Can Artificial Intelligence (AI) decrease the risk of hospital infections, increase survival rates, and reduce the workload of healthcare professionals? ^1,2

The truth is that AI can bring many benefits to healthcare and even increase the quality of life, decrease costs and waiting times, and provide a personalised follow-up to each person, thus increasing their autonomy and involvement throughout the diagnosis and treatment process. AI cannot replace health professionals but can support them in clinical decision-making and improve quality and readiness.^1,2

According to a report by Grand View Research, the compound annual growth rate in the AI healthcare market segment in the United States is expected to increase by 37.5%3 between 2023 and 2030.

With the increasing development of AI and the exponential increase in investment in recent years, we will see more of these benefits.

Let’s look at the four main categories in healthcare where AI solutions really impact⁴ :

Diagnosis and Prognosis^4,5

In diagnosis, AI increases response times, initiating the necessary treatment earlier, which may significantly improve the patient’s prognosis. Machine learning algorithms can learn to see patterns similar to how doctors see them, but they can respond more quickly than a doctor. Thus, these algorithms can prioritise patients and make a pre-selection, where a physician will later validate the results. This process reduces the workload of healthcare professionals.

Machine Learning algorithms are beneficial in areas where diagnosis is made through images, such as CTs, electrocardiograms, MRIs, dermatoscopy, and retinographies.

Developing predictive models through clinical indicators is also possible, which may be extremely important in minimising invasive examinations or helping in places without access to appropriate healthcare equipment.

Another advantage is to help the physician combine the results of multiple exams to make the diagnosis more robust and complete.

Predictive models can also be used in prognosis, helping the doctor understand the possible course of the disease, future complications, or treatment outcomes. For example, these models can help predict whether a patient will have a low or elevated risk of death after surgery, increasing their chances of survival as the doctor can adjust the treatment in advance.

^{Fig.1- Diagnosis and Prognosis}

Treatment^4,6

Another potential benefit of AI is assisting patients’ treatment or follow-up after diagnosis.

One of the main advantages is personalised medicine. Different patients respond to drugs and treatment schedules differently. So, personalised treatment has an enormous potential to increase patients’ lifespans. This type of medicine also prevents the use of drugs that produce no results in particular patients, reducing costs. With AI, it is possible to discover which clinical indicators of a patient suggest they will have a specific response to a particular therapy. A predictive model that forecasts the patient’s response to treatment can be developed, helping doctors to choose the most appropriate treatment for each person.

Another advantage is the possibility of following up with patients after some treatments or interventions at home. With the help of an AI application where the patients record relevant daily clinical data, their clinical progress can be closely monitored. If there is some reason for alarm, the system can quickly direct the patient to a healthcare unit. This at-home follow-up increases patient comfort and autonomy, reduces the risk of hospital infections, and increases the number of beds available in the healthcare system.

When all processes are automated, AI can instantaneously combine the results of multiple tests and analyses to suggest the best treatment based on the patient’s clinical indicators, which can positively impact treatment and recovery.

In recent years, robots have been developed to assist some patients in keeping their routine, helping them maintain their cognitive and motor skills through auditory, sensory, and visual cues. These robots aim to assist caregivers of people with dementia or Alzheimer’s, simplifying their tasks and increasing the patient’s quality of life. The robots can mitigate negative feelings like sadness and loneliness through the activities they promote (reading magazines and newspapers, playing music or familiar sounds, getting reminders of important dates or events, storing photographs, connecting with family and friends, etc.).

^{Fig.2- Treatment}

Drug Development^4,7

Developing drugs is a notoriously expensive and time-consuming process, and this is a problem for the pharmaceutical industry. One of the slowest phases is the analytical process, which can be accelerated with the help of AI. AI can be of use in the following four stages:

• Identifying the target molecules for intervention: To develop a drug, it is necessary to understand the biological origin of a disease to identify suitable targets where the drug can be effective and its resistance mechanisms.
• Discovering drug candidates: A compound that can interact with the identified target molecule in the desired way must be found. It is essential to be careful with the potential compounds and their effect on the target (affinity), not to mention their off-target side effects (toxicity).
• Speeding up clinical trials: Choosing suitable candidates for clinical trials may influence their duration. AI can help identify patterns that separate good candidates from wrong as well as ensure the correct distribution for groups of trial participants.
• Finding Biomarkers for diagnosis: Treating diseases is only possible when these have been correctly identified. Some screening methods are expensive and involve complex equipment. With the help of AI, it is possible to find biomarkers (molecules typically found in human blood) that provide absolute certainty as to whether a patient has a disease and to pinpoint its progression.

^{Fig.3- Drug Development}

Gene Editing^4,8

Genome editing allows the modification of the DNA of organisms. Genetic material can be added, removed, or changed. This technique identifies a target DNA sequence through a short RNA structure created by the researcher. RNA is attached to an enzyme that acts as a cutting mechanism and cuts that specific part. After the cut is done, a DNA repair mechanism is used. One of the most widely used approaches today is CRISPR-Cas9, which is based on a defence system that bacteria have against viruses.

One problem is that this technique relies on a sequence of the short guide RNA to target and edit a specific DNA location, but the guide RNA can fit multiple DNA locations, which can lead to unintended side effects. Machine Learning models have been proven to produce the best results when predicting the best RNA sequence. AI can significantly speed up the development of guide RNA, which will result in increased growth of this technique and, consequently, a reduction in costs.

^{Fig.4- Gene Editing}

Despite all the advantages AI can bring, particular attention needs to be paid to some of the existing risks to maximise its full potential. ^9,10

Some of the most critical risks to mitigate are^9,10 :

• Errors in the algorithm: AI systems are prone to mistakes that may harm the patient or cause other significant problems. It is necessary to have a large volume of data, and it must be real and reliable to minimise these errors.
• Privacy issues: As with any technology, hackers could potentially access systems and steal data. With the correct security protocols, these breaches should be less likely to occur. Protecting patients from data leaks is also integral to the safe use of AI in medicine.
• Bias: AI is not immune to bias. The faintest hint of discrimination in the training data is reflected in the results. Certain groups with distinctive characteristics (gender, ethnicity, race, convictions, etc.) may be vulnerable to wrong diagnoses/treatments if the algorithm has not considered them carefully.

To implement AI in healthcare, it is also vital to break down some barriers and face some challenges, such as:

• Data digitisation: Without massive amounts of data fed into AI systems, it is impossible to get reliable results. Therefore, it is essential to source high-quality healthcare data in digital format.
• Regular data updates: It is necessary that the data provided is always up to date to increase the performance of the algorithms and always get the best results.
• Expert interventions: machine learning algorithms should be developed with the knowledge of medical experts for better optimisation. In addition, the medical community must believe in these algorithms and be aware that AI is a just support tool that will never replace their role.

In conclusion, with all its benefits and dangers, the use of AI in healthcare can help improve the quality of healthcare services provided, reduce costs, lower waiting times, and support healthcare professionals in improving and saving lives.

References:

1. How will generative AI impact healthcare? (n.d.). World Economic Forum. https://www.weforum.org/agenda/2023/05/how-will-generative-ai-impact-healthcare/
2. Artificial Intelligence in Healthcare | EIT Health. (2022, June 15). EIT Health. https://eithealth.eu/think-tank-topic/artificial-intelligence-in-healthcare/‌
3. Artificial Intelligence in Healthcare Market Size Report, 2019-2025. (n.d.). https://www.grandviewresearch.com/industry-analysis/artificial-intelligence-ai-healthcare-market
4. Schmitt, M. (2019). Artificial Intelligence in Medicine | The Top 4 Applications. com. https://www.datarevenue.com/en-blog/artificial-intelligence-in-medicine
5. Rajpurkar, P., Chen, E., Banerjee, O., & Topol, E. J. (2022). AI in health and medicine. Nature medicine, 28(1), 31–38. https://doi.org/10.1038/s41591-021-01614-0
6. Johnson, K. B., Wei, W. Q., Weeraratne, D., Frisse, M. E., Misulis, K., Rhee, K., Zhao, J., & Snowdon, J. L. (2021). Precision Medicine, AI, and the Future of Personalized Health Care. Clinical and translational science, 14(1), 86–93. https://doi.org/10.1111/cts.12884
7. Paul, D., Sanap, G., Shenoy, S., Kalyane, D., Kalia, K., & Tekade, R. K. (2021). Artificial intelligence in drug discovery and development. Drug discovery today, 26(1), 80–93. https://doi.org/10.1016/j.drudis.2020.10.010
8. Wang, S. W., Gao, C., Zheng, Y. M., Yi, L., Lu, J. C., Huang, X. Y., Cai, J. B., Zhang, P. F., Cui, Y. H., & Ke, A. W. (2022). Current applications and future perspective of CRISPR/Cas9 gene editing in cancer. Molecular cancer, 21(1), 57. https://doi.org/10.1186/s12943-022-01518-8
9. Mills, T. (n.d.). Council Post: The Risks And Benefits Of AI In Medicine. Forbes. Retrieved January 3, 2023, from https://www.forbes.com/sites/forbestechcouncil/2021/03/23/the-risks-and-benefits-of-ai-in-medicine/?sh=7de7bf5b58d8
10. Powell,A. (2020, November 11) Risk and Benefits of an AI Revolution in Medicine. Harvard Gazette https://news.harvard.edu/gazette/story/2020/11/risks-and-benefits-of-an-ai-revolution-in-medicine/