disease and also in assessing breast cancer risk
• Portal Telemedicina, a rural healthcare provider, has teamed up with Google Clouds to provide an AI solution that helps in data aggregation, storage and analysis. This has caused a 20% reduction in admissions and a 5% reduction in healthcare costs
• Copilot, a digital assistance tool was launched by Nabla, a French tech startup to assist doctor and patient interaction. The tool converts doctor-patient conversations into prescriptions, summaries and confirmation letters
• Moderna, the producer of one of the COVID-19 vaccines, leveraged AI to facilitate the process. They conducted the majority of their preclinical quality check analysis with AI

Summary

The healthcare system is increasingly challenged by the complexity of diseases and a reduction in manpower, highlighting the need for systems that can facilitate healthcare delivery. The integration of AI in healthcare shows great promise due to its potential to accelerate and enhance patient care. To maximize AI's role, thorough evaluation is essential to ensure its intelligent and responsible implementation.

Given the need for human oversight, it's crucial to train healthcare workers to adapt to these changes by fostering a transformative shift in mindset. Responsible integration of AI, coupled with supervised use by healthcare professionals, can prevent errors and enhance its utility.