By: Tanisha Pallerla, GPHS ‘27
For a long time, radiation oncology has solely relied on the human eye to capture small changes in a patient’s condition, but this practice is changing with the emergence of artificial intelligence (AI). While the human eye is effective at detecting these changes, AI is able to capture them to a degree that is beyond the scope of the human eye’s ability. When radiologists analyze a patient’s medical images, they use qualitative methods to diagnose patients and track changes in their condition. Due to the tedious nature of this task, it takes time for radiologists to return the results of a scan to a patient. However, AI turns medical image analysis into a quantitative task by using massive amounts of data to create incredible algorithms that can distinguish between health and disease, returning its results almost immediately [1]. This feature of AI makes it a powerful tool in improving the treatment and care of cancer patients.
One widely used type of AI that analyzes medical images in oncology is deep learning, which is a type of machine learning algorithm. Deep learning algorithms are trained by large sets of data, and due to the large quantity of data available online, these algorithms have become very advanced in their ability to replicate the human task of analyzing a cancer patient’s medical images [2]. With the improved accuracy and efficiency of deep learning algorithms, radiation oncologists are able to closely monitor a patient’s cancer progression. This includes the development of tumors and the spread of cancer cells. With the ability of deep learning algorithms to track even the tiniest changes in a patient's condition, oncologists are able to monitor how effective their treatments are and respond quickly [1].
The ability of AI to replicate this human task may seem too good to be true, and in some way it is. There are limitations to the extent and ability to which AI can contribute to a cancer patient’s treatment plan. While deep learning algorithms are able to detect small changes in a patient’s condition, they are not always 100% accurate, and they cannot actually make decisions that need to be made by trained radiation oncologists [3]. Deep learning algorithms are trained by data, and the amount and quality of data used to train these algorithms is important to consider [1]. The good news is that the implementation of AI models in clinical settings is regulated. Before an algorithm can be used in a clinical setting, it must undergo testing to ensure its accuracy and get FDA approval [1]. The degree to which it is regulated is an ongoing challenge in healthcare, but the work done by AI to analyze a patient’s images will always be checked by a trained and certified physician [4]. Radiation oncologists need to verify the results produced by AI, and they are the ones who actually create and adapt their patient’s treatment plan depending on the progression of the cancer as determined by the analysis of the patient's medical images.
While the impact AI has on improving cancer diagnosis and treatment depends on how advanced machine learning algorithms become in mimicking human intelligence, one thing is certain: no matter how advanced these algorithms get, they will never be able to replace the work that radiation oncologists do. AI analysis on cancer patients' images will always need to be checked by physicians to ensure there are no errors and to ensure no diagnostic information was missed [5]. Despite its limitations, AI is a powerful tool that enhances a radiation oncologist’s ability to do their job, and will continue to advance and improve cancer diagnosis and treatment.
References
Park, A., & Johnson, V. by A. D. (2022, November 4). How AI is Changing Medical Imaging. Time. https://time.com/6227623/ai-medical-imaging-radiology/
Hosny, A., Parmar, C., Quackenbush, J., Schwartz, L. H., & Aerts, H. J. W. L. (2018). Artificial intelligence in radiology. Nature reviews. Cancer, 18(8), 500–510. https://doi.org/10.1038/s41568-018-0016-5
Jarrett, D., Stride, E., Vallis, K., & Gooding, M. J. (2019). Applications and limitations of machine learning in radiation oncology. The British journal of radiology, 92(1100), 20190001. https://doi.org/10.1259/bjr.20190001
McKee, M., & Wouters, O. J. (2023). The Challenges of Regulating Artificial Intelligence in Healthcare Comment on "Clinical Decision Support and New Regulatory Frameworks for Medical Devices: Are We Ready for It? - A Viewpoint Paper". International journal of health policy and management, 12, 7261. https://doi.org/10.34172/ijhpm.2022.7261
1, M., & Lubell, J. (2023, May 1). As health care AI advances rapidly, what role for regulators?. American Medical Association. https://www.ama-assn.org/practice-management/digital/health-care-ai-advances-rapidly-what-role-regulators
Commenti