Dra. Magda Cordeiro

Medical Oncologist 

Artificial intelligence in oncology

HPA Magazine 22 // 2024

What is Artificial Intelligence?
Artificial intelligence (AI) is the area of computer science dedicated to the development of systems capable of performing tasks normally associated with human intelligence, such as learning, reasoning, problem-solving, perception, understanding, and interaction with the environment.
Essentially, it aims to create machines capable of thinking and learning like humans, and there are 6 areas in which it is designed to achieve this.

Machine Learning: Algorithms that allow computers to learn through analysing data, which is later used to make predictions or decisions.
Deep Learning: Uses deep artificial neural networks to identify complex patterns in large volumes of data.
Natural Language Processing: Systems designed to understand, interpret the context of the conversation and respond to human language in a natural way; such as, for example, automatic translation and chatbots.
Computer Vision: Allows computers to interpret and understand the visual world by reading images and videos; such as object recognition, facial recognition, and medical image analysis.
Robotics: Integrates AI into physical systems to create robots that can perform real-world tasks, such as automatic navigation, guided manipulation of objects, or interaction with humans.
Expert Systems: Programs that use a knowledge base made up of facts and rules, which learn and allow them to solve complex problems as well as imitate the decision-making capacity of a human expert in a specific domain.

What are the applications of artificial intelligence?
AI is currently applied in our everyday lives, across a wide range of sectors.
Health: Diagnosis, drug development, robotic surgery, and treatment customization.
Finance: Fraud detection, negotiation, and risk analysis.
Transport: Automatic vehicles, route optimization, and traffic management.
Services: Virtual assistants, scheduling, and analyzing customer feedback.
Manufacturing: Process automation, predictive maintenance, and quality control.
Entertainment: Content recommendation, trend assessment, and management, AI-assisted music creation, painting, and scripting.

Artificial Intelligence in Oncology
It is therefore irrefutable that AI is not only upending our everyday lives but has also transformed medicine and offers equally promising advances in the field of oncology. This technological revolution brings us great expectations, especially with regard to screening/early detection, diagnosis, and personalized cancer treatment programs.
Early Detection and Diagnosis. Detecting cancer early is crucial to increasing individuals' survival rates. AI, through deep learning algorithms, is capable of analysing medical images with unprecedented speed and sensitivity. In this way, through in-depth analysis, it is capable of identifying anomalies in histological or imaging exams, with greater precision than that of more experienced specialists, increasing diagnostic accuracy and allowing for more targeted treatment.
Precision Oncology. Precision oncology consists of an approach that considers individual genetic variability, the environment, and lifestyle of each individual, in the personalized diagnosis and treatment of cancer patients. Using a massive amount of data and pattern analysis, based on genetic and clinical information, AI can help identify specific mutations that can be targeted by personalized therapies as well as select, among available treatments, those that may be most effective for a particular patient.
Development of New Drugs. The development of new medicines is, as we know, a long and expensive process. AI can speed up this journey, particularly helping in the statistical processing of the immense amount of data normally involved in pre-clinical studies and in the initial phases of implementation, facilitating the identification of genetic changes and potential promising drugs; which promotes not only cost reduction but also the development of new therapies more quickly.
Assistance in Daily Clinical Practice. AI-based systems can assist oncologists in making complex decisions by integrating, for example, data from multiple sources simultaneously - clinical histories, recent publications, international guidelines.
Another use of AI in oncology is the possibility of using virtual assistants to support patients. AI-based virtual assistants can provide information, monitor symptoms, and offer emotional support to patients. These systems can answer common questions, schedule appointments and reinforce treatment adherence.
Radiation oncology is another area in which AI has been used to help optimize radiotherapy plans, ensuring that the radiation dose is delivered accurately and safely to the tumour, minimizing exposure to healthy tissue.

Challenges and Ethical Considerations of Artificial Intelligence
In essence, the effectiveness of AI systems depends on the quality and diversity of the data used to train those same systems.
In this way, biased data can lead to incorrect diagnoses and inaccurate treatments, with a potentially very negative impact. In this context, it is particularly important to continue to rely on medical expertise to verify, confirm and validate AI conclusions.
From the same perspective, we know that AI analyses data objectively. However, many medical decisions depend on a clinical context, the patient's history, emotions, preferences and other variables that effectively require a subjective interpretation, based on experience and continued clinical practice, which apparently AI is not yet able to discern.
Following this, some of the ethical issues that are most debated revolve around the privacy of patient data and the transparency of the algorithms used.
Likewise, we cannot forget that the implementation of AI systems requires an entire technological infrastructure, training and continuous maintenance and that, on the other hand, their excessive dependence can make it difficult for doctors to make independent decisions.
Finally, it is not to be neglected that the use of AI can have considerable social and economic impacts, especially with regard to automating tasks and making critical decisions, which will now be carried out by machines, potentially making many possibly obsolete jobs.

What measures have been implemented to overcome these challenges?
To overcome bias in data, companies and researchers are currently investing in a more assertive and balanced collection of data, coming from diverse and representative sources in each specific area, reducing possible risks of contamination.
Likewise, advanced techniques are being developed to improve the quality of information, such as removing duplicate or irrelevant data, correcting errors and standardizing formats.
From a transparency perspective, there has been an increasing focus on developing more interpretable and explainable AI models that allow users to understand how decisions are made.
With regard to privacy and security, so-called differential privacy techniques have been developed, which add noise to data in order to prevent individual information from being easily extracted, allowing the analysis of a large amount of data without compromising the individual privacy of each patient. In this context, a valuable tool developed was called Federated Learning, in which the data remains on local devices, there is no need to send it to a centralized server and only the trained models are combined and shared, greatly reducing the risk of exposure.
With regard to regulation, governments and institutions in several countries are currently engaged in developing guidelines to ensure the ethical use of AI. The European Union, for example, is currently working on the AI Law, which aims precisely to regulate its use in different sectors.
Following this trend, many companies and research institutions around the world already have ethics committees focused on reviewing and monitoring AI projects, ensuring that they are aligned with ethical and civil responsibility principles.

In summary
AI is a reality that is revolutionizing our daily lives. Just like in medicine, it is also a promising tool in oncology, especially in early detection, diagnosis, and treatment of cancer. Although it still presents some challenges, measures are being implemented to ensure respect for human rights and guarantee social well-being.
Continuous collaboration between governments, health professionals, scientists, and computer engineers are necessary, along with the ethical development of technology. These partnerships encourage the exchange of knowledge and resources to ensure that advances can benefit all patients in an equitable and safe manner, allowing the personalization of treatments to continue evolving to offer.