The era of AI in oncology: cautious optimism

In 2020, the leading journal in research techniques Nature Medicine highlighted artificial intelligence (AI) and its potential application in cancer diagnosis and monitoring as one of the most significant recent milestones in oncology medicine.

Today, two years later, – this algorithm-based technique created for machines to develop human-like capabilities – stands as the technology of choice due to its potential to revolutionize cancer research by using machine learning models to search medical data and uncover insights that help improve health outcomes and patient experience.

However, at the same time that many oncologists are gushing with praise, many others are raising questions about its use in clinical practice. The solution? Cautious optimism.

Applications in oncology

The COVID-19 pandemic and the challenges it brought with it forced healthcare systems in different parts of the world to seek alternative solutions to treat their patients more quickly and efficiently, finding machine learning and AI as a support tool, which offer countless applications, but are still at a very early stage.

This was explained by expert doctors at the American Society of Oncology (ASCO) annual meeting in Chicago-which the LLYC Healthcare team was able to attend-and which highlighted some of the most common functions applied to oncology medicine as well as the challenges it poses.

One of these is the analysis of images such as CT scans, X-rays, MRIs and other images in search of lesions or other findings that an expert radiologist might miss. So much so that a pioneering study published in 2017 that analyzed more than 129,450 clinical images of skin diseases to classify lesions had a very encouraging result as the system’s accuracy in detecting melanomas and malignant carcinomas matched that of professional dermatologists.

Another interesting application of AI in oncology is the potential to make clinical trials more productive and inclusive, given that in the first phase researchers spend a lot of time and resources assigning medical codes to patient outcomes and updating relevant data. With the application of AI, this process can be accelerated as the machine can perform a faster and more intelligent search for these results.

In addition, recently Stanford biomedical data professor James Zou demonstrated that AI can help design appropriate eligibility rules for clinical trials. With his spaceship-named algorithm, Trial Pathfinder, he managed to combine electronic medical records and examine the details that allow some patients to be eligible for a trial and other patients to be excluded, doubling the number of potential participants and expanding the pool to include more women, minorities and older patients.

Finally, this tool can be a major breakthrough and cost-saver for pharmaceutical companies that invest millions of dollars and years of research into developing new drugs. With this big data analytics technology on their side, they can find key information extracted from millions of scientific articles that can be used to develop new treatments or find two or more existing drugs with synergistic effects, allowing pharmaceutical companies to experiment with new drug combinations, called “drug repositioning.”

A revolution, but in a measured way

Although it seems that this technology has been in our lives forever because we are accustomed to using predictive algorithms in our day-to-day lives like a GPS to get to a hotel or Apple’s SIRI application when we want to know about the weather forecast for the next day, the reality is that there is a lot of skepticism around its practice applied to medicine.

Doubt and criticism is more valid than ever because when it comes to leaving in the hands of robots the decision making that affect human beings, there is much to think about, since there may be situations in which the tool does not work well, suggesting, for example, an incorrect dosage of a drug or not applying an algorithm well, having dire consequences for the patient’s health.

For this reason, efforts to educate physicians about AI in order to integrate the use of technology into our practices are essential. Only by establishing specific measures and regulations such as ensuring that professional organizations have taken steps to evaluate algorithms in practice; obtaining quality assurance by Drug Agencies or knowing how the law will assign liability in case of malpractice, as highlighted in the journal JAMA, will we get both physicians and hospitals to widely consider using AI-based products.

In short, this is a technology with enormous potential in the field of medicine and the fight against cancer, but it will only succeed if those who use it prioritize ethics and responsibility above all else.