Artificial Intelligence is capable of powering a lot of innovations- from driverless cars to creating better video games. AI technology is progressing rapidly, and developments in the sector are influencing a lot of different aspects of our life.
AI also finds applications in healthcare. It automates processes and reduces human mistakes, helps create virtual health assistants, and assists in the detection and prevention of diseases.
In this blog, let’s look at how AI can be used in the fight against cancer.
(Recommended blog - 4 Uses of AI In Cardiology )
Why Oncology Needs AI
Approximately 19.3 million new cancer cases worldwide were reported in 2020, and the year saw almost 10 million cancer deaths. 11.7% of this is cases of female breast cancer, making it the most common kind, followed by lung cancer at 11.4%, which also causes the most deaths. It is projected that rates will increase greatly in the coming years, with a predicted 47% rise in the numbers by 2040.
Cancer is undoubtedly a menace upon the world, which, despite seeing great advancements in diagnostic and treatment methods, remains unconquered. It is still considered a death sentence by most.
In recent times, AI has proven to be a useful tool against this formidable foe, finding applications in all areas of oncology from screening to treatment.
Artificial Intelligence, and particularly its subset Machine Learning, can process large amounts of data and find patterns and hidden characteristics in them. AI can find subtle changes undetectable by human eyes- even well-trained ones.
Deep Learning systems work beyond pre-defined constraints and look at several different data sets, and so can even figure out cancer symptoms previously unheard of.
All of this means that the use of AI translates into the following results when fighting cancer -
If there is to be any hope of reducing the cancer burden on the world, AI needs to be put to use in all aspects of cancer detection, diagnosis, and treatment.
(Recommended blog - Artificial Intelligence in Healthcare: Applications and Threats )
Applications of AI in Cancer Detection and Treatment
Applications of AI have been in effect in the detection and treatment of cancer for some time. As for more advanced techniques, some of them are currently being put into practice, but most are still either implemented locally or still not moved beyond the research phase.
Let’s look at the different areas of application of artificial intelligence and its subsets, and instances of such applications -
Applications of AI in Oncology
Detecting cancer in its early stages drastically changes the survival rates in those affected. This is one of the areas where AI plays an important role. The image analysis capabilities of machine learning and deep learning to detect cancer before it becomes obvious to human specialists.
One research by Mark Schiffman on cervical cancer took 25 years worth of data showing cervical images showing cancer development, and used AI to analyze the data.
The algorithm could identify changes in the images with double the accuracy of any human specialist. This could result in the algorithm being able to predict precancerous markers before even 7 years.
Similarly, studies show AI systems being able to recognize the early stages of lung cancer. Lung cancer is usually detected in the late stages, which makes it very hard to treat with low survival rates. When it comes to breast cancer, mammograms, and even genetic tests used for early screening, are highly inaccurate. Here AI algorithms can drastically reduce the rate of false positives and negatives.
Cancer detection performed routinely and on a larger scale could be beneficial to a large portion of the population. Population screening has always been an idea too impractical to put into practice, but with AI, performing screening on a large scale becomes simpler.
A 2012 analysis showed that offering cancer screening- specifically for lung cancer- as a part of insurance benefits would be both economical and life-saving. With the power of AI, the costs go down even more.
Scientists from MIT use a deep learning system that shows even considerations of inclusivity- the “Mirai” algorithm to study breast cancer. With this one can model possible cancer risks in the future using multiple datasets, and has proven accurate to great degrees.
Large scale cancer screening programs of the future, powered by AI, could thus drastically improve the healthcare conditions of the population as a whole.
Although early detection of cancer is the most tricky part, diagnosing and classifying cancer can be just as complicated sometimes. Rightly diagnosing tumours is important to formulate the right kinds of treatment plans for patients.
This includes diagnosing anatomic anomalies of tumours as well as their genomic characteristics. Machine learning algorithms can be trained to identify genetic mutations and make diagnoses based on MRI images.
This is especially useful for things like brain tumours, where non-invasive methods are required for diagnostics.
A 2018 study created a computer program with the ability to differentiate between the two most common forms of lung cancer. Doctors often struggle to make a distinction between the two which require different types of care and treatment.
New York-based startup Paige uses machine learning to make a faster and more accurate cancer diagnosis. By helping doctors make better, faster, more accurate and more efficient decisions, ultimately helps make healthcare more accessible and affordable.
Artificial intelligence can be used to develop detailed physiological models for drug testing. In drug development, any drug, before being tested on humans, has to pass tests involving disease models.
AI tools can be used in the creation and classification of disease models and make drug development easier. For instance, an ML-powered tool called Disease Model Finder helps sift through data and find the right disease model.
AI can also help understand how drug molecules interact with and affect human cells. Bio Model Analyzer from Microsoft lets biologists model cellular networks and cell interactions. This can be used for studying interactions between cellular systems and drug molecules, and thus aid in drug development.
Predicting Treatment Efficacy
Artificial intelligence can be used to predict the efficacy and potential pitfalls of different treatment regimes.
A 2020 paper details these uses- AI can help in identifying tumour neoantigens and improving the effectiveness of tumour immunotherapy, guide radiologists in mapping target areas, automate planning of radiation treatment programs, assist in optimizing chemotherapy regimens by predicting the tolerance of chemotherapy drugs, help doctors with decision making, avoid needless surgeries, and thus help improve treatment plans.
Especially with immunotherapy, which is a relatively new tool, AI can be used to understand which patients will respond well to immunotherapy and assign alternatives otherwise.
This can be extended to create personalized treatment regimes suited to patients so that they can have the best chance possible against cancer. Microsoft’s Project InnerEye uses computer vision and machine learning to analyze medical images and to deliver precise, personalized healthcare using the results.
(Recommended blog - 8 Popular Computer Vision Applications )
Analysis of patient data, cancer statistics, and tumour features can be automatically done by algorithms saving time on manual labour. Real-time cancer data on various populations can thus be obtained. Data management can be done more efficiently, enabling patients to connect with resources more easily.
Automating pathology is a possible target that can be achieved with the help of artificial intelligence. With this, cancer care can be made several times more accessible to the population.
Modern technologies like AI- and others like nanotechnology and epigenetics- have the potential to make the complicated business of cancer prevention and detection a little easier.
(Recommended blog - Top 9 Healthcare Technologies )
Cancer rates worldwide are projected to drastically increase in the coming years, but developments like these could very well change that.