Accesa logo dark

How is AI Shaping the Future of Conventional Medicine?

Learn about the impact and future of AI in medicine. Is the human touch essential in healthcare, or is it safe to assume that AI will cover everything perfectly?

How is AI Shaping the Future of Conventional Medicine?

Written by Paul Donea (Junior Data & AI Engineer), in the February 2024 issue of Today Software Magazine. Read the article in Romanian here.

In the ever-evolving IT landscape we are going through globally, the trend of using Artificial Intelligence (AI) to streamline activities and processes where the human factor has proven to have considerable limitations increases in numerous areas. Inevitably, the accelerated expansion of this trend and the significant recent discoveries in the technology spectrum could not leave aside a vital area for humanity, such as medicine.

A staggering number of cases have been reported where poor medical performance due to the absolute management of processes by human resources have had undesirable consequences. The integration of Artificial Intelligence with its boundless capacity to process data and deliver results in much shorter timeframes into conventional medicine is entirely changing the way the patients of the future will be diagnosed and treated.

At the same time, the traditional barriers of medicine are long since forgotten and new horizons of healing are emerging so that every sufferer's hopes become a happy reality. At first sight, we can say that the combination of the two fields can only bring remarkable benefits and successes, but if we look deeper, considerable risks can be glimpsed. Identifying and minimising them is the main challenge of the coming decades.

How is AI Shaping the Future of Conventional Medicine_ -3-.png

In this article, we will explore the most important benefits of AI in medicine and examine the most worrying risks.

First of all, it is imperative to briefly describe how Artificial Intelligence can be used in medicine, starting from its very definition. If we were to look at it from a strictly practical perspective, we could say that AI is an entity with a heightened cognitive function that allows it to process enormous amounts of data that it transforms into information and results to be used in a decision-making process.

Having said that, it is much easier to understand the particular benefits it brings to medicine, a field that essentially relies on colossal volumes of data and especially on interpreting these data in concrete situations. Based on those interpretations, treatments are developed with the final scope not only to cure but also to prevent future illnesses.

So, thanks to the rapid technological advances in medicine, Artificial Intelligence has almost naturally become an integrated part of the medical universe. Its reach is continually expanding at an accelerating pace dictated by the challenges and the needs of today's patients.

Main Advantages

Predominantly, machine learning algorithms, especially those characterised by deep learning, are used for the diagnosis and detection of diseases, monitoring and creation of personalised treatments, medical imaging and for a much faster development of new drugs by automating data collection and testing processes.

Unlike doctors, algorithms do not need to rest, so they can analyse continuously, avoiding errors inherent in human activity during diagnosis.

Machine learning models are used to monitor the patient’s vital signs, alert medical staff when certain risk factors occur, and concurrently collect and process data to detect various conditions.

For example, one such algorithm was developed and later used to predict with 78% accuracy the presence of severe sepsis in babies born prematurely. Deep learning algorithms are also increasingly being used to detect arterial tumours by analysing heart rate and blood pressure data.

The creation of personalised treatments has seen a major improvement with the introduction of deep learning models that not only better capture the patients' preferences but also the risk factors they face in their environment. The proposed treatments thus became much more effective and easier to apply.

Managing large numbers of medical images, a challenge in the past has been solved using artificial neural networks that identify the presence of cancerous tumours as accurately as human radiologists but much faster.

If we are talking about the development of new medicines, we can say that this industry has undergone a real revolution with the introduction of new technology in traditional processes. The main tangible benefits of this new technique are a reduction in the cost and time taken to develop a new drug and a greater ability to analyse the chemical and biological factors leading to the final drug formulation.

Two of the most famous AI algorithms used in conventional medicine are DLAD and LYNA. DLAD (deep learning-based automated detection) was developed in South Korea at Seoul University Hospital to analyse several patients' chest X-rays and detect abnormal multiplication of potentially cancerous cells.

DLAD’s performance was compared with that of the hospital's eighteen best diagnosticians, with the algorithm demonstrating better abilities than seventeen of them.

Researchers at Google AI Healthcare created the LYNA (Lymph Node Assistant) algorithm, which analyses histological slides of tissue samples from lymph nodes to identify metastatic breast cancer tumours.

LYNA was tested on two different data sets and correctly identified samples as cancerous or non-cancerous with an amazing 99% accuracy.

Risks

However, it would be unfair for this analysis not also to highlight the main risks arising from this transition to technology and automation in the medical world. Certainly, both healthcare professionals and patients need to be properly informed about these possible risks so that mitigating them is a process in which both sides are involved.

At the time of the introduction of machine learning applications, it was believed that it would considerably alleviate the biases in the medical system due to the subjectivity of doctors in certain situations. The common assumption was that this would lead to a better, safer and fairer diagnostic process. However, it turned out that depending on the nature of the algorithm and the way it was constructed, the data could be processed in such a way that the results did not have the expected degree of objectivity.

Algorithmic biases are determined by the reflection of the author's ethical perceptions in the way of writing and the reasoning behind it. Also of particular importance in this whole process is the quality of the working data.

There are many situations where the training data are either insufficient or come from medical environments that do not coincide with the one where the algorithm will be used. The influencing factors of the two environments are not at all similar.

Algorithmic opacity is another major risk. This is characterised by the inability of the medical team to use the algorithm to understand how it works and how it is reasoning. Thus, the errors provided by the algorithm cannot be traced back to their cause, which calls into question its entire ability to provide relevant solutions and makes its performance impossible to determine.

There are many explanations for this phenomenon. One of them is related to trade secrecy and patent secrecy of the developer. Another is highlighted by the inability of medical staff to understand various automated rationing techniques and their translation into the classical way of working.

All this uncertainty goes by the name of "algorithmic black box" in which inputs and outputs are visible and easy to understand but the end-to-end process cannot be explained.

Last but not least, AI raises concerns about patient privacy and data usage. Personal data stored in the healthcare universe has high value for both research and business.

In 2017, 15% of global data breaches came from the healthcare industry, with a record high in 2021 when 45 million people were affected.

In addition, many AI technologies end up being owned and controlled by private entities, which means that such corporations play a larger than usual role in obtaining, using, and protecting patients' health information. This raises new privacy issues related to implementation and data security.

In the specialised literature, significant attention is paid to developing privacy-preserving techniques and overcoming issues that impede the adoption of AI in a real clinical setting. With the technical tools needed to merge large collections of data and distribute queries across disparate databases, researchers will need to use data anonymisation methods. This ability to anonymise data may be compromised or even negated by new algorithms that have successfully re-identified such data.

Conclusions

Although the benefits of Artificial Intelligence in the medical field have helped to accelerate its evolution, there are issues that need to be brought to the attention of the general public, and solutions need to be found so that the use of the technology becomes common practice.

On a downward slope of demographic growth among Western nations, the automation of processes by new methods generated by technological evolution is not only a normality of progress but, above all, a necessity generated by the natural reduction in the number of specialists in the medical field.

Therefore, whether or not the replacement of certain human activities by Artificial Intelligence is desired as proof of the expansion of technology, it will become a necessity in a society where the number of positions in the medical field will be significantly greater than the number of individuals ready to fill them.

How is AI Shaping the Future of Conventional Medicine_ -2-.png

Few debates, however, focus on a major risk created by AI in the medical spectrum, that of the dehumanisation of medical practice.

Empathy, sympathy, and compassion are uniquely human emotions and are currently inaccessible to AI in a transmissible way. These emotions have proven to have significant importance in the healing process. Not having this humanitarian side can profoundly affect the patient’s recovery in ways that cannot be determined yet.

Finally, we can say with certainty that conventional medicine will change entirely under the influence of artificial intelligence, which entails formidable scientific progress that would otherwise be impossible without this technology. The question that arises from all this change is: how much does the lack of human empathy affect healing?