Artificial Intelligence: A Threat to Us or to Cancer?

Reading time: 5 minutes

Aya Elmeligy

Introduction

Cancer is a constantly evolving illness. This infamous disease crosses everyone’s mind, whether through fear of developing it or worry of never being rid of it. The key to overcoming this is early diagnosis and more effective treatments that are personalized to the patient. Literature regarding the average efficacy of any cancer drug uncovered figures of only 25% success rate, leaving behind a substantial 75% of patients left in pain due to side-effects or overdoses (1). These disheartening statistics are due to an inability of some drugs to reach the cancer, increased multidrug resistance, and treatment toxicity which limits the dose a patient receives. Artificial Intelligence (AI) could be the solution to these problems.

The immense technological developments this generation has witnessed are beyond the scope of what was imaginable by our scientific predecessors. These advancements contribute to the potential of AI becoming the next step in personalized medicine. AI is a comprehensive term that describes the replication of human knowledge by a computer, that computer begins to learn independently without human interference (2). The applications for AI in cancer diagnosis and treatment are vast, branching from helping to create personalized therapy plans that avoid the use of unnecessary painful treatments with low outcomes, to understanding why cancer cells develop resistance.

Improving Anticancer Drugs with AI

When looking at the development of anticancer drugs, AI can be beneficial for assessing the association between the genomic variability of the cancer cell and the drug activity. A successful prediction model called ‘Elastic Net Regression’ was developed by Wang et al to assess the drug sensitivity of different malignancies in patients with ovarian, endometrial and gastric cancer (6). This is just the tip of the iceberg for this research. Machine learning is rapidly evolving with models being made to predict mutations within the cancer cell genome. Furthermore, developments of a machine learning AI by Sun et al produced a platform that could successfully predict the efficacy of a therapeutic drug on programmed cell death protein 1 (PD-1). By predicting the immune phenotype of the tumors the clinical outcome of a drug could be observed (5). This AI approach was tailored to improve immunotherapy efficacy for patients who are sensitive to PD-1 inhibitors.

Improving Cancer Imaging and Radiotherapy with AI

AI use within radiology has a far more specific application. Target areas in the body can be mapped out with an AI system to devise the most effective treatment plan. The 3D convolutional neural network was used to automatically outline a nasopharyngeal tumor with a successful accuracy of 79%; this is on par with the work of a radiotherapy specialist (4). This may seem hard to believe as most medical professionals know that the best treatment for a patient is not only dependent on the science but also the patient’s mental health, treatment environment, and ability to tolerate harsh treatments. This is where AI meets its match: it lacks the human touch that is needed when dealing with patients–the ability to feel empathy.

Overcoming the Challenges of AI in Oncology

There are additional challenges that face AI in healthcare. Ethical implications must be considered, including the possibility of data bias and the worry of a breach of security and privacy. These issues may be resolved by ensuring accurate reporting of the population and advanced encryption. An area of ongoing development is building an acceptable framework of ethical principles surrounding the use of AI within a medical environment. This further brings up the issue of does AI work require supervision? Is it a trustworthy process that would make patients feel safe and secure?

As with any new ideas there are two sides to the story; is AI taking over the world and making human involvement irrelevant or simply aiding us in taking care of each other? Ethically there are concerns, but do the benefits outweigh the negatives? Is there a future where patients could be treated solely by robots? These questions will be answered as we adapt and inevitably incorporate AI into medical care. Currently, AI is simply used to aid healthcare workers to treat a patient to the best of their ability, with cancer imaging being its most prominent area of use. Despite the challenges that face AI use in healthcare, in the future, it has the potential to advance several sectors in oncology, including imaging and screening, as well as diagnosis and therapy.

Edited by Deanna MacNeil

References:

1.      Cho, S.-H., Jeon, J. and Kim, S.I. (2012). Personalized Medicine in Breast Cancer: A Systematic Review. Journal of Breast Cancer, 15(3), p.265. doi:10.4048/jbc.2012.15.3.265.

2.      Hunter, B., Hindocha, S. and Lee, R.W. (2022). The Role of Artificial Intelligence in Early Cancer Diagnosis. Cancers, 14(6), p.1524. doi:10.3390/cancers14061524.

3.  Li, Q., Qi, L., Feng, Q.-X., Liu, C., Sun, S.-W., Zhang, J., Yang, G., Ge, Y.-Q., Zhang, Y.-D. and Liu, X.-S. (2019). Machine Learning–Based Computational Models Derived From Large-Scale Radiographic-Radiomic Images Can Help Predict Adverse Histopathological Status of Gastric Cancer. Clinical and Translational Gastroenterology 10(10), p.e00079. doi:10.14309/ctg.0000000000000079.

4.  Radiology. (2021). Deep Learning for Automated Contouring of Primary Tumor Volumes by MRI for Nasopharyngeal Carcinoma. Available at: https://pubs.rsna.org/doi/10.1148/radiol.2019182012

5.  Sun, R., Limkin, E.J., Vakalopoulou, M., Dercle, L., Champiat, S., Han, S.R., Verlingue, L., Brandao, D., Lancia, A., Ammari, S., Hollebecque, A., Scoazec, J.-Y., Marabelle, A., Massard, C., Soria, J.-C., Robert, C., Paragios, N., Deutsch, E. and Ferté, C. (2018). A radiomics approach to assess tumour-infiltrating CD8 cells and response to anti-PD-1 or anti-PD-L1 immunotherapy: an imaging biomarker, retrospective multicohort study. The Lancet Oncology, 19(9), pp.1180–1191. doi:10.1016/s1470-2045(18)30413-3.

6.  Wang, Y., Garbin, B., Leo, F., Coen, S., Erkintalo, M. and Murdoch, S.G. (2018). Addressing temporal Kerr cavity solitons with a single pulse of intensity modulation. Optics Letters, 43(13), p.3192. doi:10.1364/ol.43.003192.

Image created in Biorender.com