Overview

Artificial Intelligence, or AI, has been growing more and more popular lately. It is rapidly evolving and being applied to a variety of uses. Many of us have used AI or AI combined with other technologies.¹ It is becoming a part of our daily lives, such as the use of GPS guidance and generative AI tools, like ChatGPT. AI is also being applied to the field of medicine. Of particular interest is its application in oncology. AI is being used to help doctors with the diagnosis of a variety of cancers, to improve early detection of cancers and to improve treatments. Specifically, it is being used in personalised treatments to improve patient outcomes by trying to match the right patient with the right therapy option as well as monitor the responses to those treatments.

What is AI?

AI is a branch of computer science that uses computers designed to think and solve problems like humans do. Basically, AI is capable of completing tasks that would usually require human intelligence.

AI can be divided into subfields: machine learning, neural networks, and deep learning.

• Machine learning is where computers learn from data and get better over time without being given specific instructions. It usually needs more human input and structured data to learn.
• Neural networks are a type of machine learning inspired by the human brain, where connected parts (nodes) work together to process information. They send signals (data) to each other, like neurons in the brain.
• Deep learning is a more advanced type of neural network with many layers. It can learn from unstructured data, like text and images, in its raw form and automatically distinguish between different data categories. This means it needs less human input and can handle large amounts of data.

AI and oncology

AI can use large amounts of data to support doctors in their decision-making. Some of the applications of AI in oncology include:¹

• Diagnosis: AI can help doctors diagnose cancer more quickly and accurately, for example, by analysing medical images such as computed tomography (CT) or magnetic resonance imaging (MRI).
• Treatment planning: AI can help doctors create more personalised treatment plans based on a patient's medical history and genetics.
• Drug discovery and development: AI can aid the discovery of new drugs or therapies by quickly processing large amounts of scientific data that would otherwise take longer to analyse.
• Patient monitoring: AI can be used to help with monitoring patient’s health and wellbeing remotely. For instance, wearable technologies might be used to monitor if the patient is doing better or worse, allowing for prompt action if they are doing worse.

Governments are increasingly interested in using AI in healthcare as well. Recently, the UK government decided to invest into the NHS to provide AI technology across radiography departments in England. The new technology will analyse scans, such as MRIs and CTs, to help doctors diagnose cancers faster by distinguishing between cancerous and healthy cells. Not only will this help diagnose cancers more efficiently and reduce waiting times for cancer patients, but can also improve radiation treatments by targeting cancerous cells more precisely, thereby reducing the likelihood of damaging healthy organs.

How can AI help personalise cancer treatments?

AI can be a very useful tool when it comes to personalising the treatments for cancer patients. AI models can analyse and interpret large amounts and different kinds of data, including genomics, epigenomic, proteomic, metabolomic and clinical data. It can integrate all these large sets of data to better understand the complexity of the patient’s cancer. ² This data can be then used to predict the severity of the disease and tailor potential treatments to the patients. Some ways that it can help doctors personalise treatments are:

• Prognosing outcomes: AI can be used to predict the risk of developing cancer, the severity of the current disease, how the disease will progress, how will it respond to treatment, as well as how likely cancer is to return after treatment.^2,3
• Treatment recommendations: Doctors need to decide what kind of treatment combination is the best for the specific cancer patient. They need to decide if a specific drug, combination of drugs, radiotherapy and/or surgery is needed. AI can aid the process by integrating large amounts of patient data to provide evidence-based treatment options. For instance, an AI tool called LORIS was shown to accurately predict patient responses to an immune checkpoint blockade therapy (a form of immunotherapy).^4,5
• Assessment of response to treatment: Assessing how a patient is responding to a particular treatment can be quite time consuming and requires a high level of expertise and experience from doctors. AI tools can make this process easier and quicker by automating the assessment of response to treatment.⁴

Examples of current uses of AI in oncology

There are many AI tools or devices that are already being used in oncology. Here are just a few examples:⁶

• Arterys Oncology DL (Tempus, formerly Arterys)): a cloud-based AI medical imaging software that helps doctors diagnose patients by automatically measuring and tracking lesions and nodules in MRI and CT scans.
• Invasive Tumour Detection APP (Visiopharm A/S): helps with predicting how severe the cancer is by distinguishing between invasive and non-invasive tumours based on certain markers.
• ART-Plan.annotate (heraPanacea SAS): can help with delivering more precise radiotherapy by distinguishing between tumour and surrounding tissues.
• Ethos Radiotherapy Treatment (Varian Medical Systems Inc.): helps with managing and monitoring radiotherapy treatment plans.
• Cobas® EZH2 Mutation Test (Roche Molecular System, Inc.): identifies follicular lymphoma patients with an EZH2 mutation, allowing for more specific and targeted treatments.

Are there any issues with using AI in oncology?

While AI is revolutionising healthcare, including oncology, and has the potential to benefit patients in great ways, it is also important to recognise potential issues associated with it. Some of these include: ¹

• Data quality: for an AI tool to make good decisions, it needs high-quality data. However, not all data is high-quality. There needs to be special considerations regarding the quality of the data used to train AI to support more accurate and effective outcomes.
• Data privacy: AI tools use a lot of data that comes directly from the patient. This data is sensitive and should be protected. AI developers need to ensure they maintain patient privacy.
• Clinical adoption: AI is still relatively new and has a long way to obtaining full trust. This includes both clinicians and patients. Using AI also requires training, which can be time consuming for those involved, including healthcare providers and hinder adoption. Moreover, it is important to ensure that AI tools are user-friendly to enhance adoption.

It is also important not to forget the importance of doctors, as AI tools are not able to treat patients on their own and still need doctors' input and interventions.

Summary

AI is a rapidly growing field that has the potential to revolutionise and bring great benefits to the field of oncology. Of particular interest is its use in personalised cancer treatments. AI tools can be used to integrate large amounts of patient data to better understand their risks, the severity of disease, what kind of treatment options might benefit the specific patient the most, as well as their potential response to these treatments. AI has the potential to make everyone’s lives easier, including both healthcare providers and patients. It can help doctors in their decision-making and reduce the time doctors need to spend analysing patient data. As beneficial as AI tools can be, there are some considerations that need to be addressed early on. These include risks to patients' privacy and potential issues with data quality that can then affect the accuracy of AI.