Exploring what we know so far about artificial intelligence in healthcare; its upcoming trends as well as implications on the future of healthcare.
by Deborah Seah
“Artificial intelligence” (AI) a term first coined by one of its founding fathers, John McCarthy, during a conference in Dartmouth in 1956, is used to describe the mimicry of human intelligence by computer systems. By copying these abilities, these programs or machines would have the ability to make decisions, predict outcomes, and even recognize speech and facial features, much like how the human intelligence has the ability to do. Before being able to perform these tasks, these systems have to be mapped out and data has to be input in order for it to learn and generate information from the data.
With developments in AI, healthcare systems throughout the world has leveraged on its advancements to propel healthcare to be more efficient and effective for better patient outcomes. Some of the benefits of AI include, early diagnosis, prognosis, and treatment planning.
Based on statistics form the World Health Organization (WHO), leading causes of death within the Southeast Asian region include cardiovascular disease, chronic respiratory diseases, diabetes, and cancer. With an estimated 8.5 million deaths each year, this adds on to the increasing burden on healthcare resources. As such utilizing the advancements in AI will help to improve productivity in the healthcare system by assisting in decision making, data analysis, and reducing diagnostic and treatment errors.
In this issue we will discuss current research on AI in healthcare and how they contribute to improving the operations of healthcare systems. We would also look at the future of AI in healthcare through the evaluation of recently published research being done with different aspects of AI. Finally, we will consider the implications of this fast – growing field on the present landscape of the healthcare system.
AI Techniques and Devices Used in Healthcare Systems
The term AI encompasses a number of sub-categories. The two main sub – categories are machine learning, and natural language processing (NLP). Based on the type of data, the technique will be selected to best suit the aim of data analysis. There are two main types of healthcare data; structured and unstructured. Structured data can be extracted from electronic health records (EHR), while unstructured data is obtained from clinical notes of which are written in narrative text.
Structured data is analysed using machine learning, this method uses analytical algorithms to extract and analyse data from for example, patient baseline records in order to predict outcomes, and disease prognosis. A subset of machine learning known as deep learning is an expansion of the traditional neural network technique and is able to examine more complex patterns in data through more layers of neural networks.
Unstructured data however requires the translation of narrative text into a language understandable by the computer system. This is done through natural language processing by first processing the text, identifying clinical terms that are of interest to the analysis. Following which they are then classified and added to structured data that will be processed through machine learning algorithms.
These methods used by AI have shown to contribute to helping in clinical decision making, as well as disease diagnosis. The following section will discuss these contributions in further detail.
Current Research of AI in Healthcare
In a review by Fei et. al. (2017), it was found that the most common type of data found from literature on AI in healthcare is diagnostic imaging. As seen from the histogram in figure 1.
Research by Dreizin et. al. (2019), on medial AI has found that deep learning algorithms enable automated segmentation and quantification of traumatic pelvic hematomas on computed tomography (CT). The result of this research showed qualitatively that predicted labels followed the shape of hematomas and avoided muscle and displaced viscera. Even though pelvic hematomas are irregular in shape and have ill – defined margins. With the help of further development into this aspect of AI, the algorithm is able to assist in quantitating disease levels through size determination of pelvic hematomas.
Besides the use of AI in diagnostic imaging, molecular mechanisms of disease are becoming an area of interest. These molecular mechanisms can be analyzed from imaging techniques such as positron emission tomography (PET) and single photon emission computed tomography (SPECT). The complexity of the medical imaging data can be handled by AI algorithms to help us understand molecular mechanisms of disease. (Cook and Goh, 2019)
In a recent review article published by Liew et. al. (2019), they discussed the disruption of the radiology arena by AI in Singapore. It also highlighted the willingness of Singapore radiologist in being empowered by this new technology in order to achieve the objectives of delivering value – based and patient – centric care. The Radiological AI, Data Science and Imaging Informatics (RADII) was introduced in this review as a subsection of the Singapore Radiological Society. This is would allow key stakeholders to work together in developing and evaluation of AI in radiology.
Venturing Into the Future of Healthcare With AI
As AI will continue to support and shape the future of healthcare systems in developed countries, we must also then examine how these AI methods can be incorporated for developing countries or rural areas to benefit from better healthcare. In order to do so, AI systems should be drawn out to fit the requirements of healthcare in these developing countries. Guo and Li (2018), proposed a multilevel medical AI service network specifically to suit the healthcare infrastructure of rural areas of developing countries. Thereby, utilizing AI to improve efficiency of healthcare workers, making healthcare more accessible and increase quality. Wahl et. al. (2018), reviewed how AI can contribute to health in resource – poor settings. The ability of AI to help in clinical decision making and even coming up with treatments based on input patient data can support the programs and initiatives of healthcare organizations in improving medical resources in resource – poor settings.
Implications of Present Healthcare Landscape on Development of AI
No doubt with AI improvements and implementation into the healthcare system, there will be changes to its landscape in the years to come. One prominent topic of discussion is the replacement of healthcare jobs with the rise in the use of AI. But before we start being alarmed by this possibility, we have to analyse the situation and determine whether AI will take over even the jobs of human physicians.
In a study done by Li et. al. (2019), it was found that deep – learning based artificial intelligence was able to detect all malignancies from chest CT and showed a statistically significant higher sensitivity on pulmonary nodule detection.
Even though AI may prove to be more efficient than humans, quoting from Karches, 2018, AI should not replace the judgement of a human physician and that using AI may disrupt the physician – patient relationship. Fully relying on AI could possibly de – humanize healthcare to a rigid system of algorithms.
With all the promises that AI in healthcare may hold, we should always evaluate the challenges that may arise from the implementation of more AI systems. Increasingly as medical AI research looks to creating computer systems able to make clinical decisions, the question of legal liability would arise. (Loh, 2018) As with any system there will be flaws and one can never be certain that AI in healthcare would be infallible. Therefore, there is a need for governments to look into legal regulation of the use of medical AI when making clinical decisions.
As the ever – growing and exciting field of medical AI will inevitably continue to progress and be implemented to the current healthcare system, we would require the combined efforts of country leaders, healthcare professionals, physicians, and computer scientists alike for AI to achieve the aims that it was developed for. [APBN]
References
- Cook, Gary J. R. and Goh, Vicky (2019) What can artificial intelligence teach us about the molecular mechanisms underlying disease? European Journal of Nuclear Medicine and Molecular Imaging 1619-7089 https://doi.org/10.1007/s00259-019-04370-z
- Dreizin, David, Zhou, Yuyin, Zhang, Yixiao, Tirada, Nikki, and Yuille, Alan L. (2019) Performance of a Deep Learning Algorithm for Automated Segmentation and Quantification of Traumatic Pelvic Hematomas on CT, Journal of Digital Imaging, 1618-727X, https://doi.org/10.1007/s10278-019-00207-1
- Guo, J., & Li, B. (2018). The Application of Medical Artificial Intelligence Technology in Rural Areas of Developing Countries. Health equity, 2(1), 174–181. doi:10.1089/heq.2018.0037
- Jiang F, Jiang Y, Zhi H, Dong Y, Li H, Ma S, Wang Y, Dong Q, Shen H, Wang Y. (2017). Artificial intelligence in healthcare: past, present and future. Stroke and vascular neurology, 2(4), 230–243. doi:10.1136/svn-2017-000101
- Karches, Kyle E. (2018) Against the iDoctor: why artificial intelligence should not replace physician judgment, Theoretical Medicine and Bioethics, 1573-1200, https://doi.org/10.1007/s11017-018-9442-3
- Liew C.J., Krishnaswamy P., Cheng L.T., Tan C.H., Poh A.C., and Lim T.C. (2019) Artificial Intelligence and Radiology in Singapore: Championing a New Age of Augmented Imaging for Unsurpassed Patient Care. Ann Acad Med Singapore. 2019 Jan;48(1):16-24.
- LI, X., GUO, F., ZHOU, Z., ZHANG, F., WANG, Q., PENG, Z., SU, D., FAN, Y., & WANG, Y. (2019). Performance of Deep-learning-based Artificial Intelligence on Detection of Pulmonary Nodules in Chest CT. Chinese Journal Of Lung Cancer, 22(6), 336-340. doi:10.3779/10.3779/j.issn.1009-3419.2019.06.
- Loh E (2018) Medicine and the rise of the robots: a qualitative review of recent advances of artificial intelligence in health BMJ Leader 2018;2:59-63.
- Panch, T., Szolovits, P., & Atun, R. (2018). Artificial intelligence, machine learning and health systems. Journal of global health, 8(2), 020303. doi:10.7189/jogh.08.020303
- Shahid, N., Rappon, T., & Berta, W. (2019). Applications of artificial neural networks in health care organizational decision-making: A scoping review. PloS one, 14(2), e0212356. doi:10.1371/journal.pone.0212356
- Wahl, B., Cossy-Gantner, A., Germann, S., & Schwalbe, N. R. (2018). Artificial intelligence (AI) and global health: how can AI contribute to health in resource-poor settings?. BMJ global health, 3(4), e000798. doi:10.1136/bmjgh-2018-000798