Artificial intelligence is increasingly being used in various areas of health, particularly in the treatment of cancers. What are the latest applications in oncology? We provide some answers.
Artificial intelligence to advance medicine
Artificial intelligence (AI), which first emerged in the 1950s, involves using machines to perform thousands of tasks usually carried out by humans using mathematical algorithms. These machines mimic the brain activity and cognitive processes of human beings. AI can be broken down into four levels. The first two methods are the most widely used, all fields combined. These correspond to straightforward data calculations. The third level takes things further, making it possible to predict results. Lastly, the fourth level uses Machine Learning, whereby AI is capable of learning and progressing by itself.
The applications of AI in the field of medicine are very broad. In particular, AI can help practitioners make decisions, both in terms of reaching a diagnosis and prescribing treatments, recommending the most relevant therapy based on the patient’s situation. AI thereby helps to personalize treatments and sometimes even to predict a disease and its evolution, as well as the chances of success of a drug. It can also help to construct very fine-precision robots or programs, such as those used in computer-assisted surgery.
Examples of recent applications in oncology
In developed countries, cancer is a real public health challenge, with 19.3 million new cases in 2020 worldwide. It is the number one cause of death, with 10 million deaths in the same year.1 Therefore research, particularly AI, is an absolute priority to help prevent and cure cancer.
A team of Japanese scientists conducted a review of recent AI applications in oncology.2 One of the applications identified by the scientists concerned imaging using Deep Learning to improve the detection of cancers. The scientists describe a recent successful example with the classification of dermoscopy images, in which AI was found to be able to annotate skin lesions (including melanoma) with the same precision as expert dermatologists. The team also found similar or even better levels of accuracy in breast cancer screening, with AI achieving superior results to medical specialists in the interpretation of mammograms. In both cases, it is the accumulation of digital images that allows the system to use deep learning to make a pathological diagnosis.
Deep Learning is also able to describe the status of a tumor from pathological data or to assess the expression of biomarkers such as HER2 in order to predict which genes harbor mutations within the tumor tissue. Another interesting application of AI identified by the scientists is in cancer genomics, to study the thousands of possible genetic mutations without any human intervention in order to conduct a review of the literature. With more than 200,000 new articles published in 2019 alone, the lack of human resources to analyze and link genetic mutations is clear. The COSMIC database provided by the Sanger Center, an English genomic research institute, is one example of the technological performance that can be achieved, with more than 9 million mutations extracted from 26,829 papers.
The uses of AI in oncology are promising and will undoubtedly lead to further improvements in medical performance in the coming years. However the inadequate number of medical images and data available to improve training in AI is a not insignificant limitation identified by the research team.
Sung, H, Ferlay, J, Siegel, RL, Laversanne, M, Soerjomataram, I, Jemal, A, Bray, F. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2020. https://doi.org/10.3322/caac.21660